sfrench/cifs-2.6.git
5 years agoMerge tag 'mac80211-for-davem-2018-09-03' of git://git.kernel.org/pub/scm/linux/kerne...
David S. Miller [Tue, 4 Sep 2018 05:12:02 +0000 (22:12 -0700)]
Merge tag 'mac80211-for-davem-2018-09-03' of git://git./linux/kernel/git/jberg/mac80211

Johannes Berg says:

====================
Here are quite a large number of fixes, notably:
 * various A-MSDU building fixes (currently only affects mt76)
 * syzkaller & spectre fixes in hwsim
 * TXQ vs. teardown fix that was causing crashes
 * embed WMM info in reg rule, bad code here had been causing crashes
 * one compilation issue with fix from Arnd (rfkill-gpio includes)
 * fixes for a race and bad data during/after channel switch
 * nl80211: a validation fix, attribute type & unit fixes
along with other small fixes.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agotipc: correct spelling errors for tipc_topsrv_queue_evt() comments
Zhenbo Gao [Mon, 3 Sep 2018 08:36:46 +0000 (16:36 +0800)]
tipc: correct spelling errors for tipc_topsrv_queue_evt() comments

tipc_conn_queue_evt -> tipc_topsrv_queue_evt
tipc_send_work -> tipc_conn_send_work
tipc_send_to_sock -> tipc_conn_send_to_sock

Signed-off-by: Zhenbo Gao <zhenbo.gao@windriver.com>
Reviewed-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agotipc: correct spelling errors for struct tipc_bc_base's comment
Zhenbo Gao [Mon, 3 Sep 2018 08:36:45 +0000 (16:36 +0800)]
tipc: correct spelling errors for struct tipc_bc_base's comment

Trivial fix for two spelling mistakes.

Signed-off-by: Zhenbo Gao <zhenbo.gao@windriver.com>
Reviewed-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'bnxt_en-Bug-fixes'
David S. Miller [Tue, 4 Sep 2018 04:59:43 +0000 (21:59 -0700)]
Merge branch 'bnxt_en-Bug-fixes'

Michael Chan says:

====================
bnxt_en: Bug fixes.

This short series fixes resource related logic in the driver, mostly
affecting the RDMA driver under corner cases.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Do not adjust max_cp_rings by the ones used by RDMA.
Michael Chan [Mon, 3 Sep 2018 08:23:19 +0000 (04:23 -0400)]
bnxt_en: Do not adjust max_cp_rings by the ones used by RDMA.

Currently, the driver adjusts the bp->hw_resc.max_cp_rings by the number
of MSIX vectors used by RDMA.  There is one code path in open that needs
to check the true max_cp_rings including any used by RDMA.  This code
is now checking for the reduced max_cp_rings which will fail when the
number of cp rings is very small.

To fix this in a clean way, we don't adjust max_cp_rings anymore.
Instead, we add a helper bnxt_get_max_func_cp_rings_for_en() to get the
reduced max_cp_rings when appropriate.

Fixes: ec86f14ea506 ("bnxt_en: Add ULP calls to stop and restart IRQs.")
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Clean up unused functions.
Michael Chan [Mon, 3 Sep 2018 08:23:18 +0000 (04:23 -0400)]
bnxt_en: Clean up unused functions.

Remove unused bnxt_subtract_ulp_resources().  Change
bnxt_get_max_func_irqs() to static since it is only locally used.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Fix firmware signaled resource change logic in open.
Michael Chan [Mon, 3 Sep 2018 08:23:17 +0000 (04:23 -0400)]
bnxt_en: Fix firmware signaled resource change logic in open.

When the driver detects that resources have changed during open, it
should reset the rx and tx rings to 0.  This will properly setup the
init sequence to initialize the default rings again.  We also need
to signal the RDMA driver to stop and clear its interrupts.  We then
call the RoCE driver to restart if a new set of default rings is
successfully reserved.

Fixes: 25e1acd6b92b ("bnxt_en: Notify firmware about IF state changes.")
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'sctp-two-fixes-for-spp_ipv6_flowlabel-and-spp_dscp-sockopts'
David S. Miller [Tue, 4 Sep 2018 04:57:55 +0000 (21:57 -0700)]
Merge branch 'sctp-two-fixes-for-spp_ipv6_flowlabel-and-spp_dscp-sockopts'

Xin Long says:

====================
sctp: two fixes for spp_ipv6_flowlabel and spp_dscp sockopts

This patchset fixes two problems in sctp_apply_peer_addr_params()
when setting spp_ipv6_flowlabel or spp_dscp.
====================

Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agosctp: not traverse asoc trans list if non-ipv6 trans exists for ipv6_flowlabel
Xin Long [Mon, 3 Sep 2018 07:47:11 +0000 (15:47 +0800)]
sctp: not traverse asoc trans list if non-ipv6 trans exists for ipv6_flowlabel

When users set params.spp_address and get a trans, ipv6_flowlabel flag
should be applied into this trans. But even if this one is not an ipv6
trans, it should not go to apply it into all other transes of the asoc
but simply ignore it.

Fixes: 0b0dce7a36fb ("sctp: add spp_ipv6_flowlabel and spp_dscp for sctp_paddrparams")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agosctp: fix invalid reference to the index variable of the iterator
Xin Long [Mon, 3 Sep 2018 07:47:10 +0000 (15:47 +0800)]
sctp: fix invalid reference to the index variable of the iterator

Now in sctp_apply_peer_addr_params(), if SPP_IPV6_FLOWLABEL flag is set
and trans is NULL, it would use trans as the index variable to traverse
transport_addr_list, then trans is set as the last transport of it.

Later, if SPP_DSCP flag is set, it would enter into the wrong branch as
trans is actually an invalid reference.

So fix it by using a new index variable to traverse transport_addr_list
for both SPP_DSCP and SPP_IPV6_FLOWLABEL flags process.

Fixes: 0b0dce7a36fb ("sctp: add spp_ipv6_flowlabel and spp_dscp for sctp_paddrparams")
Reported-by: Julia Lawall <julia.lawall@lip6.fr>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/ibm/emac: wrong emac_calc_base call was used by typo
Ivan Mikhaylov [Mon, 3 Sep 2018 07:26:28 +0000 (10:26 +0300)]
net/ibm/emac: wrong emac_calc_base call was used by typo

__emac_calc_base_mr1 was used instead of __emac4_calc_base_mr1
by copy-paste mistake for emac4syn.

Fixes: 45d6e545505fd32edb812f085be7de45b6a5c0af ("net/ibm/emac: add 8192 rx/tx fifo size")
Signed-off-by: Ivan Mikhaylov <ivan@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: sched: null actions array pointer before releasing action
Vlad Buslov [Mon, 3 Sep 2018 07:04:55 +0000 (10:04 +0300)]
net: sched: null actions array pointer before releasing action

Currently, tcf_action_delete() nulls actions array pointer after putting
and deleting it. However, if tcf_idr_delete_index() returns an error,
pointer to action is not set to null. That results it being released second
time in error handling code of tca_action_gd().

Kasan error:

[  807.367755] ==================================================================
[  807.375844] BUG: KASAN: use-after-free in tc_setup_cb_call+0x14e/0x250
[  807.382763] Read of size 8 at addr ffff88033e636000 by task tc/2732

[  807.391289] CPU: 0 PID: 2732 Comm: tc Tainted: G        W         4.19.0-rc1+ #799
[  807.399542] Hardware name: Supermicro SYS-2028TP-DECR/X10DRT-P, BIOS 2.0b 03/30/2017
[  807.407948] Call Trace:
[  807.410763]  dump_stack+0x92/0xeb
[  807.414456]  print_address_description+0x70/0x360
[  807.419549]  kasan_report+0x14d/0x300
[  807.423582]  ? tc_setup_cb_call+0x14e/0x250
[  807.428150]  tc_setup_cb_call+0x14e/0x250
[  807.432539]  ? nla_put+0x65/0xe0
[  807.436146]  fl_dump+0x394/0x3f0 [cls_flower]
[  807.440890]  ? fl_tmplt_dump+0x140/0x140 [cls_flower]
[  807.446327]  ? lock_downgrade+0x320/0x320
[  807.450702]  ? lock_acquire+0xe2/0x220
[  807.454819]  ? is_bpf_text_address+0x5/0x140
[  807.459475]  ? memcpy+0x34/0x50
[  807.462980]  ? nla_put+0x65/0xe0
[  807.466582]  tcf_fill_node+0x341/0x430
[  807.470717]  ? tcf_block_put+0xe0/0xe0
[  807.474859]  tcf_node_dump+0xdb/0xf0
[  807.478821]  fl_walk+0x8e/0x170 [cls_flower]
[  807.483474]  tcf_chain_dump+0x35a/0x4d0
[  807.487703]  ? tfilter_notify+0x170/0x170
[  807.492091]  ? tcf_fill_node+0x430/0x430
[  807.496411]  tc_dump_tfilter+0x362/0x3f0
[  807.500712]  ? tc_del_tfilter+0x850/0x850
[  807.505104]  ? kasan_unpoison_shadow+0x30/0x40
[  807.509940]  ? __mutex_unlock_slowpath+0xcf/0x410
[  807.515031]  netlink_dump+0x263/0x4f0
[  807.519077]  __netlink_dump_start+0x2a0/0x300
[  807.523817]  ? tc_del_tfilter+0x850/0x850
[  807.528198]  rtnetlink_rcv_msg+0x46a/0x6d0
[  807.532671]  ? rtnl_fdb_del+0x3f0/0x3f0
[  807.536878]  ? tc_del_tfilter+0x850/0x850
[  807.541280]  netlink_rcv_skb+0x18d/0x200
[  807.545570]  ? rtnl_fdb_del+0x3f0/0x3f0
[  807.549773]  ? netlink_ack+0x500/0x500
[  807.553913]  netlink_unicast+0x2d0/0x370
[  807.558212]  ? netlink_attachskb+0x340/0x340
[  807.562855]  ? _copy_from_iter_full+0xe9/0x3e0
[  807.567677]  ? import_iovec+0x11e/0x1c0
[  807.571890]  netlink_sendmsg+0x3b9/0x6a0
[  807.576192]  ? netlink_unicast+0x370/0x370
[  807.580684]  ? netlink_unicast+0x370/0x370
[  807.585154]  sock_sendmsg+0x6b/0x80
[  807.589015]  ___sys_sendmsg+0x4a1/0x520
[  807.593230]  ? copy_msghdr_from_user+0x210/0x210
[  807.598232]  ? do_wp_page+0x174/0x880
[  807.602276]  ? __handle_mm_fault+0x749/0x1c10
[  807.607021]  ? __handle_mm_fault+0x1046/0x1c10
[  807.611849]  ? __pmd_alloc+0x320/0x320
[  807.615973]  ? check_chain_key+0x140/0x1f0
[  807.620450]  ? check_chain_key+0x140/0x1f0
[  807.624929]  ? __fget_light+0xbc/0xd0
[  807.628970]  ? __sys_sendmsg+0xd7/0x150
[  807.633172]  __sys_sendmsg+0xd7/0x150
[  807.637201]  ? __ia32_sys_shutdown+0x30/0x30
[  807.641846]  ? up_read+0x53/0x90
[  807.645442]  ? __do_page_fault+0x484/0x780
[  807.649949]  ? do_syscall_64+0x1e/0x2c0
[  807.654164]  do_syscall_64+0x72/0x2c0
[  807.658198]  entry_SYSCALL_64_after_hwframe+0x49/0xbe
[  807.663625] RIP: 0033:0x7f42e9870150
[  807.667568] Code: 8b 15 3c 7d 2b 00 f7 d8 64 89 02 48 c7 c0 ff ff ff ff eb cd 66 0f 1f 44 00 00 83 3d b9 d5 2b 00 00 75 10 b8 2e 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 31 c3 48 83 ec 08 e8 be cd 00 00 48 89 04 24
[  807.687328] RSP: 002b:00007ffdbf595b58 EFLAGS: 00000246 ORIG_RAX: 000000000000002e
[  807.695564] RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007f42e9870150
[  807.703083] RDX: 0000000000000000 RSI: 00007ffdbf595b80 RDI: 0000000000000003
[  807.710605] RBP: 00007ffdbf599d90 R08: 0000000000679bc0 R09: 000000000000000f
[  807.718127] R10: 00000000000005e7 R11: 0000000000000246 R12: 00007ffdbf599d88
[  807.725651] R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000

[  807.735048] Allocated by task 2687:
[  807.738902]  kasan_kmalloc+0xa0/0xd0
[  807.742852]  __kmalloc+0x118/0x2d0
[  807.746615]  tcf_idr_create+0x44/0x320
[  807.750738]  tcf_nat_init+0x41e/0x530 [act_nat]
[  807.755638]  tcf_action_init_1+0x4e0/0x650
[  807.760104]  tcf_action_init+0x1ce/0x2d0
[  807.764395]  tcf_exts_validate+0x1d8/0x200
[  807.768861]  fl_change+0x55a/0x26b4 [cls_flower]
[  807.773845]  tc_new_tfilter+0x748/0xa20
[  807.778051]  rtnetlink_rcv_msg+0x56a/0x6d0
[  807.782517]  netlink_rcv_skb+0x18d/0x200
[  807.786804]  netlink_unicast+0x2d0/0x370
[  807.791095]  netlink_sendmsg+0x3b9/0x6a0
[  807.795387]  sock_sendmsg+0x6b/0x80
[  807.799240]  ___sys_sendmsg+0x4a1/0x520
[  807.803445]  __sys_sendmsg+0xd7/0x150
[  807.807473]  do_syscall_64+0x72/0x2c0
[  807.811506]  entry_SYSCALL_64_after_hwframe+0x49/0xbe

[  807.818776] Freed by task 2728:
[  807.822283]  __kasan_slab_free+0x122/0x180
[  807.826752]  kfree+0xf4/0x2f0
[  807.830080]  __tcf_action_put+0x5a/0xb0
[  807.834281]  tcf_action_put_many+0x46/0x70
[  807.838747]  tca_action_gd+0x232/0xc40
[  807.842862]  tc_ctl_action+0x215/0x230
[  807.846977]  rtnetlink_rcv_msg+0x56a/0x6d0
[  807.851444]  netlink_rcv_skb+0x18d/0x200
[  807.855731]  netlink_unicast+0x2d0/0x370
[  807.860021]  netlink_sendmsg+0x3b9/0x6a0
[  807.864312]  sock_sendmsg+0x6b/0x80
[  807.868166]  ___sys_sendmsg+0x4a1/0x520
[  807.872372]  __sys_sendmsg+0xd7/0x150
[  807.876401]  do_syscall_64+0x72/0x2c0
[  807.880431]  entry_SYSCALL_64_after_hwframe+0x49/0xbe

[  807.887704] The buggy address belongs to the object at ffff88033e636000
                which belongs to the cache kmalloc-256 of size 256
[  807.900909] The buggy address is located 0 bytes inside of
                256-byte region [ffff88033e636000ffff88033e636100)
[  807.913155] The buggy address belongs to the page:
[  807.918322] page:ffffea000cf98d80 count:1 mapcount:0 mapping:ffff88036f80ee00 index:0x0 compound_mapcount: 0
[  807.928831] flags: 0x5fff8000008100(slab|head)
[  807.933647] raw: 005fff8000008100 ffffea000db44f00 0000000400000004 ffff88036f80ee00
[  807.942050] raw: 0000000000000000 0000000080190019 00000001ffffffff 0000000000000000
[  807.950456] page dumped because: kasan: bad access detected

[  807.958240] Memory state around the buggy address:
[  807.963405]  ffff88033e635f00: fc fc fc fc fb fb fb fb fb fb fb fc fc fc fc fb
[  807.971288]  ffff88033e635f80: fb fb fb fb fb fb fc fc fc fc fc fc fc fc fc fc
[  807.979166] >ffff88033e636000: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[  807.994882]                    ^
[  807.998477]  ffff88033e636080: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[  808.006352]  ffff88033e636100: fc fc fc fc fc fc fc fc fb fb fb fb fb fb fb fb
[  808.014230] ==================================================================
[  808.022108] Disabling lock debugging due to kernel taint

Fixes: edfaf94fa705 ("net_sched: improve and refactor tcf_action_put_many()")
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Acked-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agovhost: fix VHOST_GET_BACKEND_FEATURES ioctl request definition
Gleb Fotengauer-Malinovskiy [Mon, 3 Sep 2018 17:59:13 +0000 (20:59 +0300)]
vhost: fix VHOST_GET_BACKEND_FEATURES ioctl request definition

The _IOC_READ flag fits this ioctl request more because this request
actually only writes to, but doesn't read from userspace.
See NOTEs in include/uapi/asm-generic/ioctl.h for more information.

Fixes: 429711aec282 ("vhost: switch to use new message format")
Signed-off-by: Gleb Fotengauer-Malinovskiy <glebfm@altlinux.org>
Acked-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agor8169: add support for NCube 8168 network card
Anthony Wong [Fri, 31 Aug 2018 12:06:42 +0000 (20:06 +0800)]
r8169: add support for NCube 8168 network card

This card identifies itself as:
  Ethernet controller [0200]: NCube Device [10ff:8168] (rev 06)
  Subsystem: TP-LINK Technologies Co., Ltd. Device [7470:3468]

Adding a new entry to rtl8169_pci_tbl makes the card work.

Link: http://launchpad.net/bugs/1788730
Signed-off-by: Anthony Wong <anthony.wong@ubuntu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoip6_tunnel: respect ttl inherit for ip6tnl
Hangbin Liu [Fri, 31 Aug 2018 08:52:01 +0000 (16:52 +0800)]
ip6_tunnel: respect ttl inherit for ip6tnl

man ip-tunnel ttl section says:
0 is a special value meaning that packets inherit the TTL value.

IPv4 tunnel respect this in ip_tunnel_xmit(), but IPv6 tunnel has not
implement it yet. To make IPv6 behave consistently with IP tunnel,
add ipv6 tunnel inherit support.

Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomac80211: shorten the IBSS debug messages
Emmanuel Grumbach [Fri, 31 Aug 2018 08:31:13 +0000 (11:31 +0300)]
mac80211: shorten the IBSS debug messages

When tracing is enabled, all the debug messages are recorded and must
not exceed MAX_MSG_LEN (100) columns. Longer debug messages grant the
user with:

WARNING: CPU: 3 PID: 32642 at /tmp/wifi-core-20180806094828/src/iwlwifi-stack-dev/net/mac80211/./trace_msg.h:32 trace_event_raw_event_mac80211_msg_event+0xab/0xc0 [mac80211]
Workqueue: phy1 ieee80211_iface_work [mac80211]
 RIP: 0010:trace_event_raw_event_mac80211_msg_event+0xab/0xc0 [mac80211]
 Call Trace:
  __sdata_dbg+0xbd/0x120 [mac80211]
  ieee80211_ibss_rx_queued_mgmt+0x15f/0x510 [mac80211]
  ieee80211_iface_work+0x21d/0x320 [mac80211]

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: don't Tx a deauth frame if the AP forbade Tx
Emmanuel Grumbach [Fri, 31 Aug 2018 08:31:12 +0000 (11:31 +0300)]
mac80211: don't Tx a deauth frame if the AP forbade Tx

If the driver fails to properly prepare for the channel
switch, mac80211 will disconnect. If the CSA IE had mode
set to 1, it means that the clients are not allowed to send
any Tx on the current channel, and that includes the
deauthentication frame.

Make sure that we don't send the deauthentication frame in
this case.

In iwlwifi, this caused a failure to flush queues since the
firmware already closed the queues after having parsed the
CSA IE. Then mac80211 would wait until the deauthentication
frame would go out (drv_flush(drop=false)) and that would
never happen.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: Fix station bandwidth setting after channel switch
Ilan Peer [Fri, 31 Aug 2018 08:31:10 +0000 (11:31 +0300)]
mac80211: Fix station bandwidth setting after channel switch

When performing a channel switch flow for a managed interface, the
flow did not update the bandwidth of the AP station and the rate
scale algorithm. In case of a channel width downgrade, this would
result with the rate scale algorithm using a bandwidth that does not
match the interface channel configuration.

Fix this by updating the AP station bandwidth and rate scaling algorithm
before the actual channel change in case of a bandwidth downgrade, or
after the actual channel change in case of a bandwidth upgrade.

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: fix a race between restart and CSA flows
Emmanuel Grumbach [Fri, 31 Aug 2018 08:31:06 +0000 (11:31 +0300)]
mac80211: fix a race between restart and CSA flows

We hit a problem with iwlwifi that was caused by a bug in
mac80211. A bug in iwlwifi caused the firwmare to crash in
certain cases in channel switch. Because of that bug,
drv_pre_channel_switch would fail and trigger the restart
flow.
Now we had the hw restart worker which runs on the system's
workqueue and the csa_connection_drop_work worker that runs
on mac80211's workqueue that can run together. This is
obviously problematic since the restart work wants to
reconfigure the connection, while the csa_connection_drop_work
worker does the exact opposite: it tries to disconnect.

Fix this by cancelling the csa_connection_drop_work worker
in the restart worker.

Note that this can sound racy: we could have:

driver   iface_work   CSA_work   restart_work
+++++++++++++++++++++++++++++++++++++++++++++
              |
 <--drv_cs ---|
<FW CRASH!>
-CS FAILED-->
              |                       |
              |                 cancel_work(CSA)
           schedule                   |
           CSA work                   |
                         |            |
                        Race between those 2

But this is not possible because we flush the workqueue
in the restart worker before we cancel the CSA worker.
That would be bullet proof if we could guarantee that
we schedule the CSA worker only from the iface_work
which runs on the workqueue (and not on the system's
workqueue), but unfortunately we do have an instance
in which we schedule the CSA work outside the context
of the workqueue (ieee80211_chswitch_done).

Note also that we should probably cancel other workers
like beacon_connection_loss_work and possibly others
for different types of interfaces, at the very least,
IBSS should suffer from the exact same problem, but for
now, do the minimum to fix the actual bug that was actually
experienced and reproduced.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: fix WMM TXOP calculation
Dreyfuss, Haim [Fri, 31 Aug 2018 08:31:04 +0000 (11:31 +0300)]
mac80211: fix WMM TXOP calculation

In commit 9236c4523e5b ("mac80211: limit wmm params to comply
with ETSI requirements"), we have limited the WMM parameters to
comply with 802.11 and ETSI standard.  Mistakenly the TXOP value
was caluclated wrong.  Fix it by taking the minimum between
802.11 to ETSI to make sure we are not violating both.

Fixes: e552af058148 ("mac80211: limit wmm params to comply with ETSI requirements")
Signed-off-by: Haim Dreyfuss <haim.dreyfuss@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agocfg80211: fix a type issue in ieee80211_chandef_to_operating_class()
Dan Carpenter [Fri, 31 Aug 2018 08:10:55 +0000 (11:10 +0300)]
cfg80211: fix a type issue in ieee80211_chandef_to_operating_class()

The "chandef->center_freq1" variable is a u32 but "freq" is a u16 so we
are truncating away the high bits.  I noticed this bug because in commit
9cf0a0b4b64a ("cfg80211: Add support for 60GHz band channels 5 and 6")
we made "freq <= 56160 + 2160 * 6" a valid requency when before it was
only "freq <= 56160 + 2160 * 4" that was valid.  It introduces a static
checker warning:

    net/wireless/util.c:1571 ieee80211_chandef_to_operating_class()
    warn: always true condition '(freq <= 56160 + 2160 * 6) => (0-u16max <= 69120)'

But really we probably shouldn't have been truncating the high bits
away to begin with.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: fix an off-by-one issue in A-MSDU max_subframe computation
Lorenzo Bianconi [Thu, 30 Aug 2018 23:04:13 +0000 (01:04 +0200)]
mac80211: fix an off-by-one issue in A-MSDU max_subframe computation

Initialize 'n' to 2 in order to take into account also the first
packet in the estimation of max_subframe limit for a given A-MSDU
since frag_tail pointer is NULL when ieee80211_amsdu_aggregate
routine analyzes the second frame.

Fixes: 6e0456b54545 ("mac80211: add A-MSDU tx support")
Signed-off-by: Lorenzo Bianconi <lorenzo.bianconi@redhat.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agouapi: Fix linux/rds.h userspace compilation errors.
Vinson Lee [Sat, 1 Sep 2018 21:20:27 +0000 (21:20 +0000)]
uapi: Fix linux/rds.h userspace compilation errors.

Include linux/in6.h for struct in6_addr.

/usr/include/linux/rds.h:156:18: error: field â€˜laddr’ has incomplete type
  struct in6_addr laddr;
                  ^~~~~
/usr/include/linux/rds.h:157:18: error: field â€˜faddr’ has incomplete type
  struct in6_addr faddr;
                  ^~~~~
/usr/include/linux/rds.h:178:18: error: field â€˜laddr’ has incomplete type
  struct in6_addr laddr;
                  ^~~~~
/usr/include/linux/rds.h:179:18: error: field â€˜faddr’ has incomplete type
  struct in6_addr faddr;
                  ^~~~~
/usr/include/linux/rds.h:198:18: error: field â€˜bound_addr’ has incomplete type
  struct in6_addr bound_addr;
                  ^~~~~~~~~~
/usr/include/linux/rds.h:199:18: error: field â€˜connected_addr’ has incomplete type
  struct in6_addr connected_addr;
                  ^~~~~~~~~~~~~~
/usr/include/linux/rds.h:219:18: error: field â€˜local_addr’ has incomplete type
  struct in6_addr local_addr;
                  ^~~~~~~~~~
/usr/include/linux/rds.h:221:18: error: field â€˜peer_addr’ has incomplete type
  struct in6_addr peer_addr;
                  ^~~~~~~~~
/usr/include/linux/rds.h:245:18: error: field â€˜src_addr’ has incomplete type
  struct in6_addr src_addr;
                  ^~~~~~~~
/usr/include/linux/rds.h:246:18: error: field â€˜dst_addr’ has incomplete type
  struct in6_addr dst_addr;
                  ^~~~~~~~

Fixes: b7ff8b1036f0 ("rds: Extend RDS API for IPv6 support")
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: cadence: Fix a sleep-in-atomic-context bug in macb_halt_tx()
Jia-Ju Bai [Sat, 1 Sep 2018 12:11:05 +0000 (20:11 +0800)]
net: cadence: Fix a sleep-in-atomic-context bug in macb_halt_tx()

The kernel module may sleep with holding a spinlock.

The function call paths (from bottom to top) in Linux-4.16 are:

[FUNC] usleep_range
drivers/net/ethernet/cadence/macb_main.c, 648:
usleep_range in macb_halt_tx
drivers/net/ethernet/cadence/macb_main.c, 730:
macb_halt_tx in macb_tx_error_task
drivers/net/ethernet/cadence/macb_main.c, 721:
_raw_spin_lock_irqsave in macb_tx_error_task

To fix this bug, usleep_range() is replaced with udelay().

This bug is found by my static analysis tool DSAC.

Signed-off-by: Jia-Ju Bai <baijiaju1990@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf
David S. Miller [Sun, 2 Sep 2018 22:53:38 +0000 (15:53 -0700)]
Merge git://git./pub/scm/linux/kernel/git/bpf/bpf

Daniel Borkmann says:

====================
pull-request: bpf 2018-09-02

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

The main changes are:

1) Fix one remaining buggy offset override in sockmap's bpf_msg_pull_data()
   when linearizing multiple scatterlist elements, from Tushar.

2) Fix BPF sockmap's misuse of ULP when a collision with another ULP is
   found on map update where it would release existing ULP. syzbot found and
   triggered this couple of times now, fix from John.

3) Add missing xskmap type to bpftool so it will properly show the type
   on map dump, from Prashant.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/ipv6: Only update MTU metric if it set
David Ahern [Thu, 30 Aug 2018 21:15:43 +0000 (14:15 -0700)]
net/ipv6: Only update MTU metric if it set

Jan reported a regression after an update to 4.18.5. In this case ipv6
default route is setup by systemd-networkd based on data from an RA. The
RA contains an MTU of 1492 which is used when the route is first inserted
but then systemd-networkd pushes down updates to the default route
without the mtu set.

Prior to the change to fib6_info, metrics such as MTU were held in the
dst_entry and rt6i_pmtu in rt6_info contained an update to the mtu if
any. ip6_mtu would look at rt6i_pmtu first and use it if set. If not,
the value from the metrics is used if it is set and finally falling
back to the idev value.

After the fib6_info change metrics are contained in the fib6_info struct
and there is no equivalent to rt6i_pmtu. To maintain consistency with
the old behavior the new code should only reset the MTU in the metrics
if the route update has it set.

Fixes: d4ead6b34b67 ("net/ipv6: move metrics from dst to rt6_info")
Reported-by: Jan Janssen <medhefgo@web.de>
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ethernet: cpsw-phy-sel: prefer phandle for phy sel
Tony Lindgren [Wed, 29 Aug 2018 15:00:24 +0000 (08:00 -0700)]
net: ethernet: cpsw-phy-sel: prefer phandle for phy sel

The cpsw-phy-sel device is not a child of the cpsw interconnect target
module. It lives in the system control module.

Let's fix this issue by trying to use cpsw-phy-sel phandle first if it
exists and if not fall back to current usage of trying to find the
cpsw-phy-sel child. That way the phy sel driver can be a child of the
system control module where it belongs in the device tree.

Without this fix, we cannot have a proper interconnect target module
hierarchy in device tree for things like genpd.

Note that deferred probe is mostly not supported by cpsw and this patch
does not attempt to fix that. In case deferred probe support is needed,
this could be added to cpsw_slave_open() and phy_connect() so they start
handling and returning errors.

For documenting it, looks like the cpsw-phy-sel is used for all cpsw device
tree nodes. It's missing the related binding documentation, so let's also
update the binding documentation accordingly.

Cc: devicetree@vger.kernel.org
Cc: Andrew Lunn <andrew@lunn.ch>
Cc: Grygorii Strashko <grygorii.strashko@ti.com>
Cc: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Murali Karicheri <m-karicheri2@ti.com>
Cc: Rob Herring <robh+dt@kernel.org>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodt-bindings: net: cpsw: Document cpsw-phy-sel usage but prefer phandle
Tony Lindgren [Wed, 29 Aug 2018 15:00:23 +0000 (08:00 -0700)]
dt-bindings: net: cpsw: Document cpsw-phy-sel usage but prefer phandle

The current cpsw usage for cpsw-phy-sel is undocumented but is used for
all the boards using cpsw. And cpsw-phy-sel is not really a child of
the cpsw device, it lives in the system control module instead.

Let's document the existing usage, and improve it a bit where we prefer
to use a phandle instead of a child device for it. That way we can
properly describe the hardware in dts files for things like genpd.

Cc: devicetree@vger.kernel.org
Cc: Andrew Lunn <andrew@lunn.ch>
Cc: Grygorii Strashko <grygorii.strashko@ti.com>
Cc: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Murali Karicheri <m-karicheri2@ti.com>
Cc: Rob Herring <robh+dt@kernel.org>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'igmp-fix-two-incorrect-unsolicit-report-count-issues'
David S. Miller [Sun, 2 Sep 2018 20:39:37 +0000 (13:39 -0700)]
Merge branch 'igmp-fix-two-incorrect-unsolicit-report-count-issues'

Hangbin Liu says:

====================
igmp: fix two incorrect unsolicit report count issues

Just like the subject, fix two minor igmp unsolicit report count issues.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoigmp: fix incorrect unsolicit report count after link down and up
Hangbin Liu [Wed, 29 Aug 2018 10:06:10 +0000 (18:06 +0800)]
igmp: fix incorrect unsolicit report count after link down and up

After link down and up, i.e. when call ip_mc_up(), we doesn't init
im->unsolicit_count. So after igmp_timer_expire(), we will not start
timer again and only send one unsolicit report at last.

Fix it by initializing im->unsolicit_count in igmp_group_added(), so
we can respect igmp robustness value.

Fixes: 24803f38a5c0b ("igmp: do not remove igmp souce list info when set link down")
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoigmp: fix incorrect unsolicit report count when join group
Hangbin Liu [Wed, 29 Aug 2018 10:06:08 +0000 (18:06 +0800)]
igmp: fix incorrect unsolicit report count when join group

We should not start timer if im->unsolicit_count equal to 0 after decrease.
Or we will send one more unsolicit report message. i.e. 3 instead of 2 by
default.

Fixes: 1da177e4c3f41 ("Linux-2.6.12-rc2")
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobpf: avoid misuse of psock when TCP_ULP_BPF collides with another ULP
John Fastabend [Fri, 31 Aug 2018 04:25:02 +0000 (21:25 -0700)]
bpf: avoid misuse of psock when TCP_ULP_BPF collides with another ULP

Currently we check sk_user_data is non NULL to determine if the sk
exists in a map. However, this is not sufficient to ensure the psock
or the ULP ops are not in use by another user, such as kcm or TLS. To
avoid this when adding a sock to a map also verify it is of the
correct ULP type. Additionally, when releasing a psock verify that
it is the TCP_ULP_BPF type before releasing the ULP. The error case
where we abort an update due to ULP collision can cause this error
path.

For example,

  __sock_map_ctx_update_elem()
     [...]
     err = tcp_set_ulp_id(sock, TCP_ULP_BPF) <- collides with TLS
     if (err)                                <- so err out here
        goto out_free
     [...]
  out_free:
     smap_release_sock() <- calling tcp_cleanup_ulp releases the
                            TLS ULP incorrectly.

Fixes: 2f857d04601a ("bpf: sockmap, remove STRPARSER map_flags and add multi-map support")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
5 years agotools/bpf: bpftool, add xskmap in map types
Prashant Bhole [Fri, 31 Aug 2018 06:32:42 +0000 (15:32 +0900)]
tools/bpf: bpftool, add xskmap in map types

When listed all maps, bpftool currently shows (null) for xskmap.
Added xskmap type in map_type_name[] to show correct type.

Signed-off-by: Prashant Bhole <bhole_prashant_q7@lab.ntt.co.jp>
Acked-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
5 years agobpf: Fix bpf_msg_pull_data()
Tushar Dave [Fri, 31 Aug 2018 21:45:16 +0000 (23:45 +0200)]
bpf: Fix bpf_msg_pull_data()

Helper bpf_msg_pull_data() mistakenly reuses variable 'offset' while
linearizing multiple scatterlist elements. Variable 'offset' is used
to find first starting scatterlist element
    i.e. msg->data = sg_virt(&sg[first_sg]) + start - offset"

Use different variable name while linearizing multiple scatterlist
elements so that value contained in variable 'offset' won't get
overwritten.

Fixes: 015632bb30da ("bpf: sk_msg program helper bpf_sk_msg_pull_data")
Signed-off-by: Tushar Dave <tushar.n.dave@oracle.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
5 years agoipv6: don't get lwtstate twice in ip6_rt_copy_init()
Alexey Kodanev [Thu, 30 Aug 2018 16:11:24 +0000 (19:11 +0300)]
ipv6: don't get lwtstate twice in ip6_rt_copy_init()

Commit 80f1a0f4e0cd ("net/ipv6: Put lwtstate when destroying fib6_info")
partially fixed the kmemleak [1], lwtstate can be copied from fib6_info,
with ip6_rt_copy_init(), and it should be done only once there.

rt->dst.lwtstate is set by ip6_rt_init_dst(), at the start of the function
ip6_rt_copy_init(), so there is no need to get it again at the end.

With this patch, lwtstate also isn't copied from RTF_REJECT routes.

[1]:
unreferenced object 0xffff880b6aaa14e0 (size 64):
  comm "ip", pid 10577, jiffies 4295149341 (age 1273.903s)
  hex dump (first 32 bytes):
    01 00 04 00 04 00 00 00 10 00 00 00 00 00 00 00  ................
    00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
  backtrace:
    [<0000000018664623>] lwtunnel_build_state+0x1bc/0x420
    [<00000000b73aa29a>] ip6_route_info_create+0x9f7/0x1fd0
    [<00000000ee2c5d1f>] ip6_route_add+0x14/0x70
    [<000000008537b55c>] inet6_rtm_newroute+0xd9/0xe0
    [<000000002acc50f5>] rtnetlink_rcv_msg+0x66f/0x8e0
    [<000000008d9cd381>] netlink_rcv_skb+0x268/0x3b0
    [<000000004c893c76>] netlink_unicast+0x417/0x5a0
    [<00000000f2ab1afb>] netlink_sendmsg+0x70b/0xc30
    [<00000000890ff0aa>] sock_sendmsg+0xb1/0xf0
    [<00000000a2e7b66f>] ___sys_sendmsg+0x659/0x950
    [<000000001e7426c8>] __sys_sendmsg+0xde/0x170
    [<00000000fe411443>] do_syscall_64+0x9f/0x4a0
    [<000000001be7b28b>] entry_SYSCALL_64_after_hwframe+0x49/0xbe
    [<000000006d21f353>] 0xffffffffffffffff

Fixes: 6edb3c96a5f0 ("net/ipv6: Defer initialization of dst to data path")
Signed-off-by: Alexey Kodanev <alexey.kodanev@oracle.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoibmvnic: Include missing return code checks in reset function
Thomas Falcon [Thu, 30 Aug 2018 18:19:53 +0000 (13:19 -0500)]
ibmvnic: Include missing return code checks in reset function

Check the return codes of these functions and halt reset
in case of failure. The driver will remain in a dormant state
until the next reset event, when device initialization will be
re-attempted.

Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoselftests: pmtu: detect correct binary to ping ipv6 addresses
Sabrina Dubroca [Thu, 30 Aug 2018 14:01:18 +0000 (16:01 +0200)]
selftests: pmtu: detect correct binary to ping ipv6 addresses

Some systems don't have the ping6 binary anymore, and use ping for
everything. Detect the absence of ping6 and try to use ping instead.

Fixes: d1f1b9cbf34c ("selftests: net: Introduce first PMTU test")
Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Acked-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoselftests: pmtu: maximum MTU for vti4 is 2^16-1-20
Sabrina Dubroca [Thu, 30 Aug 2018 14:01:17 +0000 (16:01 +0200)]
selftests: pmtu: maximum MTU for vti4 is 2^16-1-20

Since commit 82612de1c98e ("ip_tunnel: restore binding to ifaces with a
large mtu"), the maximum MTU for vti4 is based on IP_MAX_MTU instead of
the mysterious constant 0xFFF8.  This makes this selftest fail.

Fixes: 82612de1c98e ("ip_tunnel: restore binding to ifaces with a large mtu")
Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Acked-by: Stefano Brivio <sbrivio@redhat.com>
Acked-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agotcp: do not restart timewait timer on rst reception
Florian Westphal [Thu, 30 Aug 2018 12:24:29 +0000 (14:24 +0200)]
tcp: do not restart timewait timer on rst reception

RFC 1337 says:
 ''Ignore RST segments in TIME-WAIT state.
   If the 2 minute MSL is enforced, this fix avoids all three hazards.''

So with net.ipv4.tcp_rfc1337=1, expected behaviour is to have TIME-WAIT sk
expire rather than removing it instantly when a reset is received.

However, Linux will also re-start the TIME-WAIT timer.

This causes connect to fail when tying to re-use ports or very long
delays (until syn retry interval exceeds MSL).

packetdrill test case:
// Demonstrate bogus rearming of TIME-WAIT timer in rfc1337 mode.
`sysctl net.ipv4.tcp_rfc1337=1`

0.000 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
0.000 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
0.000 bind(3, ..., ...) = 0
0.000 listen(3, 1) = 0

0.100 < S 0:0(0) win 29200 <mss 1460,nop,nop,sackOK,nop,wscale 7>
0.100 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK,nop,wscale 7>
0.200 < . 1:1(0) ack 1 win 257
0.200 accept(3, ..., ...) = 4

// Receive first segment
0.310 < P. 1:1001(1000) ack 1 win 46

// Send one ACK
0.310 > . 1:1(0) ack 1001

// read 1000 byte
0.310 read(4, ..., 1000) = 1000

// Application writes 100 bytes
0.350 write(4, ..., 100) = 100
0.350 > P. 1:101(100) ack 1001

// ACK
0.500 < . 1001:1001(0) ack 101 win 257

// close the connection
0.600 close(4) = 0
0.600 > F. 101:101(0) ack 1001 win 244

// Our side is in FIN_WAIT_1 & waits for ack to fin
0.7 < . 1001:1001(0) ack 102 win 244

// Our side is in FIN_WAIT_2 with no outstanding data.
0.8 < F. 1001:1001(0) ack 102 win 244
0.8 > . 102:102(0) ack 1002 win 244

// Our side is now in TIME_WAIT state, send ack for fin.
0.9 < F. 1002:1002(0) ack 102 win 244
0.9 > . 102:102(0) ack 1002 win 244

// Peer reopens with in-window SYN:
1.000 < S 1000:1000(0) win 9200 <mss 1460,nop,nop,sackOK,nop,wscale 7>

// Therefore, reply with ACK.
1.000 > . 102:102(0) ack 1002 win 244

// Peer sends RST for this ACK.  Normally this RST results
// in tw socket removal, but rfc1337=1 setting prevents this.
1.100 < R 1002:1002(0) win 244

// second syn. Due to rfc1337=1 expect another pure ACK.
31.0 < S 1000:1000(0) win 9200 <mss 1460,nop,nop,sackOK,nop,wscale 7>
31.0 > . 102:102(0) ack 1002 win 244

// .. and another RST from peer.
31.1 < R 1002:1002(0) win 244
31.2 `echo no timer restart;ss -m -e -a -i -n -t -o state TIME-WAIT`

// third syn after one minute.  Time-Wait socket should have expired by now.
63.0 < S 1000:1000(0) win 9200 <mss 1460,nop,nop,sackOK,nop,wscale 7>

// so we expect a syn-ack & 3whs to proceed from here on.
63.0 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK,nop,wscale 7>

Without this patch, 'ss' shows restarts of tw timer and last packet is
thus just another pure ack, more than one minute later.

This restores the original code from commit 283fd6cf0be690a83
("Merge in ANK networking jumbo patch") in netdev-vger-cvs.git .

For some reason the else branch was removed/lost in 1f28b683339f7
("Merge in TCP/UDP optimizations and [..]") and timer restart became
unconditional.

Reported-by: Michal Tesar <mtesar@redhat.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/rds: RDS is not Radio Data System
Pavel Machek [Thu, 30 Aug 2018 11:30:03 +0000 (13:30 +0200)]
net/rds: RDS is not Radio Data System

Getting prompt "The RDS Protocol" (RDS) is not too helpful, and it is
easily confused with Radio Data System (which we may want to support
in kernel, too).

Signed-off-by: Pavel Machek <pavel@ucw.cz>
Acked-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Acked-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agohv_netvsc: Fix a deadlock by getting rtnl lock earlier in netvsc_probe()
Dexuan Cui [Thu, 30 Aug 2018 05:42:13 +0000 (05:42 +0000)]
hv_netvsc: Fix a deadlock by getting rtnl lock earlier in netvsc_probe()

This patch fixes the race between netvsc_probe() and
rndis_set_subchannel(), which can cause a deadlock.

These are the related 3 paths which show the deadlock:

path #1:
    Workqueue: hv_vmbus_con vmbus_onmessage_work [hv_vmbus]
    Call Trace:
     schedule
     schedule_preempt_disabled
     __mutex_lock
     __device_attach
     bus_probe_device
     device_add
     vmbus_device_register
     vmbus_onoffer
     vmbus_onmessage_work
     process_one_work
     worker_thread
     kthread
     ret_from_fork

path #2:
    schedule
     schedule_preempt_disabled
     __mutex_lock
     netvsc_probe
     vmbus_probe
     really_probe
     __driver_attach
     bus_for_each_dev
     driver_attach_async
     async_run_entry_fn
     process_one_work
     worker_thread
     kthread
     ret_from_fork

path #3:
    Workqueue: events netvsc_subchan_work [hv_netvsc]
    Call Trace:
     schedule
     rndis_set_subchannel
     netvsc_subchan_work
     process_one_work
     worker_thread
     kthread
     ret_from_fork

Before path #1 finishes, path #2 can start to run, because just before
the "bus_probe_device(dev);" in device_add() in path #1, there is a line
"object_uevent(&dev->kobj, KOBJ_ADD);", so systemd-udevd can
immediately try to load hv_netvsc and hence path #2 can start to run.

Next, path #2 offloads the subchannal's initialization to a workqueue,
i.e. path #3, so we can end up in a deadlock situation like this:

Path #2 gets the device lock, and is trying to get the rtnl lock;
Path #3 gets the rtnl lock and is waiting for all the subchannel messages
to be processed;
Path #1 is trying to get the device lock, but since #2 is not releasing
the device lock, path #1 has to sleep; since the VMBus messages are
processed one by one, this means the sub-channel messages can't be
procedded, so #3 has to sleep with the rtnl lock held, and finally #2
has to sleep... Now all the 3 paths are sleeping and we hit the deadlock.

With the patch, we can make sure #2 gets both the device lock and the
rtnl lock together, gets its job done, and releases the locks, so #1
and #3 will not be blocked for ever.

Fixes: 8195b1396ec8 ("hv_netvsc: fix deadlock on hotplug")
Signed-off-by: Dexuan Cui <decui@microsoft.com>
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Cc: K. Y. Srinivasan <kys@microsoft.com>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonfp: wait for posted reconfigs when disabling the device
Jakub Kicinski [Wed, 29 Aug 2018 19:46:08 +0000 (12:46 -0700)]
nfp: wait for posted reconfigs when disabling the device

To avoid leaking a running timer we need to wait for the
posted reconfigs after netdev is unregistered.  In common
case the process of deinitializing the device will perform
synchronous reconfigs which wait for posted requests, but
especially with VXLAN ports being actively added and removed
there can be a race condition leaving a timer running after
adapter structure is freed leading to a crash.

Add an explicit flush after deregistering and for a good
measure a warning to check if timer is running just before
structures are freed.

Fixes: 3d780b926a12 ("nfp: add async reconfiguration mechanism")
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoRevert "packet: switch kvzalloc to allocate memory"
Eric Dumazet [Wed, 29 Aug 2018 18:50:12 +0000 (11:50 -0700)]
Revert "packet: switch kvzalloc to allocate memory"

This reverts commit 71e41286203c017d24f041a7cd71abea7ca7b1e0.

mmap()/munmap() can not be backed by kmalloced pages :

We fault in :

    VM_BUG_ON_PAGE(PageSlab(page), page);

    unmap_single_vma+0x8a/0x110
    unmap_vmas+0x4b/0x90
    unmap_region+0xc9/0x140
    do_munmap+0x274/0x360
    vm_munmap+0x81/0xc0
    SyS_munmap+0x2b/0x40
    do_syscall_64+0x13e/0x1c0
    entry_SYSCALL_64_after_hwframe+0x42/0xb7

Fixes: 71e41286203c ("packet: switch kvzalloc to allocate memory")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: John Sperbeck <jsperbeck@google.com>
Bisected-by: John Sperbeck <jsperbeck@google.com>
Cc: Zhang Yu <zhangyu31@baidu.com>
Cc: Li RongQing <lirongqing@baidu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomac80211: always account for A-MSDU header changes
Johannes Berg [Thu, 30 Aug 2018 08:55:49 +0000 (10:55 +0200)]
mac80211: always account for A-MSDU header changes

In the error path of changing the SKB headroom of the second
A-MSDU subframe, we would not account for the already-changed
length of the first frame that just got converted to be in
A-MSDU format and thus is a bit longer now.

Fix this by doing the necessary accounting.

It would be possible to reorder the operations, but that would
make the code more complex (to calculate the necessary pad),
and the headroom expansion should not fail frequently enough
to make that worthwhile.

Fixes: 6e0456b54545 ("mac80211: add A-MSDU tx support")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Acked-by: Lorenzo Bianconi <lorenzo.bianconi@redhat.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: do not convert to A-MSDU if frag/subframe limited
Lorenzo Bianconi [Wed, 29 Aug 2018 19:03:25 +0000 (21:03 +0200)]
mac80211: do not convert to A-MSDU if frag/subframe limited

Do not start to aggregate packets in a A-MSDU frame (converting the
first subframe to A-MSDU, adding the header) if max_tx_fragments or
max_amsdu_subframes limits are already exceeded by it. In particular,
this happens when drivers set the limit to 1 to avoid A-MSDUs at all.

Signed-off-by: Lorenzo Bianconi <lorenzo.bianconi@redhat.com>
[reword commit message to be more precise]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agocfg80211: nl80211_update_ft_ies() to validate NL80211_ATTR_IE
Arunk Khandavalli [Wed, 29 Aug 2018 21:40:16 +0000 (00:40 +0300)]
cfg80211: nl80211_update_ft_ies() to validate NL80211_ATTR_IE

nl80211_update_ft_ies() tried to validate NL80211_ATTR_IE with
is_valid_ie_attr() before dereferencing it, but that helper function
returns true in case of NULL pointer (i.e., attribute not included).
This can result to dereferencing a NULL pointer. Fix that by explicitly
checking that NL80211_ATTR_IE is included.

Fixes: 355199e02b83 ("cfg80211: Extend support for IEEE 802.11r Fast BSS Transition")
Signed-off-by: Arunk Khandavalli <akhandav@codeaurora.org>
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agoMerge branch 'net_sched-reject-unknown-tcfa_action-values'
David S. Miller [Thu, 30 Aug 2018 05:10:59 +0000 (22:10 -0700)]
Merge branch 'net_sched-reject-unknown-tcfa_action-values'

Paolo Abeni says:

====================
net_sched: reject unknown tcfa_action values

As agreed some time ago, this changeset reject unknown tcfa_action values,
instead of changing such values under the hood.

A tdc test is included to verify the new behavior.

v1 -> v2:
 - helper is now static and renamed according to act_* convention
 - updated extack message, according to the new behavior
====================

Reviewed-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agotc-testing: add test-cases for numeric and invalid control action
Paolo Abeni [Wed, 29 Aug 2018 08:22:34 +0000 (10:22 +0200)]
tc-testing: add test-cases for numeric and invalid control action

Only the police action allows us to specify an arbitrary numeric value
for the control action. This change introduces an explicit test case
for the above feature and then leverage it for testing the kernel behavior
for invalid control actions (reject).

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet_sched: reject unknown tcfa_action values
Paolo Abeni [Wed, 29 Aug 2018 08:22:33 +0000 (10:22 +0200)]
net_sched: reject unknown tcfa_action values

After the commit 802bfb19152c ("net/sched: user-space can't set
unknown tcfa_action values"), unknown tcfa_action values are
converted to TC_ACT_UNSPEC, but the common agreement is instead
rejecting such configurations.

This change also introduces a helper to simplify the destruction
of a single action, avoiding code duplication.

v1 -> v2:
 - helper is now static and renamed according to act_* convention
 - updated extack message, according to the new behavior

Fixes: 802bfb19152c ("net/sched: user-space can't set unknown tcfa_action values")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: mvpp2: initialize port of_node pointer
Baruch Siach [Wed, 29 Aug 2018 06:44:39 +0000 (09:44 +0300)]
net: mvpp2: initialize port of_node pointer

Without a valid of_node in struct device we can't find the mvpp2 port
device by its DT node. Specifically, this breaks
of_find_net_device_by_node().

For example, the Armada 8040 based Clearfog GT-8K uses Marvell 88E6141
switch connected to the &cp1_eth2 port:

&cp1_mdio {
...

switch0: switch0@4 {
compatible = "marvell,mv88e6085";
...

ports {
...

port@5 {
reg = <5>;
label = "cpu";
ethernet = <&cp1_eth2>;
};
};
};
};

Without this patch, dsa_register_switch() returns -EPROBE_DEFER because
of_find_net_device_by_node() can't find the device_node of the &cp1_eth2
device.

Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: bcmgenet: use MAC link status for fixed phy
Doug Berger [Tue, 28 Aug 2018 19:33:15 +0000 (12:33 -0700)]
net: bcmgenet: use MAC link status for fixed phy

When using the fixed PHY with GENET (e.g. MOCA) the PHY link
status can be determined from the internal link status captured
by the MAC. This allows the PHY state machine to use the correct
link state with the fixed PHY even if MAC link event interrupts
are missed when the net device is opened.

Fixes: 8d88c6ebb34c ("net: bcmgenet: enable MoCA link state change detection")
Signed-off-by: Doug Berger <opendmb@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: stmmac: build the dwmac-socfpga platform driver for Stratix10
Dinh Nguyen [Tue, 28 Aug 2018 15:19:20 +0000 (10:19 -0500)]
net: stmmac: build the dwmac-socfpga platform driver for Stratix10

The Stratix10 SoC is an AARCH64 based platform that shares the same ethernet
controller that is on other SoCFPGA platforms. Build the platform driver.

Signed-off-by: Dinh Nguyen <dinguyen@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'ipv6-fix-error-path-of-inet6_init'
David S. Miller [Thu, 30 Aug 2018 02:28:55 +0000 (19:28 -0700)]
Merge branch 'ipv6-fix-error-path-of-inet6_init'

Sabrina Dubroca says:

====================
ipv6: fix error path of inet6_init()

The error path of inet6_init() can trigger multiple kernel panics,
mostly due to wrong ordering of cleanups. This series fixes those
issues.
====================

Reviewed-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: rtnl: return early from rtnl_unregister_all when protocol isn't registered
Sabrina Dubroca [Tue, 28 Aug 2018 11:40:53 +0000 (13:40 +0200)]
net: rtnl: return early from rtnl_unregister_all when protocol isn't registered

rtnl_unregister_all(PF_INET6) gets called from inet6_init in cases when
no handler has been registered for PF_INET6 yet, for example if
ip6_mr_init() fails. Abort and avoid a NULL pointer deref in that case.

Example of panic (triggered by faking a failure of
 register_pernet_subsys):

    general protection fault: 0000 [#1] PREEMPT SMP KASAN PTI
    [...]
    RIP: 0010:rtnl_unregister_all+0x17e/0x2a0
    [...]
    Call Trace:
     ? rtnetlink_net_init+0x250/0x250
     ? sock_unregister+0x103/0x160
     ? kernel_getsockopt+0x200/0x200
     inet6_init+0x197/0x20d

Fixes: e2fddf5e96df ("[IPV6]: Make af_inet6 to check ip6_route_init return value.")
Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoipv6: fix cleanup ordering for pingv6 registration
Sabrina Dubroca [Tue, 28 Aug 2018 11:40:52 +0000 (13:40 +0200)]
ipv6: fix cleanup ordering for pingv6 registration

Commit 6d0bfe226116 ("net: ipv6: Add IPv6 support to the ping socket.")
contains an error in the cleanup path of inet6_init(): when
proto_register(&pingv6_prot, 1) fails, we try to unregister
&pingv6_prot. When rawv6_init() fails, we skip unregistering
&pingv6_prot.

Example of panic (triggered by faking a failure of
 proto_register(&pingv6_prot, 1)):

    general protection fault: 0000 [#1] PREEMPT SMP KASAN PTI
    [...]
    RIP: 0010:__list_del_entry_valid+0x79/0x160
    [...]
    Call Trace:
     proto_unregister+0xbb/0x550
     ? trace_preempt_on+0x6f0/0x6f0
     ? sock_no_shutdown+0x10/0x10
     inet6_init+0x153/0x1b8

Fixes: 6d0bfe226116 ("net: ipv6: Add IPv6 support to the ping socket.")
Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoipv6: fix cleanup ordering for ip6_mr failure
Sabrina Dubroca [Tue, 28 Aug 2018 11:40:51 +0000 (13:40 +0200)]
ipv6: fix cleanup ordering for ip6_mr failure

Commit 15e668070a64 ("ipv6: reorder icmpv6_init() and ip6_mr_init()")
moved the cleanup label for ipmr_fail, but should have changed the
contents of the cleanup labels as well. Now we can end up cleaning up
icmpv6 even though it hasn't been initialized (jump to icmp_fail or
ipmr_fail).

Simply undo things in the reverse order of their initialization.

Example of panic (triggered by faking a failure of icmpv6_init):

    kasan: GPF could be caused by NULL-ptr deref or user memory access
    general protection fault: 0000 [#1] PREEMPT SMP KASAN PTI
    [...]
    RIP: 0010:__list_del_entry_valid+0x79/0x160
    [...]
    Call Trace:
     ? lock_release+0x8a0/0x8a0
     unregister_pernet_operations+0xd4/0x560
     ? ops_free_list+0x480/0x480
     ? down_write+0x91/0x130
     ? unregister_pernet_subsys+0x15/0x30
     ? down_read+0x1b0/0x1b0
     ? up_read+0x110/0x110
     ? kmem_cache_create_usercopy+0x1b4/0x240
     unregister_pernet_subsys+0x1d/0x30
     icmpv6_cleanup+0x1d/0x30
     inet6_init+0x1b5/0x23f

Fixes: 15e668070a64 ("ipv6: reorder icmpv6_init() and ip6_mr_init()")
Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/sched: act_pedit: fix dump of extended layered op
Davide Caratti [Mon, 27 Aug 2018 20:56:22 +0000 (22:56 +0200)]
net/sched: act_pedit: fix dump of extended layered op

in the (rare) case of failure in nla_nest_start(), missing NULL checks in
tcf_pedit_key_ex_dump() can make the following command

 # tc action add action pedit ex munge ip ttl set 64

dereference a NULL pointer:

 BUG: unable to handle kernel NULL pointer dereference at 0000000000000000
 PGD 800000007d1cd067 P4D 800000007d1cd067 PUD 7acd3067 PMD 0
 Oops: 0002 [#1] SMP PTI
 CPU: 0 PID: 3336 Comm: tc Tainted: G            E     4.18.0.pedit+ #425
 Hardware name: Red Hat KVM, BIOS 0.5.1 01/01/2011
 RIP: 0010:tcf_pedit_dump+0x19d/0x358 [act_pedit]
 Code: be 02 00 00 00 48 89 df 66 89 44 24 20 e8 9b b1 fd e0 85 c0 75 46 8b 83 c8 00 00 00 49 83 c5 08 48 03 83 d0 00 00 00 4d 39 f5 <66> 89 04 25 00 00 00 00 0f 84 81 01 00 00 41 8b 45 00 48 8d 4c 24
 RSP: 0018:ffffb5d4004478a8 EFLAGS: 00010246
 RAX: ffff8880fcda2070 RBX: ffff8880fadd2900 RCX: 0000000000000000
 RDX: 0000000000000002 RSI: ffffb5d4004478ca RDI: ffff8880fcda206e
 RBP: ffff8880fb9cb900 R08: 0000000000000008 R09: ffff8880fcda206e
 R10: ffff8880fadd2900 R11: 0000000000000000 R12: ffff8880fd26cf40
 R13: ffff8880fc957430 R14: ffff8880fc957430 R15: ffff8880fb9cb988
 FS:  00007f75a537a740(0000) GS:ffff8880fda00000(0000) knlGS:0000000000000000
 CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
 CR2: 0000000000000000 CR3: 000000007a2fa005 CR4: 00000000001606f0
 Call Trace:
  ? __nla_reserve+0x38/0x50
  tcf_action_dump_1+0xd2/0x130
  tcf_action_dump+0x6a/0xf0
  tca_get_fill.constprop.31+0xa3/0x120
  tcf_action_add+0xd1/0x170
  tc_ctl_action+0x137/0x150
  rtnetlink_rcv_msg+0x263/0x2d0
  ? _cond_resched+0x15/0x40
  ? rtnl_calcit.isra.30+0x110/0x110
  netlink_rcv_skb+0x4d/0x130
  netlink_unicast+0x1a3/0x250
  netlink_sendmsg+0x2ae/0x3a0
  sock_sendmsg+0x36/0x40
  ___sys_sendmsg+0x26f/0x2d0
  ? do_wp_page+0x8e/0x5f0
  ? handle_pte_fault+0x6c3/0xf50
  ? __handle_mm_fault+0x38e/0x520
  ? __sys_sendmsg+0x5e/0xa0
  __sys_sendmsg+0x5e/0xa0
  do_syscall_64+0x5b/0x180
  entry_SYSCALL_64_after_hwframe+0x44/0xa9
 RIP: 0033:0x7f75a4583ba0
 Code: c3 48 8b 05 f2 62 2c 00 f7 db 64 89 18 48 83 cb ff eb dd 0f 1f 80 00 00 00 00 83 3d fd c3 2c 00 00 75 10 b8 2e 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 31 c3 48 83 ec 08 e8 ae cc 00 00 48 89 04 24
 RSP: 002b:00007fff60ee7418 EFLAGS: 00000246 ORIG_RAX: 000000000000002e
 RAX: ffffffffffffffda RBX: 00007fff60ee7540 RCX: 00007f75a4583ba0
 RDX: 0000000000000000 RSI: 00007fff60ee7490 RDI: 0000000000000003
 RBP: 000000005b842d3e R08: 0000000000000002 R09: 0000000000000000
 R10: 00007fff60ee6ea0 R11: 0000000000000246 R12: 0000000000000000
 R13: 00007fff60ee7554 R14: 0000000000000001 R15: 000000000066c100
 Modules linked in: act_pedit(E) ip6table_filter ip6_tables iptable_filter binfmt_misc crct10dif_pclmul ext4 crc32_pclmul mbcache ghash_clmulni_intel jbd2 pcbc snd_hda_codec_generic snd_hda_intel snd_hda_codec snd_hda_core snd_hwdep snd_seq snd_seq_device snd_pcm aesni_intel crypto_simd snd_timer cryptd glue_helper snd joydev pcspkr soundcore virtio_balloon i2c_piix4 nfsd auth_rpcgss nfs_acl lockd grace sunrpc ip_tables xfs libcrc32c ata_generic pata_acpi virtio_net net_failover virtio_blk virtio_console failover qxl crc32c_intel drm_kms_helper syscopyarea serio_raw sysfillrect sysimgblt fb_sys_fops ttm drm ata_piix virtio_pci libata virtio_ring i2c_core virtio floppy dm_mirror dm_region_hash dm_log dm_mod [last unloaded: act_pedit]
 CR2: 0000000000000000

Like it's done for other TC actions, give up dumping pedit rules and return
an error if nla_nest_start() returns NULL.

Fixes: 71d0ed7079df ("net/act_pedit: Support using offset relative to the conventional network headers")
Signed-off-by: Davide Caratti <dcaratti@redhat.com>
Acked-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agosh_eth: Add R7S9210 support
Chris Brandt [Mon, 27 Aug 2018 17:42:02 +0000 (12:42 -0500)]
sh_eth: Add R7S9210 support

Add support for the R7S9210 which is part of the RZ/A2 series.

Signed-off-by: Chris Brandt <chris.brandt@renesas.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'hns-fixes'
David S. Miller [Thu, 30 Aug 2018 01:08:20 +0000 (18:08 -0700)]
Merge branch 'hns-fixes'

Peng Li says:

====================
net: hns: fix some bugs about speed and duplex change

If there are packets in hardware when changing the spped
or duplex, it may cause hardware hang up.

This patchset adds the code for waiting chip to clean the all
pkts(TX & RX) in chip when the driver uses the function named
"adjust link".

This patchset cleans the pkts as follows:
1) close rx of chip, close tx of protocol stack.
2) wait rcb, ppe, mac to clean.
3) adjust link
4) open rx of chip, open tx of protocol stack.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns: add netif_carrier_off before change speed and duplex
Peng Li [Mon, 27 Aug 2018 01:59:30 +0000 (09:59 +0800)]
net: hns: add netif_carrier_off before change speed and duplex

If there are packets in hardware when changing the speed
or duplex, it may cause hardware hang up.

This patch adds netif_carrier_off before change speed and
duplex in ethtool_ops.set_link_ksettings, and adds
netif_carrier_on after complete the change.

Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns: add the code for cleaning pkt in chip
Peng Li [Mon, 27 Aug 2018 01:59:29 +0000 (09:59 +0800)]
net: hns: add the code for cleaning pkt in chip

If there are packets in hardware when changing the speed
or duplex, it may cause hardware hang up.

This patch adds the code for waiting chip to clean the all
pkts(TX & RX) in chip when the driver uses the function named
"adjust link".

This patch cleans the pkts as follows:
1) close rx of chip, close tx of protocol stack.
2) wait rcb, ppe, mac to clean.
3) adjust link
4) open rx of chip, open tx of protocol stack.

Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agor8169: set RxConfig after tx/rx is enabled for RTL8169sb/8110sb devices
Azat Khuzhin [Sun, 26 Aug 2018 14:03:09 +0000 (17:03 +0300)]
r8169: set RxConfig after tx/rx is enabled for RTL8169sb/8110sb devices

I have two Ethernet adapters:
  r8169 0000:03:01.0 eth0: RTL8169sb/8110sb, 00:14:d1:14:2d:49, XID 10000000, IRQ 18
  r8169 0000:01:00.0 eth0: RTL8168e/8111e, 64:66:b3:11:14:5d, XID 2c200000, IRQ 30
And after upgrading from linux 4.15 [1] to linux 4.18+ [2] RTL8169sb failed to
receive any packets. tcpdump shows a lot of checksum mismatch.

  [1]: a0f79386a4968b4925da6db2d1daffd0605a4402
  [2]: 0519359784328bfa92bf0931bf0cff3b58c16932 (4.19 merge window opened)

I started bisecting and the found that [3] breaks it. According to [4]:
  "For 8110S, 8110SB, and 8110SC series, the initial value of RxConfig
  needs to be set after the tx/rx is enabled."
So I moved rtl_init_rxcfg() after enabling tx/rs and now my adapter works
(RTL8168e works too).

  [3]: 3559d81e76bfe3803e89f2e04cf6ef7ab4f3aace
  [4]: e542a2269f232d61270ceddd42b73a4348dee2bb ("r8169: adjust the RxConfig
settings.")

Also drop "rx" from rtl_set_rx_tx_config_registers(), since it does nothing
with it already.

Fixes: 3559d81e76bfe3803e89f2e04cf6ef7ab4f3aace ("r8169: simplify
rtl_hw_start_8169")

Cc: Heiner Kallweit <hkallweit1@gmail.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: netdev@vger.kernel.org
Cc: Realtek linux nic maintainers <nic_swsd@realtek.com>
Signed-off-by: Azat Khuzhin <a3at.mail@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agotipc: switch to rhashtable iterator
Cong Wang [Fri, 24 Aug 2018 19:28:06 +0000 (12:28 -0700)]
tipc: switch to rhashtable iterator

syzbot reported a use-after-free in tipc_group_fill_sock_diag(),
where tipc_group_fill_sock_diag() still reads tsk->group meanwhile
tipc_group_delete() just deletes it in tipc_release().

tipc_nl_sk_walk() aims to lock this sock when walking each sock
in the hash table to close race conditions with sock changes like
this one, by acquiring tsk->sk.sk_lock.slock spinlock, unfortunately
this doesn't work at all. All non-BH call path should take
lock_sock() instead to make it work.

tipc_nl_sk_walk() brutally iterates with raw rht_for_each_entry_rcu()
where RCU read lock is required, this is the reason why lock_sock()
can't be taken on this path. This could be resolved by switching to
rhashtable iterator API's, where taking a sleepable lock is possible.
Also, the iterator API's are friendly for restartable calls like
diag dump, the last position is remembered behind the scence,
all we need to do here is saving the iterator into cb->args[].

I tested this with parallel tipc diag dump and thousands of tipc
socket creation and release, no crash or memory leak.

Reported-by: syzbot+b9c8f3ab2994b7cd1625@syzkaller.appspotmail.com
Cc: Jon Maloy <jon.maloy@ericsson.com>
Cc: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoRevert "net: stmmac: Do not keep rearming the coalesce timer in stmmac_xmit"
Jerome Brunet [Fri, 24 Aug 2018 09:04:40 +0000 (11:04 +0200)]
Revert "net: stmmac: Do not keep rearming the coalesce timer in stmmac_xmit"

This reverts commit 4ae0169fd1b3c792b66be58995b7e6b629919ecf.

This change in the handling of the coalesce timer is causing regression on
(at least) amlogic platforms.

Network will break down very quickly (a few seconds) after starting
a download. This can easily be reproduced using iperf3 for example.

The problem has been reported on the S805, S905, S912 and A113 SoCs
(Realtek and Micrel PHYs) and it is likely impacting all Amlogics
platforms using Gbit ethernet

No problem was seen with the platform using 10/100 only PHYs (GXL internal)

Reverting change brings things back to normal and allows to use network
again until we better understand the problem with the coalesce timer.

Cc: Jose Abreu <joabreu@synopsys.com>
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: Corentin Labbe <clabbe@baylibre.com>
Signed-off-by: Jerome Brunet <jbrunet@baylibre.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agotipc: fix a missing rhashtable_walk_exit()
Cong Wang [Thu, 23 Aug 2018 23:19:44 +0000 (16:19 -0700)]
tipc: fix a missing rhashtable_walk_exit()

rhashtable_walk_exit() must be paired with rhashtable_walk_enter().

Fixes: 40f9f4397060 ("tipc: Fix tipc_sk_reinit race conditions")
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agovti6: remove !skb->ignore_df check from vti6_xmit()
Alexey Kodanev [Thu, 23 Aug 2018 16:49:54 +0000 (19:49 +0300)]
vti6: remove !skb->ignore_df check from vti6_xmit()

Before the commit d6990976af7c ("vti6: fix PMTU caching and reporting
on xmit") '!skb->ignore_df' check was always true because the function
skb_scrub_packet() was called before it, resetting ignore_df to zero.

In the commit, skb_scrub_packet() was moved below, and now this check
can be false for the packet, e.g. when sending it in the two fragments,
this prevents successful PMTU updates in such case. The next attempts
to send the packet lead to the same tx error. Moreover, vti6 initial
MTU value relies on PMTU adjustments.

This issue can be reproduced with the following LTP test script:
    udp_ipsec_vti.sh -6 -p ah -m tunnel -s 2000

Fixes: ccd740cbc6e0 ("vti6: Add pmtu handling to vti6_xmit.")
Signed-off-by: Alexey Kodanev <alexey.kodanev@oracle.com>
Acked-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf
David S. Miller [Wed, 29 Aug 2018 23:19:38 +0000 (16:19 -0700)]
Merge git://git./pub/scm/linux/kernel/git/bpf/bpf

Daniel Borkmann says:

====================
pull-request: bpf 2018-08-29

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

The main changes are:

1) Fix a build error in sk_reuseport_convert_ctx_access() when
   compiling with clang which cannot resolve hweight_long() at
   build time inside the BUILD_BUG_ON() assertion, from Stefan.

2) Several fixes for BPF sockmap, four of them in getting the
   bpf_msg_pull_data() helper to work, one use after free case
   in bpf_tcp_close() and one refcount leak in bpf_tcp_recvmsg(),
   from Daniel.

3) Another fix for BPF sockmap where we misaccount sk_mem_uncharge()
   in the socket redirect error case from unwinding scatterlist
   twice, from John.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'bpf_msg_pull_data-fixes'
Alexei Starovoitov [Wed, 29 Aug 2018 17:47:17 +0000 (10:47 -0700)]
Merge branch 'bpf_msg_pull_data-fixes'

Daniel Borkmann says:

====================
This set contains three more fixes for the bpf_msg_pull_data()
mainly for correcting scatterlist ring wrap-arounds as well as
fixing up data pointers. For details please see individual patches.
Thanks!
====================

Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agobpf: fix sg shift repair start offset in bpf_msg_pull_data
Daniel Borkmann [Wed, 29 Aug 2018 14:50:36 +0000 (16:50 +0200)]
bpf: fix sg shift repair start offset in bpf_msg_pull_data

When we perform the sg shift repair for the scatterlist ring, we
currently start out at i = first_sg + 1. However, this is not
correct since the first_sg could point to the sge sitting at slot
MAX_SKB_FRAGS - 1, and a subsequent i = MAX_SKB_FRAGS will access
the scatterlist ring (sg) out of bounds. Add the sk_msg_iter_var()
helper for iterating through the ring, and apply the same rule
for advancing to the next ring element as we do elsewhere. Later
work will use this helper also in other places.

Fixes: 015632bb30da ("bpf: sk_msg program helper bpf_sk_msg_pull_data")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agobpf: fix shift upon scatterlist ring wrap-around in bpf_msg_pull_data
Daniel Borkmann [Wed, 29 Aug 2018 14:50:35 +0000 (16:50 +0200)]
bpf: fix shift upon scatterlist ring wrap-around in bpf_msg_pull_data

If first_sg and last_sg wraps around in the scatterlist ring, then we
need to account for that in the shift as well. E.g. crafting such msgs
where this is the case leads to a hang as shift becomes negative. E.g.
consider the following scenario:

  first_sg := 14     |=>    shift := -12     msg->sg_start := 10
  last_sg  :=  3     |                       msg->sg_end   :=  5

round  1:  i := 15, move_from :=   3, sg[15] := sg[  3]
round  2:  i :=  0, move_from := -12, sg[ 0] := sg[-12]
round  3:  i :=  1, move_from := -11, sg[ 1] := sg[-11]
round  4:  i :=  2, move_from := -10, sg[ 2] := sg[-10]
[...]
round 13:  i := 11, move_from :=  -1, sg[ 2] := sg[ -1]
round 14:  i := 12, move_from :=   0, sg[ 2] := sg[  0]
round 15:  i := 13, move_from :=   1, sg[ 2] := sg[  1]
round 16:  i := 14, move_from :=   2, sg[ 2] := sg[  2]
round 17:  i := 15, move_from :=   3, sg[ 2] := sg[  3]
[...]

This means we will loop forever and never hit the msg->sg_end condition
to break out of the loop. When we see that the ring wraps around, then
the shift should be MAX_SKB_FRAGS - first_sg + last_sg - 1. Meaning,
the remainder slots from the tail of the ring and the head until last_sg
combined.

Fixes: 015632bb30da ("bpf: sk_msg program helper bpf_sk_msg_pull_data")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agobpf: fix msg->data/data_end after sg shift repair in bpf_msg_pull_data
Daniel Borkmann [Wed, 29 Aug 2018 14:50:34 +0000 (16:50 +0200)]
bpf: fix msg->data/data_end after sg shift repair in bpf_msg_pull_data

In the current code, msg->data is set as sg_virt(&sg[i]) + start - offset
and msg->data_end relative to it as msg->data + bytes. Using iterator i
to point to the updated starting scatterlist element holds true for some
cases, however not for all where we'd end up pointing out of bounds. It
is /correct/ for these ones:

1) When first finding the starting scatterlist element (sge) where we
   find that the page is already privately owned by the msg and where
   the requested bytes and headroom fit into the sge's length.

However, it's /incorrect/ for the following ones:

2) After we made the requested area private and updated the newly allocated
   page into first_sg slot of the scatterlist ring; when we find that no
   shift repair of the ring is needed where we bail out updating msg->data
   and msg->data_end. At that point i will point to last_sg, which in this
   case is the next elem of first_sg in the ring. The sge at that point
   might as well be invalid (e.g. i == msg->sg_end), which we use for
   setting the range of sg_virt(&sg[i]). The correct one would have been
   first_sg.

3) Similar as in 2) but when we find that a shift repair of the ring is
   needed. In this case we fix up all sges and stop once we've reached the
   end. In this case i will point to will point to the new msg->sg_end,
   and the sge at that point will be invalid. Again here the requested
   range sits in first_sg.

Fixes: 015632bb30da ("bpf: sk_msg program helper bpf_sk_msg_pull_data")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agomac80211: avoid kernel panic when building AMSDU from non-linear SKB
Sara Sharon [Wed, 29 Aug 2018 06:57:02 +0000 (08:57 +0200)]
mac80211: avoid kernel panic when building AMSDU from non-linear SKB

When building building AMSDU from non-linear SKB, we hit a
kernel panic when trying to push the padding to the tail.
Instead, put the padding at the head of the next subframe.
This also fixes the A-MSDU subframes to not have the padding
accounted in the length field and not have pad at all for
the last subframe, both required by the spec.

Fixes: 6e0456b54545 ("mac80211: add A-MSDU tx support")
Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Reviewed-by: Lorenzo Bianconi <lorenzo.bianconi@redhat.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: mesh: fix HWMP sequence numbering to follow standard
Yuan-Chi Pang [Wed, 29 Aug 2018 01:30:08 +0000 (09:30 +0800)]
mac80211: mesh: fix HWMP sequence numbering to follow standard

IEEE 802.11-2016 14.10.8.3 HWMP sequence numbering says:
If it is a target mesh STA, it shall update its own HWMP SN to
maximum (current HWMP SN, target HWMP SN in the PREQ element) + 1
immediately before it generates a PREP element in response to a
PREQ element.

Signed-off-by: Yuan-Chi Pang <fu3mo6goo@gmail.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agobpf: fix several offset tests in bpf_msg_pull_data
Daniel Borkmann [Tue, 28 Aug 2018 14:15:35 +0000 (16:15 +0200)]
bpf: fix several offset tests in bpf_msg_pull_data

While recently going over bpf_msg_pull_data(), I noticed three
issues which are fixed in here:

1) When we attempt to find the first scatterlist element (sge)
   for the start offset, we add len to the offset before we check
   for start < offset + len, whereas it should come after when
   we iterate to the next sge to accumulate the offsets. For
   example, given a start offset of 12 with a sge length of 8
   for the first sge in the list would lead us to determine this
   sge as the first sge thinking it covers first 16 bytes where
   start is located, whereas start sits in subsequent sges so
   we would end up pulling in the wrong data.

2) After figuring out the starting sge, we have a short-cut test
   in !msg->sg_copy[i] && bytes <= len. This checks whether it's
   not needed to make the page at the sge private where we can
   just exit by updating msg->data and msg->data_end. However,
   the length test is not fully correct. bytes <= len checks
   whether the requested bytes (end - start offsets) fit into the
   sge's length. The part that is missing is that start must not
   be sge length aligned. Meaning, the start offset into the sge
   needs to be accounted as well on top of the requested bytes
   as otherwise we can access the sge out of bounds. For example
   the sge could have length of 8, our requested bytes could have
   length of 8, but at a start offset of 4, so we also would need
   to pull in 4 bytes of the next sge, when we jump to the out
   label we do set msg->data to sg_virt(&sg[i]) + start - offset
   and msg->data_end to msg->data + bytes which would be oob.

3) The subsequent bytes < copy test for finding the last sge has
   the same issue as in point 2) but also it tests for less than
   rather than less or equal to. Meaning if the sge length is of
   8 and requested bytes of 8 while having the start aligned with
   the sge, we would unnecessarily go and pull in the next sge as
   well to make it private.

Fixes: 015632bb30da ("bpf: sk_msg program helper bpf_sk_msg_pull_data")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agonl80211: Pass center frequency in kHz instead of MHz
Haim Dreyfuss [Tue, 21 Aug 2018 06:22:20 +0000 (09:22 +0300)]
nl80211: Pass center frequency in kHz instead of MHz

freq_reg_info expects to get the frequency in kHz. Instead we
accidently pass it in MHz.  Thus, currently the function always
return ERR rule. Fix that.

Fixes: 50f32718e125 ("nl80211: Add wmm rule attribute to NL80211_CMD_GET_WIPHY dump command")
Signed-off-by: Haim Dreyfuss <haim.dreyfuss@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
[fix kHz/MHz in commit message]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agonl80211: Fix nla_put_u8 to u16 for NL80211_WMMR_TXOP
Haim Dreyfuss [Tue, 21 Aug 2018 06:22:19 +0000 (09:22 +0300)]
nl80211: Fix nla_put_u8 to u16 for NL80211_WMMR_TXOP

TXOP (also known as Channel Occupancy Time) is u16 and should be
added using nla_put_u16 instead of u8, fix that.

Fixes: 50f32718e125 ("nl80211: Add wmm rule attribute to NL80211_CMD_GET_WIPHY dump command")
Signed-off-by: Haim Dreyfuss <haim.dreyfuss@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211_hwsim: Fix possible Spectre-v1 for hwsim_world_regdom_custom
Jinbum Park [Tue, 31 Jul 2018 14:10:40 +0000 (23:10 +0900)]
mac80211_hwsim: Fix possible Spectre-v1 for hwsim_world_regdom_custom

User controls @idx which to be used as index of hwsim_world_regdom_custom.
So, It can be exploited via Spectre-like attack. (speculative execution)

This kind of attack leaks address of hwsim_world_regdom_custom,
It leads an attacker to bypass security mechanism such as KASLR.

So sanitize @idx before using it to prevent attack.

I leveraged strategy [1] to find and exploit this gadget.

[1] https://github.com/jinb-park/linux-exploit/tree/master/exploit-remaining-spectre-gadget/

Signed-off-by: Jinbum Park <jinb.park7@gmail.com>
[johannes: unwrap URL]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: don't update the PM state of a peer upon a multicast frame
Emmanuel Grumbach [Mon, 20 Aug 2018 10:56:07 +0000 (13:56 +0300)]
mac80211: don't update the PM state of a peer upon a multicast frame

I changed the way mac80211 updates the PM state of the peer.
I forgot that we could also have multicast frames from the
peer and that those frame should of course not change the
PM state of the peer: A peer goes to power save when it
needs to scan, but it won't send the broadcast Probe Request
with the PM bit set.

This made us mark the peer as awake when it wasn't and then
Intel's firmware would fail to transmit because the peer is
asleep according to its database. The driver warned about
this and it looked like this:

 WARNING: CPU: 0 PID: 184 at /usr/src/linux-4.16.14/drivers/net/wireless/intel/iwlwifi/mvm/tx.c:1369 iwl_mvm_rx_tx_cmd+0x53b/0x860
 CPU: 0 PID: 184 Comm: irq/124-iwlwifi Not tainted 4.16.14 #1
 RIP: 0010:iwl_mvm_rx_tx_cmd+0x53b/0x860
 Call Trace:
  iwl_pcie_rx_handle+0x220/0x880
  iwl_pcie_irq_handler+0x6c9/0xa20
  ? irq_forced_thread_fn+0x60/0x60
  ? irq_thread_dtor+0x90/0x90

The relevant code that spits the WARNING is:

        case TX_STATUS_FAIL_DEST_PS:
                /* the FW should have stopped the queue and not
                 * return this status
                 */
                WARN_ON(1);
                info->flags |= IEEE80211_TX_STAT_TX_FILTERED;

This fixes https://bugzilla.kernel.org/show_bug.cgi?id=199967.

Fixes: 9fef65443388 ("mac80211: always update the PM state of a peer on MGMT / DATA frames")
Cc: <stable@vger.kernel.org> #4.16+
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agocfg80211: make wmm_rule part of the reg_rule structure
Stanislaw Gruszka [Wed, 22 Aug 2018 11:52:21 +0000 (13:52 +0200)]
cfg80211: make wmm_rule part of the reg_rule structure

Make wmm_rule be part of the reg_rule structure. This simplifies the
code a lot at the cost of having bigger memory usage. However in most
cases we have only few reg_rule's and when we do have many like in
iwlwifi we do not save memory as it allocates a separate wmm_rule for
each channel anyway.

This also fixes a bug reported in various places where somewhere the
pointers were corrupted and we ended up doing a null-dereference.

Fixes: 230ebaa189af ("cfg80211: read wmm rules from regulatory database")
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
[rephrase commit message slightly]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211_hwsim: correct use of IEEE80211_VHT_CAP_RXSTBC_X
Danek Duvall [Wed, 22 Aug 2018 23:01:05 +0000 (16:01 -0700)]
mac80211_hwsim: correct use of IEEE80211_VHT_CAP_RXSTBC_X

The mac80211_hwsim driver intends to say that it supports up to four
STBC receive streams, but instead it ends up saying something undefined.
The IEEE80211_VHT_CAP_RXSTBC_X macros aren't independent bits that can
be ORed together, but values.  In this case, _4 is the appropriate one
to use.

Signed-off-by: Danek Duvall <duvall@comfychair.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: correct use of IEEE80211_VHT_CAP_RXSTBC_X
Danek Duvall [Wed, 22 Aug 2018 23:01:04 +0000 (16:01 -0700)]
mac80211: correct use of IEEE80211_VHT_CAP_RXSTBC_X

The mod mask for VHT capabilities intends to say that you can override
the number of STBC receive streams, and it does, but only by accident.
The IEEE80211_VHT_CAP_RXSTBC_X aren't bits to be set, but values (albeit
left-shifted).  ORing the bits together gets the right answer, but we
should use the _MASK macro here instead.

Signed-off-by: Danek Duvall <duvall@comfychair.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agobpf: sockmap, decrement copied count correctly in redirect error case
John Fastabend [Sat, 25 Aug 2018 00:37:00 +0000 (17:37 -0700)]
bpf: sockmap, decrement copied count correctly in redirect error case

Currently, when a redirect occurs in sockmap and an error occurs in
the redirect call we unwind the scatterlist once in the error path
of bpf_tcp_sendmsg_do_redirect() and then again in sendmsg(). Then
in the error path of sendmsg we decrement the copied count by the
send size.

However, its possible we partially sent data before the error was
generated. This can happen if do_tcp_sendpages() partially sends the
scatterlist before encountering a memory pressure error. If this
happens we need to decrement the copied value (the value tracking
how many bytes were actually sent to TCP stack) by the number of
remaining bytes _not_ the entire send size. Otherwise we risk
confusing userspace.

Also we don't need two calls to free the scatterlist one is
good enough. So remove the one in bpf_tcp_sendmsg_do_redirect() and
then properly reduce copied by the number of remaining bytes which
may in fact be the entire send size if no bytes were sent.

To do this use bool to indicate if free_start_sg() should do mem
accounting or not.

Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
5 years agobpf: fix build error with clang
Stefan Agner [Mon, 27 Aug 2018 19:30:42 +0000 (21:30 +0200)]
bpf: fix build error with clang

Building the newly introduced BPF_PROG_TYPE_SK_REUSEPORT leads to
a compile time error when building with clang:
net/core/filter.o: In function `sk_reuseport_convert_ctx_access':
  ../net/core/filter.c:7284: undefined reference to `__compiletime_assert_7284'

It seems that clang has issues resolving hweight_long at compile
time. Since SK_FL_PROTO_MASK is a constant, we can use the interface
for known constant arguments which works fine with clang.

Fixes: 2dbb9b9e6df6 ("bpf: Introduce BPF_PROG_TYPE_SK_REUSEPORT")
Signed-off-by: Stefan Agner <stefan@agner.ch>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agobpf, sockmap: fix psock refcount leak in bpf_tcp_recvmsg
Daniel Borkmann [Fri, 24 Aug 2018 20:08:51 +0000 (22:08 +0200)]
bpf, sockmap: fix psock refcount leak in bpf_tcp_recvmsg

In bpf_tcp_recvmsg() we first took a reference on the psock, however
once we find that there are skbs in the normal socket's receive queue
we return with processing them through tcp_recvmsg(). Problem is that
we leak the taken reference on the psock in that path. Given we don't
really do anything with the psock at this point, move the skb_queue_empty()
test before we fetch the psock to fix this case.

Fixes: 8934ce2fd081 ("bpf: sockmap redirect ingress support")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agobpf, sockmap: fix potential use after free in bpf_tcp_close
Daniel Borkmann [Fri, 24 Aug 2018 20:08:50 +0000 (22:08 +0200)]
bpf, sockmap: fix potential use after free in bpf_tcp_close

bpf_tcp_close() we pop the psock linkage to a map via psock_map_pop().
A parallel update on the sock hash map can happen between psock_map_pop()
and lookup_elem_raw() where we override the element under link->hash /
link->key. In bpf_tcp_close()'s lookup_elem_raw() we subsequently only
test whether an element is present, but we do not test whether the
element is infact the element we were looking for.

We lock the sock in bpf_tcp_close() during that time, so do we hold
the lock in sock_hash_update_elem(). However, the latter locks the
sock which is newly updated, not the one we're purging from the hash
table. This means that while one CPU is doing the lookup from bpf_tcp_close(),
another CPU is doing the map update in parallel, dropped our sock from
the hlist and released the psock.

Subsequently the first CPU will find the new sock and attempts to drop
and release the old sock yet another time. Fix is that we need to check
the elements for a match after lookup, similar as we do in the sock map.
Note that the hash tab elems are freed via RCU, so access to their
link->hash / link->key is fine since we're under RCU read side there.

Fixes: e9db4ef6bf4c ("bpf: sockhash fix omitted bucket lock in sock_close")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agonet/rds: Use rdma_read_gids to get connection SGID/DGID in IPv6
Zhu Yanjun [Sat, 25 Aug 2018 07:19:05 +0000 (15:19 +0800)]
net/rds: Use rdma_read_gids to get connection SGID/DGID in IPv6

In IPv4, the newly introduced rdma_read_gids is used to read the SGID/DGID
for the connection which returns GID correctly for RoCE transport as well.

In IPv6, rdma_read_gids is also used. The following are why rdma_read_gids
is introduced.

rdma_addr_get_dgid() for RoCE for client side connections returns MAC
address, instead of DGID.
rdma_addr_get_sgid() for RoCE doesn't return correct SGID for IPv6 and
when more than one IP address is assigned to the netdevice.

So the transport agnostic rdma_read_gids() API is provided by rdma_cm
module.

Signed-off-by: Zhu Yanjun <yanjun.zhu@oracle.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: Drop GPIO includes
Linus Walleij [Sun, 26 Aug 2018 22:20:11 +0000 (00:20 +0200)]
net: dsa: Drop GPIO includes

Commit 52638f71fcff ("dsa: Move gpio reset into switch driver")
moved the GPIO handling into the switch drivers but forgot
to remove the GPIO header includes.

Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agotipc: fix the big/little endian issue in tipc_dest
Haiqing Bai [Mon, 27 Aug 2018 01:32:26 +0000 (09:32 +0800)]
tipc: fix the big/little endian issue in tipc_dest

In function tipc_dest_push, the 32bit variables 'node' and 'port'
are stored separately in uppper and lower part of 64bit 'value'.
Then this value is assigned to dst->value which is a union like:
union
{
  struct {
    u32 port;
    u32 node;
  };
  u64 value;
}
This works on little-endian machines like x86 but fails on big-endian
machines.

The fix remove the 'value' stack parameter and even the 'value'
member of the union in tipc_dest, assign the 'node' and 'port' member
directly with the input parameter to avoid the endian issue.

Fixes: a80ae5306a73 ("tipc: improve destination linked list")
Signed-off-by: Zhenbo Gao <zhenbo.gao@windriver.com>
Acked-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Haiqing Bai <Haiqing.Bai@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'net-sched-fixes'
David S. Miller [Mon, 27 Aug 2018 22:16:16 +0000 (15:16 -0700)]
Merge branch 'net-sched-fixes'

Jiri Pirko says:

====================
net: sched: couple of small fixes

Jiri Pirko (2):
  net: sched: fix extack error message when chain is failed to be
    created
  net: sched: return -ENOENT when trying to remove filter from
    non-existent chain
====================

Acked-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: sched: return -ENOENT when trying to remove filter from non-existent chain
Jiri Pirko [Mon, 27 Aug 2018 18:58:44 +0000 (20:58 +0200)]
net: sched: return -ENOENT when trying to remove filter from non-existent chain

When chain 0 was implicitly created, removal of non-existent filter from
chain 0 gave -ENOENT. Once chain 0 became non-implicit, the same call is
giving -EINVAL. Fix this by returning -ENOENT in that case.

Reported-by: Roman Mashak <mrv@mojatatu.com>
Fixes: f71e0ca4db18 ("net: sched: Avoid implicit chain 0 creation")
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: sched: fix extack error message when chain is failed to be created
Jiri Pirko [Mon, 27 Aug 2018 18:58:43 +0000 (20:58 +0200)]
net: sched: fix extack error message when chain is failed to be created

Instead "Cannot find" say "Cannot create".

Fixes: c35a4acc2985 ("net: sched: cls_api: handle generic cls errors")
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoerspan: set erspan_ver to 1 by default when adding an erspan dev
Xin Long [Mon, 27 Aug 2018 10:41:32 +0000 (18:41 +0800)]
erspan: set erspan_ver to 1 by default when adding an erspan dev

After erspan_ver is introudced, if erspan_ver is not set in iproute, its
value will be left 0 by default. Since Commit 02f99df1875c ("erspan: fix
invalid erspan version."), it has broken the traffic due to the version
check in erspan_xmit if users are not aware of 'erspan_ver' param, like
using an old version of iproute.

To fix this compatibility problem, it sets erspan_ver to 1 by default
when adding an erspan dev in erspan_setup. Note that we can't do it in
ipgre_netlink_parms, as this function is also used by ipgre_changelink.

Fixes: 02f99df1875c ("erspan: fix invalid erspan version.")
Reported-by: Jianlin Shi <jishi@redhat.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agosctp: remove useless start_fail from sctp_ht_iter in proc
Xin Long [Mon, 27 Aug 2018 10:40:18 +0000 (18:40 +0800)]
sctp: remove useless start_fail from sctp_ht_iter in proc

After changing rhashtable_walk_start to return void, start_fail would
never be set other value than 0, and the checking for start_fail is
pointless, so remove it.

Fixes: 97a6ec4ac021 ("rhashtable: Change rhashtable_walk_start to return void")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agosctp: hold transport before accessing its asoc in sctp_transport_get_next
Xin Long [Mon, 27 Aug 2018 10:38:31 +0000 (18:38 +0800)]
sctp: hold transport before accessing its asoc in sctp_transport_get_next

As Marcelo noticed, in sctp_transport_get_next, it is iterating over
transports but then also accessing the association directly, without
checking any refcnts before that, which can cause an use-after-free
Read.

So fix it by holding transport before accessing the association. With
that, sctp_transport_hold calls can be removed in the later places.

Fixes: 626d16f50f39 ("sctp: export some apis or variables for sctp_diag and reuse some for proc")
Reported-by: syzbot+fe62a0c9aa6a85c6de16@syzkaller.appspotmail.com
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Mon, 27 Aug 2018 18:59:39 +0000 (11:59 -0700)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) ICE, E1000, IGB, IXGBE, and I40E bug fixes from the Intel folks.

 2) Better fix for AB-BA deadlock in packet scheduler code, from Cong
    Wang.

 3) bpf sockmap fixes (zero sized key handling, etc.) from Daniel
    Borkmann.

 4) Send zero IPID in TCP resets and SYN-RECV state ACKs, to prevent
    attackers using it as a side-channel. From Eric Dumazet.

 5) Memory leak in mediatek bluetooth driver, from Gustavo A. R. Silva.

 6) Hook up rt->dst.input of ipv6 anycast routes properly, from Hangbin
    Liu.

 7) hns and hns3 bug fixes from Huazhong Tan.

 8) Fix RIF leak in mlxsw driver, from Ido Schimmel.

 9) iova range check fix in vhost, from Jason Wang.

10) Fix hang in do_tcp_sendpages() with tls, from John Fastabend.

11) More r8152 chips need to disable RX aggregation, from Kai-Heng Feng.

12) Memory exposure in TCA_U32_SEL handling, from Kees Cook.

13) TCP BBR congestion control fixes from Kevin Yang.

14) hv_netvsc, ignore non-PCI devices, from Stephen Hemminger.

15) qed driver fixes from Tomer Tayar.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (77 commits)
  net: sched: Fix memory exposure from short TCA_U32_SEL
  qed: fix spelling mistake "comparsion" -> "comparison"
  vhost: correctly check the iova range when waking virtqueue
  qlge: Fix netdev features configuration.
  net: macb: do not disable MDIO bus at open/close time
  Revert "net: stmmac: fix build failure due to missing COMMON_CLK dependency"
  net: macb: Fix regression breaking non-MDIO fixed-link PHYs
  mlxsw: spectrum_switchdev: Do not leak RIFs when removing bridge
  i40e: fix condition of WARN_ONCE for stat strings
  i40e: Fix for Tx timeouts when interface is brought up if DCB is enabled
  ixgbe: fix driver behaviour after issuing VFLR
  ixgbe: Prevent unsupported configurations with XDP
  ixgbe: Replace GFP_ATOMIC with GFP_KERNEL
  igb: Replace mdelay() with msleep() in igb_integrated_phy_loopback()
  igb: Replace GFP_ATOMIC with GFP_KERNEL in igb_sw_init()
  igb: Use an advanced ctx descriptor for launchtime
  e1000: ensure to free old tx/rx rings in set_ringparam()
  e1000: check on netif_running() before calling e1000_up()
  ixgb: use dma_zalloc_coherent instead of allocator/memset
  ice: Trivial formatting fixes
  ...

5 years agoFix up libata MAINTAINERS entry
Jens Axboe [Mon, 27 Aug 2018 14:37:46 +0000 (08:37 -0600)]
Fix up libata MAINTAINERS entry

The email was botched in one entry, and I also forgot to update the
location of the git tree. It'll be under the linux-block umbrella, just
with different branches.

Reported-by: Baruch Siach <baruch@tkos.co.il>
Fixes: 7634ccd2da97 ("libata: maintainership update")
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agonet: sched: Fix memory exposure from short TCA_U32_SEL
Kees Cook [Sun, 26 Aug 2018 05:58:01 +0000 (22:58 -0700)]
net: sched: Fix memory exposure from short TCA_U32_SEL

Via u32_change(), TCA_U32_SEL has an unspecified type in the netlink
policy, so max length isn't enforced, only minimum. This means nkeys
(from userspace) was being trusted without checking the actual size of
nla_len(), which could lead to a memory over-read, and ultimately an
exposure via a call to u32_dump(). Reachability is CAP_NET_ADMIN within
a namespace.

Reported-by: Al Viro <viro@zeniv.linux.org.uk>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Jiri Pirko <jiri@resnulli.us>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: netdev@vger.kernel.org
Signed-off-by: Kees Cook <keescook@chromium.org>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoLinux 4.19-rc1 v4.19-rc1
Linus Torvalds [Sun, 26 Aug 2018 21:11:59 +0000 (14:11 -0700)]
Linux 4.19-rc1

5 years agoMerge branch 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 26 Aug 2018 20:39:05 +0000 (13:39 -0700)]
Merge branch 'timers-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull timer update from Thomas Gleixner:
 "New defines for the compat time* types so they can be shared between
  32bit and 64bit builds. Not used yet, but merging them now allows the
  actual conversions to be merged through different maintainer trees
  without dependencies

  We still have compat interfaces for 32bit on 64bit even with the new
  2038 safe timespec/val variants because pointer size is different. And
  for the old style timespec/val interfaces we need yet another 'compat'
  interface for both 32bit native and 32bit on 64bit"

* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  y2038: Provide aliases for compat helpers

5 years agoMerge branch 'ida-4.19' of git://git.infradead.org/users/willy/linux-dax
Linus Torvalds [Sun, 26 Aug 2018 18:48:42 +0000 (11:48 -0700)]
Merge branch 'ida-4.19' of git://git.infradead.org/users/willy/linux-dax

Pull IDA updates from Matthew Wilcox:
 "A better IDA API:

      id = ida_alloc(ida, GFP_xxx);
      ida_free(ida, id);

  rather than the cumbersome ida_simple_get(), ida_simple_remove().

  The new IDA API is similar to ida_simple_get() but better named.  The
  internal restructuring of the IDA code removes the bitmap
  preallocation nonsense.

  I hope the net -200 lines of code is convincing"

* 'ida-4.19' of git://git.infradead.org/users/willy/linux-dax: (29 commits)
  ida: Change ida_get_new_above to return the id
  ida: Remove old API
  test_ida: check_ida_destroy and check_ida_alloc
  test_ida: Convert check_ida_conv to new API
  test_ida: Move ida_check_max
  test_ida: Move ida_check_leaf
  idr-test: Convert ida_check_nomem to new API
  ida: Start new test_ida module
  target/iscsi: Allocate session IDs from an IDA
  iscsi target: fix session creation failure handling
  drm/vmwgfx: Convert to new IDA API
  dmaengine: Convert to new IDA API
  ppc: Convert vas ID allocation to new IDA API
  media: Convert entity ID allocation to new IDA API
  ppc: Convert mmu context allocation to new IDA API
  Convert net_namespace to new IDA API
  cb710: Convert to new IDA API
  rsxx: Convert to new IDA API
  osd: Convert to new IDA API
  sd: Convert to new IDA API
  ...