openwrt/staging/blogic.git
8 years agoMerge tag 'wireless-drivers-for-davem-2016-03-04' of git://git.kernel.org/pub/scm...
David S. Miller [Mon, 7 Mar 2016 19:58:11 +0000 (14:58 -0500)]
Merge tag 'wireless-drivers-for-davem-2016-03-04' of git://git./linux/kernel/git/kvalo/wireless-drivers

Kalle Valo says:

====================
wireless-drivers fixes for 4.5

iwlwifi

* free firmware paging memory when the module is unloaded or device removed
* fix pending frames counter to fix an issue when removing stations

ssb

* fix a build problem related to ssb_fill_sprom_with_fallback()
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: Add missing MFD_SYSCON dependency on HAS_IOMEM
Krzysztof Kozlowski [Fri, 4 Mar 2016 01:04:52 +0000 (10:04 +0900)]
net: ethernet: Add missing MFD_SYSCON dependency on HAS_IOMEM

The MFD_SYSCON depends on HAS_IOMEM so when selecting it avoid unmet
direct dependencies.

Signed-off-by: Krzysztof Kozlowski <k.kozlowski@samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoibmveth: check return of skb_linearize in ibmveth_start_xmit
Thomas Falcon [Thu, 3 Mar 2016 21:22:36 +0000 (15:22 -0600)]
ibmveth: check return of skb_linearize in ibmveth_start_xmit

If skb_linearize fails, the driver should drop the packet
instead of trying to copy it into the bounce buffer.

Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agocdc_ncm: toggle altsetting to force reset before setup
Bjørn Mork [Thu, 3 Mar 2016 21:20:53 +0000 (22:20 +0100)]
cdc_ncm: toggle altsetting to force reset before setup

Some devices will silently fail setup unless they are reset first.
This is necessary even if the data interface is already in
altsetting 0, which it will be when the device is probed for the
first time.  Briefly toggling the altsetting forces a function
reset regardless of the initial state.

This fixes a setup problem observed on a number of Huawei devices,
appearing to operate in NTB-32 mode even if we explicitly set them
to NTB-16 mode.

Signed-off-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agousbnet: cleanup after bind() in probe()
Oliver Neukum [Mon, 7 Mar 2016 10:31:10 +0000 (11:31 +0100)]
usbnet: cleanup after bind() in probe()

In case bind() works, but a later error forces bailing
in probe() in error cases work and a timer may be scheduled.
They must be killed. This fixes an error case related to
the double free reported in
http://www.spinics.net/lists/netdev/msg367669.html
and needs to go on top of Linus' fix to cdc-ncm.

Signed-off-by: Oliver Neukum <ONeukum@suse.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'mlxsw-fixes'
David S. Miller [Mon, 7 Mar 2016 16:39:16 +0000 (11:39 -0500)]
Merge branch 'mlxsw-fixes'

Jiri Pirko says:

====================
mlxsw: couple of fixes

Couple of fixes from Ido.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomlxsw: pci: Correctly determine if descriptor queue is full
Ido Schimmel [Mon, 7 Mar 2016 14:15:30 +0000 (15:15 +0100)]
mlxsw: pci: Correctly determine if descriptor queue is full

The descriptor queues for sending (SDQs) and receiving (RDQs) packets
are managed by two counters - producer and consumer - which are both
16-bit in size. A queue is considered full when the difference between
the two equals the queue's maximum number of descriptors.

However, if the producer counter overflows, then it's possible for the
full queue check to fail, as it doesn't take the overflow into account.
In such a case, descriptors already passed to the device - but for which
a completion has yet to be posted - will be overwritten, thereby causing
undefined behavior. The above can be achieved under heavy load (~30
netperf instances).

Fix that by casting the subtraction result to u16, preventing it from
being treated as a signed integer.

Fixes: eda6500a987a ("mlxsw: Add PCI bus implementation")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomlxsw: spectrum: Always decrement bridge's ref count
Ido Schimmel [Mon, 7 Mar 2016 14:15:29 +0000 (15:15 +0100)]
mlxsw: spectrum: Always decrement bridge's ref count

Since we only support one VLAN filtering bridge we need to associate a
reference count with it, so that when the last port netdev leaves it, we
would know that a different bridge can be offloaded to hardware.

When a LAG device is memeber in a bridge and port netdevs are leaving
the LAG, we should always decrement the bridge's reference count, as it's
incremented for any port in the LAG.

Fixes: 4dc236c31733 ("mlxsw: spectrum: Handle port leaving LAG while bridged")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agotipc: fix nullptr crash during subscription cancel
Parthasarathy Bhuvaragan [Thu, 3 Mar 2016 16:54:54 +0000 (17:54 +0100)]
tipc: fix nullptr crash during subscription cancel

commit 4d5cfcba2f6e ('tipc: fix connection abort during subscription
cancel'), removes the check for a valid subscription before calling
tipc_nametbl_subscribe().

This will lead to a nullptr exception when we process a
subscription cancel request. For a cancel request, a null
subscription is passed to tipc_nametbl_subscribe() resulting
in exception.

In this commit, we call tipc_nametbl_subscribe() only for
a valid subscription.

Fixes: 4d5cfcba2f6e ('tipc: fix connection abort during subscription cancel')
Reported-by: Anders Widell <anders.widell@ericsson.com>
Signed-off-by: Parthasarathy Bhuvaragan <parthasarathy.bhuvaragan@ericsson.com>
Acked-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: eth: altera: do not free array priv->mdio->irq
Colin Ian King [Thu, 3 Mar 2016 13:47:18 +0000 (13:47 +0000)]
net: eth: altera: do not free array priv->mdio->irq

priv->mdio->irq used to be allocated and required freeing, but it
is now a fixed sized array and should no longer be free'd.

Issue detected using static analysis with CoverityScan

Fixes: e7f4dc3536a400 ("mdio: Move allocation of interrupts into core")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/ethoc: do not free array priv->mdio->irq
Colin Ian King [Thu, 3 Mar 2016 13:43:34 +0000 (13:43 +0000)]
net/ethoc: do not free array priv->mdio->irq

priv->mdio->irq used to be allocated and required freeing, but it
is now a fixed sized array and should no longer be free'd.

Issue detected using static analysis with CoverityScan

Fixes: e7f4dc3536a400 ("mdio: Move allocation of interrupts into core")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Reviewed-by: Tobias Klauser <tklauser@distanz.ch>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: sched: fix act_ipt for LOG target
Phil Sutter [Thu, 3 Mar 2016 13:34:14 +0000 (14:34 +0100)]
net: sched: fix act_ipt for LOG target

Before calling the destroy() or target() callbacks, the family parameter
field has to be initialized. Otherwise at least the LOG target will
refuse to work and upon removal oops the kernel.

Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Phil Sutter <phil@nwl.cc>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoasix: do not free array priv->mdio->irq
Colin Ian King [Thu, 3 Mar 2016 13:27:56 +0000 (13:27 +0000)]
asix: do not free array priv->mdio->irq

Used to be allocated and required freeing, but now
priv->mdio->irq is now a fixed sized array and should no longer be
free'd.

Issue detected using static analysis with CoverityScan

Fixes: e7f4dc3536a400 ("mdio: Move allocation of interrupts into core")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agogianfar: Enable eTSEC-106 erratum w/a for MPC8548E Rev2
Atsushi Nemoto [Thu, 3 Mar 2016 00:07:51 +0000 (09:07 +0900)]
gianfar: Enable eTSEC-106 erratum w/a for MPC8548E Rev2

Enable workaround for MPC8548E erratum eTSEC 106,
"Excess delays when transmitting TOE=1 large frames".
(see commit 53fad77375ce "gianfar: Enable eTSEC-20 erratum w/a
for P2020 Rev1")

This erratum was fixed in Rev 3.1.x.

Signed-off-by: Atsushi Nemoto <nemoto@toshiba-tops.co.jp>
Acked-by: Claudiu Manoil <claudiu.manoil@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge tag 'linux-can-fixes-for-4.5-20160304' of git://git.kernel.org/pub/scm/linux...
David S. Miller [Fri, 4 Mar 2016 19:32:47 +0000 (14:32 -0500)]
Merge tag 'linux-can-fixes-for-4.5-20160304' of git://git./linux/kernel/git/mkl/linux-can

Marc Kleine-Budde says:

====================
pull-request: can 2016-03-04

this is a pull request for net/master.

There is one patch from Ed Spiridonov, which increases the performance of the
mcp251x SPI CAN driver, by avoiding to write to error flag register if it's
unnecessary.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agocan: mcp251x: avoid write to error flag register if it's unnecessary
Ed Spiridonov [Fri, 4 Mar 2016 06:07:27 +0000 (09:07 +0300)]
can: mcp251x: avoid write to error flag register if it's unnecessary

Only two bits (RX0OVR and RX1OVR) are writable in EFLG, write is useless
if these bits aren't set.

Signed-off-by: Ed Spiridonov <edo.rus@gmail.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
8 years agobe2net: don't enable multicast flag in be_enable_if_filters() routine
Venkat Duvvuru [Wed, 2 Mar 2016 11:00:28 +0000 (06:00 -0500)]
be2net: don't enable multicast flag in be_enable_if_filters() routine

When the interface is opened (in be_open()) the routine
be_enable_if_filters() must be called to switch on the basic filtering
capabilities of an interface that are not changed at run-time.
These include the flags UNTAGGED, BROADCAST and PASS_L3L4_ERRORS.
Other flags such as MULTICAST and PROMISC must be enabled later by
be_set_rx_mode() based on the state in the netdev/adapter struct.

be_enable_if_filters() routine is wrongly trying to enable MULTICAST flag
without checking the current adapter state. This can cause the RX_FILTER
cmds to the FW to fail.  This patch fixes this problem by only enabling
the basic filtering flags in be_enable_if_filters().

The VF must be able to issue RX_FILTER cmd with any filter flag, as long
as the PF allowed those flags (if_cap_flags) in the iface it provisioned
for the VF. This rule is applicable even when the VF doesn't have the
FILTMGMT privilege. There is a bug in BE3 FW that wrongly fails RX_FILTER
multicast programming cmds on VFs that don't have FILTMGMT privilege.
This patch also helps in insulating the VF driver from be_open failures due
to the FW bug. A fix for the BE3 FW issue will be available in
versions >= 11.0.283.0 and 10.6.334.0

Reported-by: Ivan Vecera <ivecera@redhat.com>
Signed-off-by: Venkat Duvvuru <venkatkumar.duvvuru@avagotech.com>
Signed-off-by: Sathya Perla <sathya.perla@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: moxa: fix an error code
Dan Carpenter [Wed, 2 Mar 2016 10:11:10 +0000 (13:11 +0300)]
net: moxa: fix an error code

We accidentally return IS_ERR(priv->base) which is 1 instead of
PTR_ERR(priv->base) which is the error code.

Fixes: 6c821bd9edc9 ('net: Add MOXA ART SoCs ethernet driver')
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMAINTAINERS: add maintainer entry for FREESCALE FEC ethernet driver
Nimrod Andy [Wed, 2 Mar 2016 09:24:53 +0000 (17:24 +0800)]
MAINTAINERS: add maintainer entry for FREESCALE FEC ethernet driver

Add a maintainer entry for FREESCALE FEC ethernet driver and add myself
as a maintainer.

Signed-off-by: Fugang Duan <fugang.duan@nxp.com>
Acked-by: Frank Li <Frank.Li@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agovxlan: fix missing options_len update on RX with collect metadata
Daniel Borkmann [Wed, 2 Mar 2016 01:32:08 +0000 (02:32 +0100)]
vxlan: fix missing options_len update on RX with collect metadata

When signalling to metadata consumers that the metadata_dst entry
carries additional GBP extension data for vxlan (TUNNEL_VXLAN_OPT),
the dst's vxlan_metadata information is populated, but options_len
is left to zero. F.e. in ovs, ovs_flow_key_extract() checks for
options_len before extracting the data through ip_tunnel_info_opts_get().

Geneve uses ip_tunnel_info_opts_set() helper in receive path, which
sets options_len internally, vxlan however uses ip_tunnel_info_opts(),
so when filling vxlan_metadata, we do need to update options_len.

Fixes: 4c22279848c5 ("ip-tunnel: Use API to access tunnel metadata options.")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: renesas: sh_eth: don't open code of_device_get_match_data()
Wolfram Sang [Tue, 1 Mar 2016 16:37:59 +0000 (17:37 +0100)]
net: ethernet: renesas: sh_eth: don't open code of_device_get_match_data()

This change will also make Coverity happy by avoiding a theoretical NULL
pointer dereference; yet another reason is to use the above helper function
to tighten the code and make it more readable.

Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Reviewed-by: Simon Horman <horms+renesas@verge.net.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: renesas: ravb_main: don't open code of_device_get_match_data()
Wolfram Sang [Tue, 1 Mar 2016 16:37:58 +0000 (17:37 +0100)]
net: ethernet: renesas: ravb_main: don't open code of_device_get_match_data()

This change will also make Coverity happy by avoiding a theoretical NULL
pointer dereference; yet another reason is to use the above helper function
to tighten the code and make it more readable.

Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Reviewed-by: Simon Horman <horms+renesas@verge.net.au>
Acked-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoibmvnic: Fix ibmvnic_capability struct
Thomas Falcon [Tue, 1 Mar 2016 16:20:09 +0000 (10:20 -0600)]
ibmvnic: Fix ibmvnic_capability struct

The ibmvnic_capability struct was defined incorrectly. The last two
elements of the struct are in the wrong order.  In addition, the number
element should be 64-bit. Byteswapping functions are updated
as well.

Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoipv6: re-enable fragment header matching in ipv6_find_hdr
Florian Westphal [Tue, 1 Mar 2016 15:15:16 +0000 (16:15 +0100)]
ipv6: re-enable fragment header matching in ipv6_find_hdr

When ipv6_find_hdr is used to find a fragment header
(caller specifies target NEXTHDR_FRAGMENT) we erronously return
-ENOENT for all fragments with nonzero offset.

Before commit 9195bb8e381d, when target was specified, we did not
enter the exthdr walk loop as nexthdr == target so this used to work.

Now we do (so we can skip empty route headers). When we then stumble upon
a frag with nonzero frag_off we must return -ENOENT ("header not found")
only if the caller did not specifically request NEXTHDR_FRAGMENT.

This allows nfables exthdr expression to match ipv6 fragments, e.g. via

nft add rule ip6 filter input frag frag-off gt 0

Fixes: 9195bb8e381d ("ipv6: improve ipv6_find_hdr() to skip empty routing headers")
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoqmi_wwan: add Sierra Wireless EM74xx device ID
Bjørn Mork [Tue, 1 Mar 2016 13:31:02 +0000 (14:31 +0100)]
qmi_wwan: add Sierra Wireless EM74xx device ID

The MC74xx and EM74xx modules use different IDs by default, according
to the Lenovo EM7455 driver for Windows.

Signed-off-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agotipc: Revert "tipc: use existing sk_write_queue for outgoing packet chain"
Parthasarathy Bhuvaragan [Tue, 1 Mar 2016 10:07:09 +0000 (11:07 +0100)]
tipc: Revert "tipc: use existing sk_write_queue for outgoing packet chain"

reverts commit 94153e36e709e ("tipc: use existing sk_write_queue for
outgoing packet chain")

In Commit 94153e36e709e, we assume that we fill & empty the socket's
sk_write_queue within the same lock_sock() session.

This is not true if the link is congested. During congestion, the
socket lock is released while we wait for the congestion to cease.
This implementation causes a nullptr exception, if the user space
program has several threads accessing the same socket descriptor.

Consider two threads of the same program performing the following:
     Thread1                                  Thread2
--------------------                    ----------------------
Enter tipc_sendmsg()                    Enter tipc_sendmsg()
lock_sock()                             lock_sock()
Enter tipc_link_xmit(), ret=ELINKCONG   spin on socket lock..
sk_wait_event()                             :
release_sock()                          grab socket lock
    :                                   Enter tipc_link_xmit(), ret=0
    :                                   release_sock()
Wakeup after congestion
lock_sock()
skb = skb_peek(pktchain);
!! TIPC_SKB_CB(skb)->wakeup_pending = tsk->link_cong;

In this case, the second thread transmits the buffers belonging to
both thread1 and thread2 successfully. When the first thread wakeup
after the congestion it assumes that the pktchain is intact and
operates on the skb's in it, which leads to the following exception:

[2102.439969] BUG: unable to handle kernel NULL pointer dereference at 00000000000000d0
[2102.440074] IP: [<ffffffffa005f330>] __tipc_link_xmit+0x2b0/0x4d0 [tipc]
[2102.440074] PGD 3fa3f067 PUD 3fa6b067 PMD 0
[2102.440074] Oops: 0000 [#1] SMP
[2102.440074] CPU: 2 PID: 244 Comm: sender Not tainted 3.12.28 #1
[2102.440074] RIP: 0010:[<ffffffffa005f330>]  [<ffffffffa005f330>] __tipc_link_xmit+0x2b0/0x4d0 [tipc]
[...]
[2102.440074] Call Trace:
[2102.440074]  [<ffffffff8163f0b9>] ? schedule+0x29/0x70
[2102.440074]  [<ffffffffa006a756>] ? tipc_node_unlock+0x46/0x170 [tipc]
[2102.440074]  [<ffffffffa005f761>] tipc_link_xmit+0x51/0xf0 [tipc]
[2102.440074]  [<ffffffffa006d8ae>] tipc_send_stream+0x11e/0x4f0 [tipc]
[2102.440074]  [<ffffffff8106b150>] ? __wake_up_sync+0x20/0x20
[2102.440074]  [<ffffffffa006dc9c>] tipc_send_packet+0x1c/0x20 [tipc]
[2102.440074]  [<ffffffff81502478>] sock_sendmsg+0xa8/0xd0
[2102.440074]  [<ffffffff81507895>] ? release_sock+0x145/0x170
[2102.440074]  [<ffffffff815030d8>] ___sys_sendmsg+0x3d8/0x3e0
[2102.440074]  [<ffffffff816426ae>] ? _raw_spin_unlock+0xe/0x10
[2102.440074]  [<ffffffff81115c2a>] ? handle_mm_fault+0x6ca/0x9d0
[2102.440074]  [<ffffffff8107dd65>] ? set_next_entity+0x85/0xa0
[2102.440074]  [<ffffffff816426de>] ? _raw_spin_unlock_irq+0xe/0x20
[2102.440074]  [<ffffffff8107463c>] ? finish_task_switch+0x5c/0xc0
[2102.440074]  [<ffffffff8163ea8c>] ? __schedule+0x34c/0x950
[2102.440074]  [<ffffffff81504e12>] __sys_sendmsg+0x42/0x80
[2102.440074]  [<ffffffff81504e62>] SyS_sendmsg+0x12/0x20
[2102.440074]  [<ffffffff8164aed2>] system_call_fastpath+0x16/0x1b

In this commit, we maintain the skb list always in the stack.

Signed-off-by: Parthasarathy Bhuvaragan <parthasarathy.bhuvaragan@ericsson.com>
Acked-by: Ying Xue <ying.xue@windriver.com>
Acked-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomld, igmp: Fix reserved tailroom calculation
Benjamin Poirier [Mon, 29 Feb 2016 23:03:33 +0000 (15:03 -0800)]
mld, igmp: Fix reserved tailroom calculation

The current reserved_tailroom calculation fails to take hlen and tlen into
account.

skb:
[__hlen__|__data____________|__tlen___|__extra__]
^                                               ^
head                                            skb_end_offset

In this representation, hlen + data + tlen is the size passed to alloc_skb.
"extra" is the extra space made available in __alloc_skb because of
rounding up by kmalloc. We can reorder the representation like so:

[__hlen__|__data____________|__extra__|__tlen___]
^                                               ^
head                                            skb_end_offset

The maximum space available for ip headers and payload without
fragmentation is min(mtu, data + extra). Therefore,
reserved_tailroom
= data + extra + tlen - min(mtu, data + extra)
= skb_end_offset - hlen - min(mtu, skb_end_offset - hlen - tlen)
= skb_tailroom - min(mtu, skb_tailroom - tlen) ; after skb_reserve(hlen)

Compare the second line to the current expression:
reserved_tailroom = skb_end_offset - min(mtu, skb_end_offset)
and we can see that hlen and tlen are not taken into account.

The min() in the third line can be expanded into:
if mtu < skb_tailroom - tlen:
reserved_tailroom = skb_tailroom - mtu
else:
reserved_tailroom = tlen

Depending on hlen, tlen, mtu and the number of multicast address records,
the current code may output skbs that have less tailroom than
dev->needed_tailroom or it may output more skbs than needed because not all
space available is used.

Fixes: 4c672e4b ("ipv6: mld: fix add_grhead skb_over_panic for devs with large MTUs")
Signed-off-by: Benjamin Poirier <bpoirier@suse.com>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agofsl/fman: Initialize fman->dev earlier
Igal Liberman [Sun, 28 Feb 2016 21:59:53 +0000 (23:59 +0200)]
fsl/fman: Initialize fman->dev earlier

Currently, in a case of error, dev_err is using fman->dev
before its initialization and "(NULL device *)" is printed.
This patch fixes this issue.

Signed-off-by: Igal Liberman <igal.liberman@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agostmmac: Fix 'eth0: No PHY found' regression
Gabriel Fernandez [Mon, 29 Feb 2016 16:18:22 +0000 (17:18 +0100)]
stmmac: Fix 'eth0: No PHY found' regression

This patch manages the case when you have an Ethernet MAC with
a "fixed link", and not connected to a normal MDIO-managed PHY device.

The test of phy_bus_name was not helpful because it was never affected
and replaced by the mdio test node.

Signed-off-by: Gabriel Fernandez <gabriel.fernandez@linaro.org>
Acked-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'dwc_eth_qos'
David S. Miller [Wed, 2 Mar 2016 19:57:16 +0000 (14:57 -0500)]
Merge branch 'dwc_eth_qos'

Lars Persson says:

====================
dwc_eth_qos: stability fixes and support for CMA

This series has bug fixes for the dwc_eth_qos ethernet driver.

Mainly two stability fixes for problems found by Rabin Vincent:
- Successive starts and stops of the interface would trigger a DMA reset timeout.
- A race condition in the TX DMA handling could trigger a netdev watchdog
 timeout.

The memory allocation was improved to support use of the CMA as DMA allocator
backend.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodwc_eth_qos: do phy_start before resetting hardware
Lars Persson [Mon, 29 Feb 2016 15:22:34 +0000 (16:22 +0100)]
dwc_eth_qos: do phy_start before resetting hardware

This reverts the changed init order from commit 3647bc35bd42
("dwc_eth_qos: Reset hardware before PHY start") and makes another fix
for the race.

It turned out that the reset state machine of the dwceqos hardware
requires PHY clocks to be present in order to complete the reset
cycle.

To plug the race with the phy state machine we defer link speed
setting until the hardware init has finished.

Signed-off-by: Lars Persson <larper@axis.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodwc_eth_qos: use DWCEQOS_MSG_DEFAULT
Rabin Vincent [Mon, 29 Feb 2016 15:22:33 +0000 (16:22 +0100)]
dwc_eth_qos: use DWCEQOS_MSG_DEFAULT

Since debug is hardcoded to 3, the defaults in the DWCEQOS_MSG_DEFAULT
macro are never used, which does not seem to be the intended behaviour
here.  Set debug to -1 like other drivers so that DWCEQOS_MSG_DEFAULT is
actually used by default.

Signed-off-by: Rabin Vincent <rabinv@axis.com>
Signed-off-by: Lars Persson <larper@axis.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodwc_eth_qos: use GFP_KERNEL in dma_alloc_coherent()
Rabin Vincent [Mon, 29 Feb 2016 15:22:32 +0000 (16:22 +0100)]
dwc_eth_qos: use GFP_KERNEL in dma_alloc_coherent()

Since we are in non-atomic context here we can pass GFP_KERNEL to
dma_alloc_coherent(). This enables use of the CMA.

Signed-off-by: Rabin Vincent <rabinv@axis.com>
Signed-off-by: Lars Persson <larper@axis.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodwc_eth_qos: release descriptors outside netif_tx_lock
Lars Persson [Mon, 29 Feb 2016 15:22:31 +0000 (16:22 +0100)]
dwc_eth_qos: release descriptors outside netif_tx_lock

To prepare for using the CMA, we can not be in atomic context when
de-allocating DMA buffers.

The tx lock was needed only to protect the hw reset against the xmit
handler. Now we briefly grab the tx lock while stopping the queue to
make sure no thread is inside or will enter the xmit handler.

Signed-off-by: Lars Persson <larper@axis.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodwc_eth_qos: fix race condition in dwceqos_start_xmit
Rabin Vincent [Mon, 29 Feb 2016 15:22:30 +0000 (16:22 +0100)]
dwc_eth_qos: fix race condition in dwceqos_start_xmit

The xmit handler and the tx_reclaim tasklet had a race on the tx_free
variable which could lead to a tx timeout if tx_free was updated after
the tx complete interrupt.

Signed-off-by: Rabin Vincent <rabinv@axis.com>
Signed-off-by: Lars Persson <larper@axis.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'mlx4-fixes'
David S. Miller [Wed, 2 Mar 2016 19:42:46 +0000 (14:42 -0500)]
Merge branch 'mlx4-fixes'

Or Gerlitz says:

====================
Mellanox 10/40G mlx4 driver fixes for 4.5-rc6

This series contains two fixes for the SRIOV HW LAG that was
introduced in 4.5-rc1 and one fix that allows to revoke the
administrative MAC that was assigned to VF through the PF.

The VF mac fix needs to go for stable too.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx4_core: Allow resetting VF admin mac to zero
Jack Morgenstein [Wed, 2 Mar 2016 15:47:46 +0000 (17:47 +0200)]
net/mlx4_core: Allow resetting VF admin mac to zero

The VF administrative mac addresses (stored in the PF driver) are
initialized to zero when the PF driver starts up.

These addresses may be modified in the PF driver through ndo calls
initiated by iproute2 or libvirt.

While we allow the PF/host to change the VF admin mac address from zero
to a valid unicast mac, we do not allow restoring the VF admin mac to
zero. We currently only allow changing this mac to a different unicast mac.

This leads to problems when libvirt scripts are used to deal with
VF mac addresses, and libvirt attempts to revoke the mac so this
host will not use it anymore.

Fix this by allowing resetting a VF administrative MAC back to zero.

Fixes: 8f7ba3ca12f6 ('net/mlx4: Add set VF mac address support')
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Reported-by: Moshe Levi <moshele@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx4_core: Check the correct limitation on VFs for HA mode
Moni Shoua [Wed, 2 Mar 2016 15:47:45 +0000 (17:47 +0200)]
net/mlx4_core: Check the correct limitation on VFs for HA mode

The limit of 63 is only for virtual functions while the actual enforcement
was for VFs plus physical functions, fix that.

Fixes: e57968a10bc1 ('net/mlx4_core: Support the HA mode for SRIOV VFs too')
Signed-off-by: Moni Shoua <monis@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx4_core: Fix lockdep warning in handling of mac/vlan tables
Jack Morgenstein [Wed, 2 Mar 2016 15:47:44 +0000 (17:47 +0200)]
net/mlx4_core: Fix lockdep warning in handling of mac/vlan tables

In the mac and vlan register/unregister/replace functions, the driver locks
the mac table mutex (or vlan table mutex) on both ports.

We move to use mutex_lock_nested() to prevent warnings, such as the one below.

[ 101.828445] =============================================
[ 101.834820] [ INFO: possible recursive locking detected ]
[ 101.841199] 4.5.0-rc2+  #49 Not tainted
[ 101.850251] ---------------------------------------------
[ 101.856621] modprobe/3054 is trying to acquire lock:
[ 101.862514] (&table->mutex#2){+.+.+.}, at: [<ffffffffa079c10e>] __mlx4_register_mac+0x87e/0xa90 [mlx4_core]
[ 101.874598]
[ 101.874598] but task is already holding lock:
[ 101.881703] (&table->mutex#2){+.+.+.}, at: [<ffffffffa079c0f0>] __mlx4_register_mac+0x860/0xa90 [mlx4_core]
[ 101.893776]
[ 101.893776] other info that might help us debug this:
[ 101.901658] Possible unsafe locking scenario:
[ 101.901658]
[ 101.908859] CPU0
[ 101.911923] ----
[ 101.914985] lock(&table->mutex#2);
[ 101.919595] lock(&table->mutex#2);
[ 101.924199]
[ 101.924199] * DEADLOCK *
[ 101.924199]
[ 101.931643] May be due to missing lock nesting notation

Fixes: 5f61385d2ebc ('net/mlx4_core: Keep VLAN/MAC tables mirrored in multifunc HA mode')
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Suggested-by: Doron Tsur <doront@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'mlx5-fixes'
David S. Miller [Wed, 2 Mar 2016 19:37:27 +0000 (14:37 -0500)]
Merge branch 'mlx5-fixes'

Saeed Mahameed says:

====================
Mellanox 100G mlx5 driver fixes

This series has few bug fixes for the mlx5 Ethernet driver.

Eran fixed a locking issue with time-stamping that could cause a
soft-lockup when time-stamping is enabled.

Gal fixed the rx/tx packets/bytes counters returned by the driver to
actually went through the network stack.

Tariq removed a poll CQ optimization which could lead the driver to
stop getting interrupts for some of the rings, and a did also fix to
HW LRO which is currently broken.

He also provided RSS and RX hash fixes for the case of changing the
number of rx rings the RX hash/RSS configuration will be out of sync.

The time stamping fix from Eran is not for -stable as the feature was
only introduced in 4.5 but all of the others are.

Changes fro V0:
- Eran addressed the irqsave/restore comments from "Dave" and fixed them.

This series is generated against net commit 4c0b6eaf373a 'net:
thunderx: Fix for Qset error due to CQ full'
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Provide correct packet/bytes statistics
Gal Pressman [Mon, 29 Feb 2016 19:17:15 +0000 (21:17 +0200)]
net/mlx5e: Provide correct packet/bytes statistics

Using the HW VPort counters for traffic (rx/tx packets/bytes)
statistics is wrong. This is because frames dropped due to steering or
out of buffer will be counted as received. To fix that, we move to use
the packet/bytes accounting done by the driver for what the netdev
reports out.

Fixes: f62b8bb8f2d3 ('net/mlx5: Extend mlx5_core to support [...]')
Signed-off-by: Gal Pressman <galp@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Add rx/tx bytes software counters
Gal Pressman [Mon, 29 Feb 2016 19:17:14 +0000 (21:17 +0200)]
net/mlx5e: Add rx/tx bytes software counters

Sum up rx/tx bytes in software as we do for rx/tx packets, to be reported
in upcoming statistics fix.

Signed-off-by: Gal Pressman <galp@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Correctly handle RSS indirection table when changing number of channels
Tariq Toukan [Mon, 29 Feb 2016 19:17:13 +0000 (21:17 +0200)]
net/mlx5e: Correctly handle RSS indirection table when changing number of channels

Upon changing num_channels, reset the RSS indirection table to
match the new value.

Fixes: 2d75b2bc8a8c ('net/mlx5e: Add ethtool RSS configuration options')
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Fix ethtool RX hash func configuration change
Tariq Toukan [Mon, 29 Feb 2016 19:17:12 +0000 (21:17 +0200)]
net/mlx5e: Fix ethtool RX hash func configuration change

We should modify TIRs explicitly to apply the new RSS configuration.
The light ndo close/open calls do not "refresh" them.

Fixes: 2d75b2bc8a8c ('net/mlx5e: Add ethtool RSS configuration options')
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Fix soft lockup when HW Timestamping is enabled
Eran Ben Elisha [Mon, 29 Feb 2016 19:17:11 +0000 (21:17 +0200)]
net/mlx5e: Fix soft lockup when HW Timestamping is enabled

Readers/Writers lock for SW timecounter was acquired without disabling
interrupts on local CPU.

The problematic scenario:
* HW timestamping is enabled
* Timestamp overflow periodic service task is running on local CPU and
  holding write_lock for SW timecounter
* Completion arrives, triggers interrupt for local CPU.
  Interrupt routine calls napi_schedule(), which triggers rx/tx
  skb process.
  An attempt to read SW timecounter using read_lock is done, which is
  already locked by a writer on the same CPU and cause soft lockup.

Add irqsave/irqrestore for when using the readers/writers lock for
writing.

Fixes: ef9814deafd0 ('net/mlx5e: Add HW timestamping (TS) support')
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Fix LRO modify
Tariq Toukan [Mon, 29 Feb 2016 19:17:10 +0000 (21:17 +0200)]
net/mlx5e: Fix LRO modify

Ethtool LRO enable/disable is broken, as of today we only modify TCP
TIRs in order to apply the requested configuration.

Hardware requires that all TIRs pointing to the same RQ should share the
same LRO configuration. For that all other TIRs' LRO fields must be
modified as well.

Fixes: 5c50368f3831 ('net/mlx5e: Light-weight netdev open/stop')
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Remove wrong poll CQ optimization
Tariq Toukan [Mon, 29 Feb 2016 19:17:09 +0000 (21:17 +0200)]
net/mlx5e: Remove wrong poll CQ optimization

With the MLX5E_CQ_HAS_CQES optimization flag, the following buggy
flow might occur:
- Suppose RX is always busy, TX has a single packet every second.
- We poll a single TX cqe and clear its flag.
- We never arm it again as RX is always busy.
- TX CQ flag is never changed, and new TX cqes are not polled.

We revert this optimization.

Fixes: e586b3b0baee ('net/mlx5: Ethernet Datapath files')
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agor8169: Enable RX_MULTI_EN for RTL_GIGA_MAC_VER_41~48
Chunhao Lin [Wed, 2 Mar 2016 16:59:15 +0000 (00:59 +0800)]
r8169: Enable RX_MULTI_EN for RTL_GIGA_MAC_VER_41~48

For RTL8168G/RTL8168H/RTL8411B/RTL8107E, enable this flag to eliminate
message "AMD-Vi: Event logged [IO_PAGE_FAULT device=01:00.0 domain=0x0002
address=0x0000000000003000 flags=0x0050] in dmesg.

Signed-off-by: Chunhao Lin <hau@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge tag 'mac80211-for-davem-2016-03-02' of git://git.kernel.org/pub/scm/linux/kerne...
David S. Miller [Wed, 2 Mar 2016 18:35:31 +0000 (13:35 -0500)]
Merge tag 'mac80211-for-davem-2016-03-02' of git://git./linux/kernel/git/jberg/mac80211

Johannes Berg says:

====================
Here are a few more fixes for the current cycle:
 * check GCMP encryption vs. fragmentation properly; we'd found
   this problem quite a while ago but waited for the 802.11 spec
   to be updated
 * fix RTS/CTS logic in minstrel_ht
 * fix RX of certain public action frames in AP mode
 * add mac80211_hwsim to MAC80211 in MAINTAINERS, this helps
   the kbuild robot pick up the right tree for it
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'bnx2x-fixes'
David S. Miller [Wed, 2 Mar 2016 18:31:58 +0000 (13:31 -0500)]
Merge branch 'bnx2x-fixes'

Michal Schmidt says:

====================
bnx2x: endianness fixes

this fixes a VLAN crash and some SRIOV bugs in bnx2x observed on ppc64.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobnx2x: fix indentation in bnx2x_sp_task()
Michal Schmidt [Wed, 2 Mar 2016 12:47:11 +0000 (13:47 +0100)]
bnx2x: fix indentation in bnx2x_sp_task()

Fix a case of misleading code indentation.

Signed-off-by: Michal Schmidt <mschmidt@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobnx2x: define event data reserved fields as little-endian
Michal Schmidt [Wed, 2 Mar 2016 12:47:10 +0000 (13:47 +0100)]
bnx2x: define event data reserved fields as little-endian

For consistency with other event data structs and to lessen
the chance of a mistake should one of the reserved fields become
used in the future, define the reserved fields as little-endian.

Signed-off-by: Michal Schmidt <mschmidt@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobnx2x: define fields of struct cfc_del_event_data as little-endian
Michal Schmidt [Wed, 2 Mar 2016 12:47:09 +0000 (13:47 +0100)]
bnx2x: define fields of struct cfc_del_event_data as little-endian

There were no missing endianness conversions in this case, but the
fields of struct cfc_del_event_data should be defined as little-endian
to get rid of the ugly (__force __le32) casts.

Signed-off-by: Michal Schmidt <mschmidt@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobnx2x: access cfc_del_event only if the opcode is CFC_DEL
Michal Schmidt [Wed, 2 Mar 2016 12:47:08 +0000 (13:47 +0100)]
bnx2x: access cfc_del_event only if the opcode is CFC_DEL

It's not really a bug, but it was odd that bnx2x_eq_int() read the
message data as if it were a cfc_del_event regardless of the event type.
It's cleaner to access only the appropriate member of union event_data
after checking the event opcode.

Signed-off-by: Michal Schmidt <mschmidt@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobnx2x: fix receive of VF->PF mailbox messages by the PF on big-endian
Michal Schmidt [Wed, 2 Mar 2016 12:47:07 +0000 (13:47 +0100)]
bnx2x: fix receive of VF->PF mailbox messages by the PF on big-endian

On ppc64 the PF did not receive messages from VFs correctly.

Fields of struct vf_pf_event_data are little-endian.

Signed-off-by: Michal Schmidt <mschmidt@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobnx2x: fix sending VF->PF messages on big-endian
Michal Schmidt [Wed, 2 Mar 2016 12:47:06 +0000 (13:47 +0100)]
bnx2x: fix sending VF->PF messages on big-endian

When a VF is sending a message to the PF, it needs to trigger the PF
to tell it the message is ready.
The trigger did not work on ppc64. No interrupt appeared in the PF.

The bug is due to confusion about the layout of struct trigger_vf_zone.
In bnx2x_send_msg2pf() the trigger is written using writeb(), not
writel(), so the attempt to define the struct with a reversed layout on
big-endian is counter-productive.

Signed-off-by: Michal Schmidt <mschmidt@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobnx2x: fix crash on big-endian when adding VLAN
Michal Schmidt [Wed, 2 Mar 2016 12:47:05 +0000 (13:47 +0100)]
bnx2x: fix crash on big-endian when adding VLAN

bnx2x crashes during the initialization of the 8021q module on ppc64.
The bug is a missing conversion from le32 in
bnx2x_handle_classification_eqe() when obtaining the cid value from
struct eth_event_data.

The fields in struct eth_event_data should all be declared as
little-endian and conversions added where missing.

Signed-off-by: Michal Schmidt <mschmidt@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomac80211_hwsim: treat as part of mac80211 for MAINTAINERS
Johannes Berg [Fri, 26 Feb 2016 09:50:34 +0000 (10:50 +0100)]
mac80211_hwsim: treat as part of mac80211 for MAINTAINERS

Since I maintain this driver as part of mac80211, add it to
the file list for mac80211; this helps submitters send it to
me instead of Kalle and also makes the build robot apply the
patches for it on the right tree for build attempts.

Acked-by: Kalle Valo <kvalo@codeaurora.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
8 years agoof_mdio: fix kernel-doc for of_phy_connect()
Sergei Shtylyov [Sun, 28 Feb 2016 19:50:59 +0000 (22:50 +0300)]
of_mdio: fix kernel-doc for of_phy_connect()

The 'flags' parameter of the of_phy_connect() function wasn't described
in  the kernel-doc comment...

Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agosctp: sctp_remaddr_seq_show use the wrong variable to dump transport info
Xin Long [Sun, 28 Feb 2016 02:33:11 +0000 (10:33 +0800)]
sctp: sctp_remaddr_seq_show use the wrong variable to dump transport info

Now in sctp_remaddr_seq_show(), we use variable *tsp to get the param *v.
but *tsp is also used to traversal transport_addr_list, which will cover
the previous value, and make sctp_transport_put work on the wrong transport.

So fix it by adding a new variable to get the param *v.

Fixes: fba4c330c5b9 ("sctp: hold transport before we access t->asoc in sctp proc")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agosctp: lack the check for ports in sctp_v6_cmp_addr
Xin Long [Sun, 28 Feb 2016 02:03:51 +0000 (10:03 +0800)]
sctp: lack the check for ports in sctp_v6_cmp_addr

As the member .cmp_addr of sctp_af_inet6, sctp_v6_cmp_addr should also check
the port of addresses, just like sctp_v4_cmp_addr, cause it's invoked by
sctp_cmp_addr_exact().

Now sctp_v6_cmp_addr just check the port when two addresses have different
family, and lack the port check for two ipv6 addresses. that will make
sctp_hash_cmp() cannot work well.

so fix it by adding ports comparison in sctp_v6_cmp_addr().

Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'phy-micrel-fixes'
David S. Miller [Tue, 1 Mar 2016 21:52:09 +0000 (16:52 -0500)]
Merge branch 'phy-micrel-fixes'

Alexandre Belloni says:

====================
phy: micrel: fix issues with interrupt on atmel boards

Since the phy is not polled anymore, there were issues getting a link on the
sama5d* xplained boards.

I'm not too sure about were those fixes should go and I'm wondering whether the
first one shoud be made generic.

For the second one, I found the PHY_HAS_MAGICANEG flag that is not used and I
wondering whether this is related to that kind of issue. I had a quick look at
the history and could'nt find its use.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agophy: micrel: Disable auto negotiation on startup
Alexandre Belloni [Fri, 26 Feb 2016 18:18:23 +0000 (19:18 +0100)]
phy: micrel: Disable auto negotiation on startup

Disable auto negotiation on init to properly detect an already plugged
cable at boot.

At boot, when the phy is started, it is in the PHY_UP state.
However, if a cable is plugged at boot, because auto negociation is already
enabled at the time we get the first interrupt, the phy is already running.
But the state machine then switches from PHY_UP to PHY_AN and calls
phy_start_aneg(). phy_start_aneg() will not do anything because aneg is
already enabled on the phy. It will then wait for a interrupt before going
further. This interrupt will never happen unless the cable is unplugged and
then replugged.

It was working properly before 321beec5047a (net: phy: Use interrupts when
available in NOLINK state) because switching to NOLINK meant starting
polling the phy, even if IRQ were enabled.

Fixes: 321beec5047a (net: phy: Use interrupts when available in NOLINK state)
Signed-off-by: Alexandre Belloni <alexandre.belloni@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agophy: micrel: Ensure interrupts are reenabled on resume
Alexandre Belloni [Fri, 26 Feb 2016 18:18:22 +0000 (19:18 +0100)]
phy: micrel: Ensure interrupts are reenabled on resume

At least on ksz8081, when getting back from power down, interrupts are
disabled. ensure they are reenabled if they were previously enabled.

This fixes resuming which is failing on the xplained boards from atmel
since 321beec5047a (net: phy: Use interrupts when available in NOLINK
state)

Fixes: 321beec5047a (net: phy: Use interrupts when available in NOLINK state)
Signed-off-by: Alexandre Belloni <alexandre.belloni@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoppp: lock ppp->flags in ppp_read() and ppp_poll()
Guillaume Nault [Fri, 26 Feb 2016 17:45:34 +0000 (18:45 +0100)]
ppp: lock ppp->flags in ppp_read() and ppp_poll()

ppp_read() and ppp_poll() can be called concurrently with ppp_ioctl().
In this case, ppp_ioctl() might call ppp_ccp_closed(), which may update
ppp->flags while ppp_read() or ppp_poll() is reading it.
The update done by ppp_ccp_closed() isn't atomic due to the bit mask
operation ('ppp->flags &= ~(SC_CCP_OPEN | SC_CCP_UP)'), so concurrent
readers might get transient values.
Reading incorrect ppp->flags may disturb the 'ppp->flags & SC_LOOP_TRAFFIC'
test in ppp_read() and ppp_poll(), which in turn can lead to improper
decision on whether the PPP unit file is ready for reading or not.

Since ppp_ccp_closed() is protected by the Rx and Tx locks (with
ppp_lock()), taking the Rx lock is enough for ppp_read() and ppp_poll()
to guarantee that ppp_ccp_closed() won't update ppp->flags
concurrently.

The same reasoning applies to ppp->n_channels. The 'n_channels' field
can also be written to concurrently by ppp_ioctl() (through
ppp_connect_channel() or ppp_disconnect_channel()). These writes aren't
atomic (simple increment/decrement), but are protected by both the Rx
and Tx locks (like in the ppp->flags case). So holding the Rx lock
before reading ppp->n_channels also prevents concurrent writes.

Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobna: fix list corruption
Ivan Vecera [Fri, 26 Feb 2016 08:15:50 +0000 (09:15 +0100)]
bna: fix list corruption

Use list_move_tail() to move MAC address entry from list of pending
to list of active entries. Simple list_add_tail() leaves the entry
also in the first list, this leads to list corruption.

Cc: Rasesh Mody <rasesh.mody@qlogic.com>
Signed-off-by: Ivan Vecera <ivecera@redhat.com>
Acked-by: Rasesh Mody <rasesh.mody@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge tag 'linux-can-fixes-for-4.5-20160226' of git://git.kernel.org/pub/scm/linux...
David S. Miller [Tue, 1 Mar 2016 20:10:38 +0000 (15:10 -0500)]
Merge tag 'linux-can-fixes-for-4.5-20160226' of git://git./linux/kernel/git/mkl/linux-can

Marc Kleine-Budde says:

====================
pull-request: can 2016-02-26

this is a pull request of one patch for net.

The patch by Maximilain Schneider fixes a kfree() problem during disconnect in
the gs_usb driver.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomac80211: minstrel_ht: fix a logic error in RTS/CTS handling
Felix Fietkau [Wed, 24 Feb 2016 11:07:17 +0000 (12:07 +0100)]
mac80211: minstrel_ht: fix a logic error in RTS/CTS handling

RTS/CTS needs to be enabled if the rate is a fallback rate *or* if it's
a dual-stream rate and the sta is in dynamic SMPS mode.

Cc: stable@vger.kernel.org
Fixes: a3ebb4e1b763 ("mac80211: minstrel_ht: handle peers in dynamic SMPS")
Reported-by: Matías Richart <mrichart@fing.edu.uy>
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
8 years agomac80211: Fix Public Action frame RX in AP mode
Jouni Malinen [Mon, 29 Feb 2016 22:29:00 +0000 (00:29 +0200)]
mac80211: Fix Public Action frame RX in AP mode

Public Action frames use special rules for how the BSSID field (Address
3) is set. A wildcard BSSID is used in cases where the transmitter and
recipient are not members of the same BSS. As such, we need to accept
Public Action frames with wildcard BSSID.

Commit db8e17324553 ("mac80211: ignore frames between TDLS peers when
operating as AP") added a rule that drops Action frames to TDLS-peers
based on an Action frame having different DA (Address 1) and BSSID
(Address 3) values. This is not correct since it misses the possibility
of BSSID being a wildcard BSSID in which case the Address 1 would not
necessarily match.

Fix this by allowing mac80211 to accept wildcard BSSID in an Action
frame when in AP mode.

Fixes: db8e17324553 ("mac80211: ignore frames between TDLS peers when operating as AP")
Cc: stable@vger.kernel.org
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
8 years agomac80211: check PN correctly for GCMP-encrypted fragmented MPDUs
Johannes Berg [Fri, 26 Feb 2016 21:13:40 +0000 (22:13 +0100)]
mac80211: check PN correctly for GCMP-encrypted fragmented MPDUs

Just like for CCMP we need to check that for GCMP the fragments
have PNs that increment by one; the spec was updated to fix this
security issue and now has the following text:

The receiver shall discard MSDUs and MMPDUs whose constituent
MPDU PN values are not incrementing in steps of 1.

Adapt the code for CCMP to work for GCMP as well, luckily the
relevant fields already alias each other so no code duplication
is needed (just check the aliasing with BUILD_BUG_ON.)

Cc: stable@vger.kernel.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
8 years ago3c59x: mask LAST_FRAG bit from length field in ring
Neil Horman [Thu, 25 Feb 2016 18:02:50 +0000 (13:02 -0500)]
3c59x: mask LAST_FRAG bit from length field in ring

Recently, I fixed a bug in 3c59x:

commit 6e144419e4da11a9a4977c8d899d7247d94ca338
Author: Neil Horman <nhorman@tuxdriver.com>
Date:   Wed Jan 13 12:43:54 2016 -0500

    3c59x: fix another page map/single unmap imbalance

Which correctly rebalanced dma mapping and unmapping types.  Unfortunately it
introduced a new bug which causes oopses on older systems.

When mapping dma regions, the last entry for a packet in the 3c59x tx ring
encodes a LAST_FRAG bit, which is encoded as the high order bit of the buffers
length field.  When it is unmapped the LAST_FRAG bit is cleared prior to being
passed to the unmap function.  Unfortunately the commit above fails to do that
masking.  It was missed in testing because the system on which I tested it had
an intel iommu, the driver for which ignores the size field, using only the DMA
address as the token to identify the mapping to be released.  However, on older
systems that rely on swiotlb (or other dma drivers that key off that length
field), not masking off that LAST_FRAG high order bit results in parsing a huge
size to be release, leading to all sorts of odd corruptions and the like.

Fix is easy, just mask the length with 0xFFF.  It should really be
&(LAST_FRAG-1), but 0xFFF is the style of the file, and I'd like to make this
fix minimal and correct before making it prettier.

Appies to the net tree cleanly.  All testing on both iommu and swiommu based
systems produce good results

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
CC: Steffen Klassert <klassert@mathematik.tu-chemnitz.de>
CC: "David S. Miller" <davem@davemloft.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ndo_fdb_dump should report -EMSGSIZE to rtnl_fdb_dump.
MINOURA Makoto / 箕浦 真 [Thu, 25 Feb 2016 05:20:48 +0000 (14:20 +0900)]
net: ndo_fdb_dump should report -EMSGSIZE to rtnl_fdb_dump.

When the send skbuff reaches the end, nlmsg_put and friends returns
-EMSGSIZE but it is silently thrown away in ndo_fdb_dump. It is called
within a for_each_netdev loop and the first fdb entry of a following
netdev could fit in the remaining skbuff.  This breaks the mechanism
of cb->args[0] and idx to keep track of the entries that are already
dumped, which results missing entries in bridge fdb show command.

Signed-off-by: Minoura Makoto <minoura@valinux.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agossb: host_soc depends on sprom
Arnd Bergmann [Thu, 25 Feb 2016 20:57:43 +0000 (21:57 +0100)]
ssb: host_soc depends on sprom

Drivers that use the SSB sprom functionality typically 'select SSB_SPROM'
from Kconfig, but CONFIG_SSB_HOST_SOC misses this, which results in
a build failure unless at least one of the other drivers that selects
it is enabled:

drivers/built-in.o: In function `ssb_host_soc_get_invariants':
(.text+0x459494): undefined reference to `ssb_fill_sprom_with_fallback'

This adds the same select statement that is used elsewhere.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: 541c9a84cd85 ("ssb: pick SoC invariants code from MIPS BCM47xx arch")
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agoMerge tag 'iwlwifi-for-kalle-2016-02-25' of https://git.kernel.org/pub/scm/linux...
Kalle Valo [Fri, 26 Feb 2016 10:43:40 +0000 (12:43 +0200)]
Merge tag 'iwlwifi-for-kalle-2016-02-25' of https://git./linux/kernel/git/iwlwifi/iwlwifi-fixes

Two fixes for 4.5:
* We forgot to free the paging memory (Matti)
* Fix the frames in flight accounting (Liad)

8 years agocan: gs_usb: fixed disconnect bug by removing erroneous use of kfree()
Maximilain Schneider [Tue, 23 Feb 2016 01:17:28 +0000 (01:17 +0000)]
can: gs_usb: fixed disconnect bug by removing erroneous use of kfree()

gs_destroy_candev() erroneously calls kfree() on a struct gs_can *, which is
allocated through alloc_candev() and should instead be freed using
free_candev() alone.

The inappropriate use of kfree() causes the kernel to hang when
gs_destroy_candev() is called.

Only the struct gs_usb * which is allocated through kzalloc() should be freed
using kfree() when the device is disconnected.

Signed-off-by: Maximilian Schneider <max@schneidersoft.net>
Cc: linux-stable <stable@vger.kernel.org>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
8 years agonet: thunderx: Fix for Qset error due to CQ full
Sunil Goutham [Wed, 24 Feb 2016 11:10:50 +0000 (16:40 +0530)]
net: thunderx: Fix for Qset error due to CQ full

On Thunderx pass 1.x and pass2 due to a HW errata default CQ
DROP_LEVEL of 0x80 is not sufficient to avoid CQ_WR_FULL Qset
error when packets are being received at >20Mpps resulting in
complete stall of packet reception.

This patch will configure it to 0x100 which is what is expected
by HW on Thunderx. On future passes of thunderx and other chips
HW default/reset value will be 0x100 or higher hence not overwritten.

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Signed-off-by: Sunil Goutham <sgoutham@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agor8169:fix "rtl_counters_cond == 1 (loop: 1000, delay: 10)" log spam.
Chun-Hao Lin [Wed, 24 Feb 2016 06:18:42 +0000 (14:18 +0800)]
r8169:fix "rtl_counters_cond == 1 (loop: 1000, delay: 10)" log spam.

There will be a log spam when there is no cable plugged. Please refer to
following links. https://bugzilla.kernel.org/show_bug.cgi?id=104351
https://bugzilla.kernel.org/show_bug.cgi?id=107421

This issue is caused by runtime power management. When there is no cable
plugged, the driver will be suspend (runtime suspend) by OS and NIC will be
put into the D3 state. During this time, if OS call rtl8169_get_stats64()
to dump tally counter, because NIC is in D3 state, the register value read
by driver will return all 0xff. This will let driver think tally counter
flag is not toggled and then sends the warning message "rtl_counters_cond
== 1 (loop: 1000, delay: 10)" to kernel log.

For fixing this issue, 1.add checking driver's pm runtime status in
rtl8169_get_stats64(). 2.dump tally counter before going runtime suspend
for counter accuracy in runtime suspend.

Signed-off-by: Chunhao Lin <hau@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: fix bridge multicast packet checksum validation
Linus Lüssing [Wed, 24 Feb 2016 03:21:42 +0000 (04:21 +0100)]
net: fix bridge multicast packet checksum validation

We need to update the skb->csum after pulling the skb, otherwise
an unnecessary checksum (re)computation can ocure for IGMP/MLD packets
in the bridge code. Additionally this fixes the following splats for
network devices / bridge ports with support for and enabled RX checksum
offloading:

[...]
[   43.986968] eth0: hw csum failure
[   43.990344] CPU: 3 PID: 0 Comm: swapper/3 Not tainted 4.4.0 #2
[   43.996193] Hardware name: BCM2709
[   43.999647] [<800204e0>] (unwind_backtrace) from [<8001cf14>] (show_stack+0x10/0x14)
[   44.007432] [<8001cf14>] (show_stack) from [<801ab614>] (dump_stack+0x80/0x90)
[   44.014695] [<801ab614>] (dump_stack) from [<802e4548>] (__skb_checksum_complete+0x6c/0xac)
[   44.023090] [<802e4548>] (__skb_checksum_complete) from [<803a055c>] (ipv6_mc_validate_checksum+0x104/0x178)
[   44.032959] [<803a055c>] (ipv6_mc_validate_checksum) from [<802e111c>] (skb_checksum_trimmed+0x130/0x188)
[   44.042565] [<802e111c>] (skb_checksum_trimmed) from [<803a06e8>] (ipv6_mc_check_mld+0x118/0x338)
[   44.051501] [<803a06e8>] (ipv6_mc_check_mld) from [<803b2c98>] (br_multicast_rcv+0x5dc/0xd00)
[   44.060077] [<803b2c98>] (br_multicast_rcv) from [<803aa510>] (br_handle_frame_finish+0xac/0x51c)
[...]

Fixes: 9afd85c9e455 ("net: Export IGMP/MLD message validation code")
Reported-by: Álvaro Fernández Rojas <noltari@gmail.com>
Signed-off-by: Linus Lüssing <linus.luessing@c0d3.blue>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: qca_spi: clear IFF_TX_SKB_SHARING
Stefan Wahren [Tue, 23 Feb 2016 19:23:24 +0000 (19:23 +0000)]
net: qca_spi: clear IFF_TX_SKB_SHARING

ether_setup sets IFF_TX_SKB_SHARING but this is not supported by
qca_spi as it modifies the skb on xmit.

Signed-off-by: Stefan Wahren <stefan.wahren@i2se.com>
Fixes: 291ab06ecf67 (net: qualcomm: new Ethernet over SPI driver for QCA7000)
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: qca_spi: Don't clear IFF_BROADCAST
Stefan Wahren [Tue, 23 Feb 2016 19:23:23 +0000 (19:23 +0000)]
net: qca_spi: Don't clear IFF_BROADCAST

Currently qcaspi_netdev_setup accidentally clears IFF_BROADCAST.
So fix this by keeping the flags from ether_setup.

Reported-by: Michael Heimpold <michael.heimpold@i2se.com>
Signed-off-by: Stefan Wahren <stefan.wahren@i2se.com>
Fixes: 291ab06ecf67 (net: qualcomm: new Ethernet over SPI driver for QCA7000)
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: vrf: Remove direct access to skb->data
David Ahern [Tue, 23 Feb 2016 18:10:26 +0000 (10:10 -0800)]
net: vrf: Remove direct access to skb->data

Nik pointed that the VRF driver should be using skb_header_pointer
instead of accessing skb->data and bits beyond directly which can
be garbage.

Fixes: 35402e313663 ("net: Add IPv6 support to VRF device")
Cc: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: jme: fix suspend/resume on JMC260
Diego Viola [Tue, 23 Feb 2016 15:04:04 +0000 (12:04 -0300)]
net: jme: fix suspend/resume on JMC260

The JMC260 network card fails to suspend/resume because the call to
jme_start_irq() was too early, moving the call to jme_start_irq() after
the call to jme_reset_link() makes it work.

Prior this change suspend/resume would fail unless /sys/power/pm_async=0
was explicitly specified.

Relevant bug report: https://bugzilla.kernel.org/show_bug.cgi?id=112351

Signed-off-by: Diego Viola <diego.viola@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge tag 'mac80211-for-davem-2016-02-23' of git://git.kernel.org/pub/scm/linux/kerne...
David S. Miller [Wed, 24 Feb 2016 21:58:12 +0000 (16:58 -0500)]
Merge tag 'mac80211-for-davem-2016-02-23' of git://git./linux/kernel/git/jberg/mac80211

Johannes Berg says:

====================
Another small set of fixes:
 * stop critical protocol session on disconnect to avoid
   it getting stuck
 * wext: fix two RTNL message ordering issues
 * fix an uninitialized value (found by KASAN)
 * fix an out-of-bounds access (also found by KASAN)
 * clear connection keys when freeing them in all cases
   (IBSS, all other places already did so)
 * fix expected throughput unit to get consistent values
 * set default TX aggregation timeout to 0 in minstrel
   to avoid (really just hide) issues and perform better
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobpf: fix csum setting for bpf_set_tunnel_key
Daniel Borkmann [Tue, 23 Feb 2016 01:05:26 +0000 (02:05 +0100)]
bpf: fix csum setting for bpf_set_tunnel_key

The fix in 35e2d1152b22 ("tunnels: Allow IPv6 UDP checksums to be correctly
controlled.") changed behavior for bpf_set_tunnel_key() when in use with
IPv6 and thus uncovered a bug that TUNNEL_CSUM needed to be set but wasn't.
As a result, the stack dropped ingress vxlan IPv6 packets, that have been
sent via eBPF through collect meta data mode due to checksum now being zero.

Since after LCO, we enable IPv4 checksum by default, so make that analogous
and only provide a flag BPF_F_ZERO_CSUM_TX for the user to turn it off in
IPv4 case.

Fixes: 35e2d1152b22 ("tunnels: Allow IPv6 UDP checksums to be correctly controlled.")
Fixes: c6c33454072f ("bpf: support ipv6 for bpf_skb_{set,get}_tunnel_key")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoipv4: only create late gso-skb if skb is already set up with CHECKSUM_PARTIAL
Hannes Frederic Sowa [Mon, 22 Feb 2016 17:43:25 +0000 (18:43 +0100)]
ipv4: only create late gso-skb if skb is already set up with CHECKSUM_PARTIAL

Otherwise we break the contract with GSO to only pass CHECKSUM_PARTIAL
skbs down. This can easily happen with UDP+IPv4 sockets with the first
MSG_MORE write smaller than the MTU, second write is a sendfile.

Returning -EOPNOTSUPP lets the callers fall back into normal sendmsg path,
were we calculate the checksum manually during copying.

Commit d749c9cbffd6 ("ipv4: no CHECKSUM_PARTIAL on MSG_MORE corked
sockets") started to exposes this bug.

Fixes: d749c9cbffd6 ("ipv4: no CHECKSUM_PARTIAL on MSG_MORE corked sockets")
Reported-by: Jiri Benc <jbenc@redhat.com>
Cc: Jiri Benc <jbenc@redhat.com>
Reported-by: Wakko Warner <wakko@animx.eu.org>
Cc: Wakko Warner <wakko@animx.eu.org>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: nb8800: support fixed-link DT node
Sebastian Frias [Mon, 22 Feb 2016 12:33:14 +0000 (13:33 +0100)]
net: ethernet: nb8800: support fixed-link DT node

Under some circumstances, e.g. when connecting to a switch, the ethernet
port will not be connected to a PHY. In that case a "fixed-link" DT node
can be used to replace it.

https://stackoverflow.com/questions/31046172/device-tree-for-phy-less-connection-to-a-dsa-switch

This patch adds support for the "fixed-link" node to the nb8800 driver.

Signed-off-by: Sebastian Frias <sf84@laposte.net>
Acked-by: Mans Rullgard <mans@mansr.com>
Cc: Mason <slash.tmp@free.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobnxt_en: Fix zero padding of tx push data.
Michael Chan [Mon, 22 Feb 2016 07:10:26 +0000 (02:10 -0500)]
bnxt_en: Fix zero padding of tx push data.

The arithmetic to zero pad the last 64-bit word in the push buffer is not
correct.

1. It should be pdata + length to get to the end.
2. 'pdata' is void pointer and passing it to PTR_ALIGN() will cast the
aligned pointer to void.  Pass 'end' which is u64 pointer to PTR_ALIGN()
instead so that the aligned pointer - 1 is the last 64-bit pointer to data.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agotunnel: Clear IPCB(skb)->opt before dst_link_failure called
Bernie Harris [Sun, 21 Feb 2016 23:58:05 +0000 (12:58 +1300)]
tunnel: Clear IPCB(skb)->opt before dst_link_failure called

IPCB may contain data from previous layers (in the observed case the
qdisc layer). In the observed scenario, the data was misinterpreted as
ip header options, which later caused the ihl to be set to an invalid
value (<5). This resulted in an infinite loop in the mips implementation
of ip_fast_csum.

This patch clears IPCB(skb)->opt before dst_link_failure can be called for
various types of tunnels. This change only applies to encapsulated ipv4
packets.

The code introduced in 11c21a30 which clears all of IPCB has been removed
to be consistent with these changes, and instead the opt field is cleared
unconditionally in ip_tunnel_xmit. The change in ip_tunnel_xmit applies to
SIT, GRE, and IPIP tunnels.

The relevant vti, l2tp, and pptp functions already contain similar code for
clearing the IPCB.

Signed-off-by: Bernie Harris <bernie.harris@alliedtelesis.co.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agotcp: convert cached rtt from usec to jiffies when feeding initial rto
Konstantin Khlebnikov [Sun, 21 Feb 2016 07:12:39 +0000 (10:12 +0300)]
tcp: convert cached rtt from usec to jiffies when feeding initial rto

Currently it's converted into msecs, thus HZ=1000 intact.

Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Fixes: 740b0f1841f6 ("tcp: switch rtt estimations to usec resolution")
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoAdd Dell Wireless 5809e Gobi 4G HSPA+ Mobile Broadband Card (rev3) to qmi_wwan
Patrik Halfar [Sat, 20 Feb 2016 17:49:40 +0000 (18:49 +0100)]
Add Dell Wireless 5809e Gobi 4G HSPA+ Mobile Broadband Card (rev3) to qmi_wwan

New revison of Dell Wireless 5809e Gobi 4G HSPA+ Mobile Broadband Card has new idProduct

Bus 002 Device 006: ID 413c:81b3 Dell Computer Corp.
Device Descriptor:
  bLength                18
  bDescriptorType         1
  bcdUSB               2.00
  bDeviceClass            0
  bDeviceSubClass         0
  bDeviceProtocol         0
  bMaxPacketSize0        64
  idVendor           0x413c Dell Computer Corp.
  idProduct          0x81b3
  bcdDevice            0.06
  iManufacturer           1 Sierra Wireless, Incorporated
  iProduct                2 Dell Wireless 5809e Gobi™ 4G HSPA+ Mobile Broadband Card
  iSerial                 3
  bNumConfigurations      2

Signed-off-by: Patrik Halfar <patrik_halfar@halfarit.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoiwlwifi: mvm: Fix paging memory leak
Matti Gottlieb [Sun, 14 Feb 2016 15:05:39 +0000 (17:05 +0200)]
iwlwifi: mvm: Fix paging memory leak

If the opmode is stopped and started again we did not free
the paging buffers. Fix that.
In addition when freeing the firmware's paging download
buffer, set the pointer to NULL.

Signed-off-by: Matti Gottlieb <matti.gottlieb@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
8 years agocfg80211: stop critical protocol session upon disconnect event
Arend van Spriel [Mon, 15 Feb 2016 13:35:53 +0000 (14:35 +0100)]
cfg80211: stop critical protocol session upon disconnect event

When user-space has started a critical protocol session and a disconnect
event occurs, the rdev::crit_prot_nlportid remains set. This caused a
subsequent NL80211_CMD_CRIT_PROTO_START to fail (-EBUSY). Fix this by
clearing the rdev attribute and call .crit_proto_stop() callback upon
disconnect event.

Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
8 years agonl80211: Zero out the connection keys memory when freeing them.
Ola Olsson [Thu, 11 Feb 2016 00:00:22 +0000 (01:00 +0100)]
nl80211: Zero out the connection keys memory when freeing them.

The connection keys are zeroed out in all other cases except this
one. Let's fix the last one as well.

Signed-off-by: Ola Olsson <ola.olsson@sonymobile.com>
Reviewed-by: Julian Calaby <julian.calaby@gmail.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
8 years agomac80211: minstrel_ht: set default tx aggregation timeout to 0
Felix Fietkau [Thu, 18 Feb 2016 18:49:18 +0000 (19:49 +0100)]
mac80211: minstrel_ht: set default tx aggregation timeout to 0

The value 5000 was put here with the addition of the timeout field to
ieee80211_start_tx_ba_session. It was originally added in mac80211 to
save resources for drivers like iwlwifi, which only supports a limited
number of concurrent aggregation sessions.

Since iwlwifi does not use minstrel_ht and other drivers don't need
this, 0 is a better default - especially since there have been
recent reports of aggregation setup related issues reproduced with
ath9k. This should improve stability without causing any adverse
effects.

Cc: stable@vger.kernel.org
Acked-by: Avery Pennarun <apenwarr@gmail.com>
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
8 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Mon, 22 Feb 2016 20:18:07 +0000 (12:18 -0800)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:
 "Looks like a lot, but mostly driver fixes scattered all over as usual.

  Of note:

   1) Add conditional sched in nf conntrack in cleanup to avoid NMI
      watchdogs.  From Florian Westphal.

   2) Fix deadlock in nfnetlink cttimeout, also from Floarian.

   3) Fix handling of slaves in bonding ARP monitor validation, from Jay
      Vosburgh.

   4) Callers of ip_cmsg_send() are responsible for freeing IP options,
      some were not doing so.  Fix from Eric Dumazet.

   5) Fix per-cpu bugs in mvneta driver, from Gregory CLEMENT.

   6) Fix vlan handling in mv88e6xxx DSA driver, from Vivien Didelot.

   7) bcm7xxx PHY driver bug fixes from Florian Fainelli.

   8) Avoid unaligned accesses to protocol headers wrt.  GRE, from
      Alexander Duyck.

   9) SKB leaks and other problems in arc_emac driver, from Alexander
      Kochetkov.

  10) tcp_v4_inbound_md5_hash() releases listener socket instead of
      request socket on error path, oops.  Fix from Eric Dumazet.

  11) Missing socket release in pppoe_rcv_core() that seems to have
      existed basically forever.  From Guillaume Nault.

  12) Missing slave_dev unregister in dsa_slave_create() error path,
      from Florian Fainelli.

  13) crypto_alloc_hash() never returns NULL, fix return value check in
      __tcp_alloc_md5sig_pool.  From Insu Yun.

  14) Properly expire exception route entries in ipv4, from Xin Long.

  15) Fix races in tcp/dccp listener socket dismantle, from Eric
      Dumazet.

  16) Don't set IFF_TX_SKB_SHARING in vxlan, geneve, or GRE, it's not
      legal.  These drivers modify the SKB on transmit.  From Jiri Benc.

  17) Fix regression in the initialziation of netdev->tx_queue_len.
      From Phil Sutter.

  18) Missing unlock in tipc_nl_add_bc_link() error path, from Insu Yun.

  19) SCTP port hash sizing does not properly ensure that table is a
      power of two in size.  From Neil Horman.

  20) Fix initializing of software copy of MAC address in fmvj18x_cs
      driver, from Ken Kawasaki"

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (129 commits)
  bnx2x: Fix 84833 phy command handler
  bnx2x: Fix led setting for 84858 phy.
  bnx2x: Correct 84858 PHY fw version
  bnx2x: Fix 84833 RX CRC
  bnx2x: Fix link-forcing for KR2
  net: ethernet: davicom: fix devicetree irq resource
  fmvj18x_cs: fix incorrect indexing of dev->dev_addr[] when copying the MAC address
  Driver: Vmxnet3: Update Rx ring 2 max size
  net: netcp: rework the code for get/set sw_data in dma desc
  soc: ti: knav_dma: rename pad in struct knav_dma_desc to sw_data
  net: ti: netcp: restore get/set_pad_info() functionality
  MAINTAINERS: Drop myself as xen netback maintainer
  sctp: Fix port hash table size computation
  can: ems_usb: Fix possible tx overflow
  Bluetooth: hci_core: Avoid mixing up req_complete and req_complete_skb
  net: bcmgenet: Fix internal PHY link state
  af_unix: Don't use continue to re-execute unix_stream_read_generic loop
  unix_diag: fix incorrect sign extension in unix_lookup_by_ino
  bnxt_en: Failure to update PHY is not fatal condition.
  bnxt_en: Remove unnecessary call to update PHY settings.
  ...

8 years agoMerge tag 'hwmon-for-linus-v4.5-rc6' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Mon, 22 Feb 2016 20:12:46 +0000 (12:12 -0800)]
Merge tag 'hwmon-for-linus-v4.5-rc6' of git://git./linux/kernel/git/groeck/linux-staging

Pull hwmon fixes from Guenter Roeck:
 "Two fixes headed for stable:

   - Remove an unnecessary speed_index lookup for thermal hook in the
     gpio-fan driver.  The unnecessary speed lookup can hog the system.

   - Handle negative conversion values correctly in the ads1015 driver"

* tag 'hwmon-for-linus-v4.5-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck/linux-staging:
  hwmon: (gpio-fan) Remove un-necessary speed_index lookup for thermal hook
  hwmon: (ads1015) Handle negative conversion values correctly

8 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma
Linus Torvalds [Mon, 22 Feb 2016 20:04:11 +0000 (12:04 -0800)]
Merge tag 'for-linus' of git://git./linux/kernel/git/dledford/rdma

Pull rdma fixes from Doug Ledford:
 "One ocrdma fix:

   - The new CQ API support was added to ocrdma, but they got the arming
     logic wrong, so without this, transfers eventually fail when they
     fail to arm the interrupt properly under load

  Two related fixes for mlx4:

   - When we added the 64bit extended counters support to the core IB
     code, they forgot to update the RoCE side of the mlx4 driver (the
     IB side they properly updated).

     I debated whether or not to include these patches as they could be
     considered feature enablement patches, but the existing code will
     blindy copy the 32bit counters, whether any counters were requested
     at all (a bug).

     These two patches make it (a) check to see that counters were
     requested and (b) copy the right counters (the 64bit support is
     new, the 32bit is not).  For that reason I went ahead and took
     them"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma:
  IB/mlx4: Add support for the port info class for RoCE ports
  IB/mlx4: Add support for extended counters over RoCE ports
  RDMA/ocrdma: Fix arm logic to align with new cq API

8 years agoMerge branch 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa...
Linus Torvalds [Mon, 22 Feb 2016 19:55:18 +0000 (11:55 -0800)]
Merge branch 'i2c/for-current' of git://git./linux/kernel/git/wsa/linux

Pull i2c fixes from Wolfram Sang:
 "Some bugfixes from I2C for you:

  A fix for a RuntimePM regression with OMAP, a fix to enable TCO for
  Lewisburg platforms, and a typo fix while we are here"

* 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  i2c: i801: Adding Intel Lewisburg support for iTCO
  i2c: uniphier: fix typos in error messages
  i2c: omap: Fix PM regression with deferred probe for pm_runtime_reinit

8 years agoMerge tag 'linux-can-fixes-for-4.5-20160221' of git://git.kernel.org/pub/scm/linux...
David S. Miller [Mon, 22 Feb 2016 03:51:55 +0000 (22:51 -0500)]
Merge tag 'linux-can-fixes-for-4.5-20160221' of git://git./linux/kernel/git/mkl/linux-can

Marc Kleine-Budde says:

====================
pull-request: can 2016-02-21

this is a pull reqeust of one patch for net/master.

The patch is by Gerhard Uttenthaler and fixes a potential tx overflow in the
ems_usb driver.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'bnx2x-848xx-phy-fixes'
David S. Miller [Mon, 22 Feb 2016 03:50:55 +0000 (22:50 -0500)]
Merge branch 'bnx2x-848xx-phy-fixes'

Yuval Mintz says:

====================
bnx2x: Fix 848xx phys

This series contains link-related fixes, mostly for the 848xx phys
[2 patches are for 84833, and 2 patches are for 84858].
====================

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