OSDN Git Service

android-x86/kernel.git
6 years agonetem: add uapi to express delay and jitter in nanoseconds
Dave Taht [Wed, 8 Nov 2017 23:12:27 +0000 (15:12 -0800)]
netem: add uapi to express delay and jitter in nanoseconds

netem userspace has long relied on a horrible /proc/net/psched hack
to translate the current notion of "ticks" to nanoseconds.

Expressing latency and jitter instead, in well defined nanoseconds,
increases the dynamic range of emulated delays and jitter in netem.

It will also ease a transition where reducing a tick to nsec
equivalence would constrain the max delay in prior versions of
netem to only 4.3 seconds.

Signed-off-by: Dave Taht <dave.taht@gmail.com>
Suggested-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonetem: convert to qdisc_watchdog_schedule_ns
Dave Taht [Wed, 8 Nov 2017 23:12:26 +0000 (15:12 -0800)]
netem: convert to qdisc_watchdog_schedule_ns

Upgrade the internal netem scheduler to use nanoseconds rather than
ticks throughout.

Convert to and from the std "ticks" userspace api automatically,
while allowing for finer grained scheduling to take place.

Signed-off-by: Dave Taht <dave.taht@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoipv6: try not to take rtnl_lock in ip6mr_sk_done
Francesco Ruggeri [Wed, 8 Nov 2017 19:23:46 +0000 (11:23 -0800)]
ipv6: try not to take rtnl_lock in ip6mr_sk_done

Avoid traversing the list of mr6_tables (which requires the
rtnl_lock) in ip6mr_sk_done(), when we know in advance that
a match will not be found.
This can happen when rawv6_close()/ip6mr_sk_done() is invoked
on non-mroute6 sockets.
This patch helps reduce rtnl_lock contention when destroying
a large number of net namespaces, each having a non-mroute6
raw socket.

v2: same patch, only fixed subject line and expanded comment.

Signed-off-by: Francesco Ruggeri <fruggeri@arista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: realtek: r8169: remove redundant assignment to giga_ctrl
Colin Ian King [Wed, 8 Nov 2017 13:23:23 +0000 (13:23 +0000)]
net: realtek: r8169: remove redundant assignment to giga_ctrl

The variable giga_ctrl is being assigned to zero however this is
never read and hence the assignment is redundant, so remove it.
Cleans up clang warning:

drivers/net/ethernet/realtek/r8169.c:1978:3: warning: Value stored
to 'giga_ctrl' is never read

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: lan9303: Documentation: Add missing word "Mbps"
Egil Hjelmeland [Wed, 8 Nov 2017 10:55:14 +0000 (11:55 +0100)]
net: dsa: lan9303: Documentation: Add missing word "Mbps"

Signed-off-by: Egil Hjelmeland <privat@egil-hjelmeland.no>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: lan9303: Fix lan9303_alr_del_port()
Egil Hjelmeland [Wed, 8 Nov 2017 10:44:36 +0000 (11:44 +0100)]
net: dsa: lan9303: Fix lan9303_alr_del_port()

Fix embarrassing bug in lan9303_alr_del_port(): Instead of zeroing
entr->mac_addr, I destroyed the next cache entry. Affected .port_fdb_del and
.port_mdb_del.

Fixes: 0620427ea0d6 ("net: dsa: lan9303: Add fdb/mdb manipulation")
Signed-off-by: Egil Hjelmeland <privat@egil-hjelmeland.no>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
David S. Miller [Sun, 12 Nov 2017 00:17:05 +0000 (09:17 +0900)]
Merge git://git./linux/kernel/git/davem/net

6 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Sat, 11 Nov 2017 17:10:39 +0000 (09:10 -0800)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) Use after free in vlan, from Cong Wang.

 2) Handle NAPI poll with a zero budget properly in mlx5 driver, from
    Saeed Mahameed.

 3) If DMA mapping fails in mlx5 driver, NULL out page, from Inbar
    Karmy.

 4) Handle overrun in RX FIFO of sun4i CAN driver, from Gerhard
    Bertelsmann.

 5) Missing return in mdb and vlan prepare phase of DSA layer, from
    Vivien Didelot.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net:
  vlan: fix a use-after-free in vlan_device_event()
  net: dsa: return after vlan prepare phase
  net: dsa: return after mdb prepare phase
  can: ifi: Fix transmitter delay calculation
  tcp: fix tcp_fastretrans_alert warning
  tcp: gso: avoid refcount_t warning from tcp_gso_segment()
  can: peak: Add support for new PCIe/M2 CAN FD interfaces
  can: sun4i: handle overrun in RX FIFO
  can: c_can: don't indicate triple sampling support for D_CAN
  net/mlx5e: Increase Striding RQ minimum size limit to 4 multi-packet WQEs
  net/mlx5e: Set page to null in case dma mapping fails
  net/mlx5e: Fix napi poll with zero budget
  net/mlx5: Cancel health poll before sending panic teardown command
  net/mlx5: Loop over temp list to release delay events
  rds: ib: Fix NULL pointer dereference in debug code

6 years agoMerge tag 'wireless-drivers-next-for-davem-2017-11-11' of git://git.kernel.org/pub...
David S. Miller [Sat, 11 Nov 2017 13:37:22 +0000 (22:37 +0900)]
Merge tag 'wireless-drivers-next-for-davem-2017-11-11' of git://git./linux/kernel/git/kvalo/wireless-drivers-next

Kalle Valo says:

====================
wireless-drivers-next patches for 4.15

Last minute patches before the merge window. Not really anything
special standing out, mostly fixes or cleanup and some minor new
features.

Major changes:

iwlwifi

* some new PCI IDs
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: Remove unused skb_shared_info member
Mat Martineau [Fri, 10 Nov 2017 22:03:51 +0000 (14:03 -0800)]
net: Remove unused skb_shared_info member

ip6_frag_id was only used by UFO, which has been removed.
ipv6_proxy_select_ident() only existed to set ip6_frag_id and has no
in-tree callers.

Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'l2tp-avoid-aliasing-tunnels-socket-pointer'
David S. Miller [Sat, 11 Nov 2017 13:08:24 +0000 (22:08 +0900)]
Merge branch 'l2tp-avoid-aliasing-tunnels-socket-pointer'

Guillaume Nault says:

====================
l2tp: avoid aliasing tunnels socket pointer

We don't need to copy the tunnel's socket pointer in the pseudo-wire
specific session structures. This uselessly complicates the code
and hampers evolution.

This series was part of an effort to protect tunnels socket pointer
with RCU. But since it provides nice cleanup, I submit it separately.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agol2tp: remove the .tunnel_sock field from struct pppol2tp_session
Guillaume Nault [Fri, 10 Nov 2017 21:06:37 +0000 (06:06 +0900)]
l2tp: remove the .tunnel_sock field from struct pppol2tp_session

The last user of .tunnel_sock is pppol2tp_connect() which defensively
uses it to verify internal data consistency.

This check isn't necessary: l2tp_session_get() guarantees that the
returned session belongs to the tunnel passed as parameter. And
.tunnel_sock is never updated, so checking that it still points to
the parent tunnel socket is useless; that test can never fail.

Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agol2tp: avoid using ->tunnel_sock for getting session's parent tunnel
Guillaume Nault [Fri, 10 Nov 2017 21:06:31 +0000 (06:06 +0900)]
l2tp: avoid using ->tunnel_sock for getting session's parent tunnel

Sessions don't need to use l2tp_sock_to_tunnel(xxx->tunnel_sock) for
accessing their parent tunnel. They have the .tunnel field in the
l2tp_session structure for that. Furthermore, in all these cases, the
session is registered, so we're guaranteed that .tunnel isn't NULL and
that the session properly holds a reference on the tunnel.

Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agol2tp: remove .tunnel_sock from struct l2tp_eth
Guillaume Nault [Fri, 10 Nov 2017 21:06:29 +0000 (06:06 +0900)]
l2tp: remove .tunnel_sock from struct l2tp_eth

This field has never been used.

Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'dsa-b53-Turn-on-Broadcom-tags'
David S. Miller [Sat, 11 Nov 2017 12:55:16 +0000 (21:55 +0900)]
Merge branch 'dsa-b53-Turn-on-Broadcom-tags'

Florian Fainelli says:

====================
net: dsa: b53: Turn on Broadcom tags

This was long overdue, with this patch series, the b53 driver now
turns on Broadcom tags except for 5325 and 5365 which use an older
format that we do not support yet (TBD).

First patch is necessary in order for bgmac, used on BCM5301X and Northstar
Plus to work correctly and successfully send ARP packets back to the requsester.

Second patch is actually a bug fix, but because net/master and net-next/master
diverge in that area, I am targeting net-next/master here.

Finally, the last patch enables Broadcom tags after checking that the CPU port
selected is either, 5, 7 or 8, since those are the only valid combinations
given currently supported HW.

Changes in v3:

- guarded padding with netdev_uses_dsa() to let the non-DSA use cases
  not have a performance hit for smaller packets

- added missing select NET_DSA_TAG_BRCM to drivers/net/dsa/b53/Kconfig

Changes in v2:

- moved a hunk between patch 2 and patch 3 to avoid a bisectability issue
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: b53: Turn on Broadcom tags
Florian Fainelli [Fri, 10 Nov 2017 19:33:27 +0000 (11:33 -0800)]
net: dsa: b53: Turn on Broadcom tags

Enable Broadcom tags for b53 devices, except 5325 and 5365 which use a
different Broadcom tag format not yet supported by net/dsa/tag_brcm.c.

We also make sure that we can turn on Broadcom tags on a CPU port number
that is capable of that: 5, 7 or 8.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: b53: Stop using dev->cpu_port incorrectly
Florian Fainelli [Fri, 10 Nov 2017 19:33:26 +0000 (11:33 -0800)]
net: dsa: b53: Stop using dev->cpu_port incorrectly

dev->cpu_port is the driver local information that should only be used
to look up register offsets for a particular port, when they differ
(e.g: IMP port override), but it should certainly not be used in place
of the DSA configured CPU port.

Since the DSA switch layer calls port_vlan_{add,del}() on the CPU port
as well, we can remove the specific setting of the CPU port within
port_vlan_{add,del}.

Fixes: ff39c2d68679 ("net: dsa: b53: Add bridge support")
Fixes: 967dd82ffc52 ("net: dsa: b53: Add support for Broadcom RoboSwitch")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: bgmac: Pad packets to a minimum size
Florian Fainelli [Fri, 10 Nov 2017 19:33:25 +0000 (11:33 -0800)]
net: bgmac: Pad packets to a minimum size

In preparation for enabling Broadcom tags with b53, pad packets to a
minimum size of 64 bytes (sans FCS) in order for the Broadcom switch to
accept ingressing frames. Without this, we would typically be able to
DHCP, but not resolve with ARP because packets are too small and get
rejected by the switch.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge tag 'linux-can-fixes-for-4.14-20171110' of git://git.kernel.org/pub/scm/linux...
David S. Miller [Sat, 11 Nov 2017 12:52:01 +0000 (21:52 +0900)]
Merge tag 'linux-can-fixes-for-4.14-20171110' of git://git./linux/kernel/git/mkl/linux-can

Marc Kleine-Budde says:

====================
pull-request: can 2017-11-10

this is a pull request for net/master.

The first patch by Richard Schütz for the c_can driver removes the false
indication to support triple sampling for d_can. Gerhard Bertelsmann's
patch for the sun4i driver improves the RX overrun handling. The patch
by Stephane Grosjean for the peak_canfd driver adds the PCI ids for
various new PCIe/M2 interfaces. Marek Vasut's patch for the ifi driver
fix transmitter delay calculation.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'dsa-lan9303-IGMP-handling'
David S. Miller [Sat, 11 Nov 2017 12:50:15 +0000 (21:50 +0900)]
Merge branch 'dsa-lan9303-IGMP-handling'

Egil Hjelmeland says:

====================
net: dsa: lan9303: IGMP handling

Set up the HW switch to trap IGMP packets to CPU port.
And make sure skb->offload_fwd_mark is cleared for incoming IGMP packets.

skb->offload_fwd_mark calculation is a candidate for consolidation into the
DSA core. The calculation can probably be more polished when done at a point
where DSA has updated skb.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: lan9303: Clear offload_fwd_mark for IGMP
Egil Hjelmeland [Fri, 10 Nov 2017 11:54:35 +0000 (12:54 +0100)]
net: dsa: lan9303: Clear offload_fwd_mark for IGMP

Now that IGMP packets no longer is flooded in HW, we want the SW bridge to
forward packets based on bridge configuration. To make that happen,
IGMP packets must have skb->offload_fwd_mark = 0.

Signed-off-by: Egil Hjelmeland <privat@egil-hjelmeland.no>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: lan9303: Set up trapping of IGMP to CPU port
Egil Hjelmeland [Fri, 10 Nov 2017 11:54:34 +0000 (12:54 +0100)]
net: dsa: lan9303: Set up trapping of IGMP to CPU port

IGMP packets should be trapped to the CPU port. The SW bridge knows
whether to forward to other ports.

With "IGMP snooping for local traffic" merged, IGMP trapping is also
required for stable IGMPv2 operation.

LAN9303 does not trap IGMP packets by default.

Enable IGMP trapping in lan9303_setup.

Signed-off-by: Egil Hjelmeland <privat@egil-hjelmeland.no>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agocxgb4: collect vpd info directly from hardware
Rahul Lakkireddy [Fri, 10 Nov 2017 07:33:37 +0000 (13:03 +0530)]
cxgb4: collect vpd info directly from hardware

Collect vpd information directly from hardware instead of software
adapter context. Move EEPROM physical address to virtual address
translation logic to t4_hw.c and update relevant files.

Fixes: 6f92a6544f1a ("cxgb4: collect hardware misc dumps")
Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge tag 'mlx5-fixes-2017-11-08' of git://git.kernel.org/pub/scm/linux/kernel/git...
David S. Miller [Sat, 11 Nov 2017 10:40:05 +0000 (19:40 +0900)]
Merge tag 'mlx5-fixes-2017-11-08' of git://git./linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
Mellanox, mlx5 fixes 2017-11-08

The following series includes some fixes for mlx5 core and etherent
driver.

Sorry for the late submission but as you can see i have some very
critical fixes below that i would like them merged into this RC.

Please pull and let me know if there is any problem.

For -stable:
('net/mlx5e: Set page to null in case dma mapping fails') kernels >= 4.13
('net/mlx5: FPGA, return -EINVAL if size is zero') kernels >= 4.13
('net/mlx5: Cancel health poll before sending panic teardown command') kernels >= 4.13

V1->V2:
- Fix Reviewed-by tag of the 2nd patch.
- Drop the FPGA 0 size fix, it needs some more change log info.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agovlan: fix a use-after-free in vlan_device_event()
Cong Wang [Fri, 10 Nov 2017 00:43:13 +0000 (16:43 -0800)]
vlan: fix a use-after-free in vlan_device_event()

After refcnt reaches zero, vlan_vid_del() could free
dev->vlan_info via RCU:

RCU_INIT_POINTER(dev->vlan_info, NULL);
call_rcu(&vlan_info->rcu, vlan_info_rcu_free);

However, the pointer 'grp' still points to that memory
since it is set before vlan_vid_del():

        vlan_info = rtnl_dereference(dev->vlan_info);
        if (!vlan_info)
                goto out;
        grp = &vlan_info->grp;

Depends on when that RCU callback is scheduled, we could
trigger a use-after-free in vlan_group_for_each_dev()
right following this vlan_vid_del().

Fix it by moving vlan_vid_del() before setting grp. This
is also symmetric to the vlan_vid_add() we call in
vlan_device_event().

Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Fixes: efc73f4bbc23 ("net: Fix memory leak - vlan_info struct")
Cc: Alexander Duyck <alexander.duyck@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Girish Moodalbail <girish.moodalbail@oracle.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Reviewed-by: Girish Moodalbail <girish.moodalbail@oracle.com>
Tested-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: mv88e6xxx: Fix stats histogram mode
Andrew Lunn [Thu, 9 Nov 2017 23:36:41 +0000 (00:36 +0100)]
net: dsa: mv88e6xxx: Fix stats histogram mode

The statistics histogram mode was not being explicitly initialized on
devices other than the 6390 family. Clearing the statistics then
overwrote the default setting, setting the histogram to a reserved
mode.

Explicitly set the histogram mode for all devices. Change the
statistics clear into a read/modify/write, and since it is now more
complex, move it into global1.c.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'mv88e6xxx-broadcast-flooding-in-hardware'
David S. Miller [Sat, 11 Nov 2017 10:33:11 +0000 (19:33 +0900)]
Merge branch 'mv88e6xxx-broadcast-flooding-in-hardware'

Andrew Lunn says:

====================
mv88e6xxx broadcast flooding in hardware

This patchset makes the mv88e6xxx driver perform flooding in hardware,
rather than let the software bridge perform the flooding. This is a
prerequisite for IGMP snooping on the bridge interface.

In order to make hardware broadcasting work, a few other issues need
fixing or improving. SWITCHDEV_ATTR_ID_PORT_PARENT_ID is broken, which
is apparent when testing on the ZII devel board with multiple
switches.

Some of these patches are taken from a previous RFC patchset of IGMP
support.

Rebased onto net-next, with fixup for Vivien's refactoring.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: mv88e6xxx: Flood broadcast frames in hardware
Andrew Lunn [Thu, 9 Nov 2017 21:29:56 +0000 (22:29 +0100)]
net: dsa: mv88e6xxx: Flood broadcast frames in hardware

By default, the switch does not flood broadcast frames. Instead the
broadcast address is unknown in the ATU, so the frame gets forwarded
out the cpu port. The software bridge then floods it back to the
individual switch ports which are members of the bridge.

Add an ATU entry in the switch so that it floods broadcast frames out
ports, rather than have the software bridge do it. Also, send a copy
out the cpu port and any dsa ports. Rely on the port vectors to
prevent broadcast frames leaking between bridges, and separated ports.

Additionally, when a VLAN is added, a new FID is allocated.  This
represents a new table of ATU entries. A broadcast entry is added to
the new FID.

With offload_fwd_mark being set, the software bridge will not flood
the frames it receives back to the switch.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: mv88e6xxx: Move mv88e6xxx_port_db_load_purge()
Andrew Lunn [Thu, 9 Nov 2017 21:29:55 +0000 (22:29 +0100)]
net: dsa: mv88e6xxx: Move mv88e6xxx_port_db_load_purge()

This function is going to be needed by a soon to be added new
function. Move it earlier so we can avoid a forward declaration.
No functional changes.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: mv88e6xxx: Print offending port when vlan check fails
Andrew Lunn [Thu, 9 Nov 2017 21:29:54 +0000 (22:29 +0100)]
net: dsa: mv88e6xxx: Print offending port when vlan check fails

When testing if a VLAN is one more than one bridge, we print an error
message that the VLAN is already in use somewhere else. Print both the
new port which would like the VLAN, and the port which already has it,
to aid debugging.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: mv88e6xxx: Fixed port netdev check for VLANs
Andrew Lunn [Thu, 9 Nov 2017 21:29:53 +0000 (22:29 +0100)]
net: dsa: mv88e6xxx: Fixed port netdev check for VLANs

Having the same VLAN on multiple bridges is currently unsupported as
an offload. mv88e6xxx_port_check_hw_vlan() is used to ensure that a
VLAN is not on multiple bridges when adding a VLAN range to a port. It
loops the ports and checks to see if there are ports in a different
bridge with the same VLAN.

While walking all switch ports, the code was checking if the new port
has a netdev slave attached to it. If not, skip checking the port
being walked. This seems like a typ0. If the new port does not have a
slave, how has a VLAN been added to it in the first place, requiring
this check be performed at all? More likely, we should be checking if
the port being walked has a slave. Without the port having a slave, it
cannot have a VLAN on it, so there is no need to check further for
that particular port.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: {e}dsa: set offload_fwd_mark on received packets
Andrew Lunn [Thu, 9 Nov 2017 21:29:52 +0000 (22:29 +0100)]
net: dsa: {e}dsa: set offload_fwd_mark on received packets

The software bridge needs to know if a packet has already been bridged
by hardware offload to ports in the same hardware offload, in order
that it does not re-flood them, causing duplicates. This is
particularly true for broadcast and multicast traffic which the host
has requested.

By setting offload_fwd_mark in the skb the bridge will only flood to
ports in other offloads and other netifs. Set this flag in the DSA and
EDSA tag driver.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: Fix SWITCHDEV_ATTR_ID_PORT_PARENT_ID
Andrew Lunn [Thu, 9 Nov 2017 21:29:51 +0000 (22:29 +0100)]
net: dsa: Fix SWITCHDEV_ATTR_ID_PORT_PARENT_ID

SWITCHDEV_ATTR_ID_PORT_PARENT_ID is used by the software bridge when
determining which ports to flood a packet out. If the packet
originated from a switch, it assumes the switch has already flooded
the packet out the switches ports, so the bridge should not flood the
packet itself out switch ports. Ports on the same switch are expected
to return the same parent ID when SWITCHDEV_ATTR_ID_PORT_PARENT_ID is
called.

DSA gets this wrong with clusters of switches. As far as the software
bridge is concerned, the cluster is all one switch. A packet from any
switch in the cluster can be assumed to have been flooded as needed
out of all ports of the cluster, not just the switch it originated
from. Hence all ports of a cluster should return the same parent. The
old implementation did not, each switch in the cluster had its own ID.

Also wrong was that the ID was not unique if multiple DSA instances
are in operation.

Use the tree ID as the parent ID, which is the same for all switches
in a cluster and unique across switch clusters.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'ieee802154-for-davem-2017-11-09' of git://git.kernel.org/pub/scm/linux...
David S. Miller [Sat, 11 Nov 2017 10:30:16 +0000 (19:30 +0900)]
Merge branch 'ieee802154-for-davem-2017-11-09' of git://git./linux/kernel/git/sschmidt/wpan-next

Stefan Schmidt says:

====================
pull-request: net-next: ieee802154 2017-11-09

A small update on ieee802154 patches for net-next. Nothing dramatic, but simply
housekeeping this time around.
A fix for the correct mask to be applied in the mrf24j40 driver by Gustavo A. R. Silva
Removal of a non existing email user for the ca8210 driver by Harry Morris
A bunch of checkpatch cleanups across the subsystem from myself
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agobindings: net: stmmac: correctify note about LPI interrupt
Niklas Cassel [Thu, 9 Nov 2017 17:09:26 +0000 (18:09 +0100)]
bindings: net: stmmac: correctify note about LPI interrupt

There are two different combined signal for various interrupt events:
In EQOS-CORE and EQOS-MTL configurations, mci_intr_o is the interrupt
signal.
In EQOS-DMA, EQOS-AHB and EQOS-AXI configurations, these interrupt events
are combined with the events in the DMA on the sbd_intr_o signal.

Depending on configuration, the device tree irq "macirq" will refer to
either mci_intr_o or sbd_intr_o.

The databook states:
"The MAC generates the LPI interrupt when the Tx or Rx side enters or exits
the LPI state. The interrupt mci_intr_o (sbd_intr_o in certain
configurations) is asserted when the LPI interrupt status is set.

When the MAC exits the Rx LPI state, then in addition to the mci_intr_o
(sbd_intr_o in certain configurations), the sideband signal lpi_intr_o is
asserted.

If you do not want to gate-off the application clock during the Rx LPI
state, you can leave the lpi_intr_o signal unconnected and use the
mci_intr_o (sbd_intr_o in certain configurations) signal to detect Rx LPI
exit."

Since the "macirq" is always raised when Tx or Rx enters/exits the LPI
state, "eth_lpi" must therefore refer to lpi_intr_o, which is only raised
when Rx exits the LPI state. Update the DT binding description to reflect
reality.

Signed-off-by: Niklas Cassel <niklas.cassel@axis.com>
Acked-by: Alexandre TORGUE <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoipvlan: fix ipv6 outbound device
Keefe Liu [Thu, 9 Nov 2017 12:09:31 +0000 (20:09 +0800)]
ipvlan: fix ipv6 outbound device

When process the outbound packet of ipv6, we should assign the master
device to output device other than input device.

Signed-off-by: Keefe Liu <liuqifa@huawei.com>
Acked-by: Mahesh Bandewar <maheshb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: thunderx: fix double free error
Aleksey Makarov [Thu, 9 Nov 2017 11:58:57 +0000 (14:58 +0300)]
net: thunderx: fix double free error

This patch fixes an error in memory allocation/freeing in
ThunderX PF driver.

I moved the allocation to the probe() function and made it managed.

>From the Colin's email:

While running static analysis on linux-next with CoverityScan I found 3
double free errors in the Cavium thunder driver.

The issue occurs on the err_disable_device: label of function nic_probe
when nic_free_lmacmem(nic) is called and a double free occurs on
nic->duplex, nic->link and nic->speed.  This occurs when nic_init_hw()
fails:

        /* Initialize hardware */
        err = nic_init_hw(nic);
        if (err)
                goto err_release_regions;

nic_init_hw() calls nic_get_hw_info() and this calls nic_free_lmacmem()
if any of the allocations fail. This free'ing occurs again by the call
to nic_free_lmacmem() on the err_release_regions exit path in nic_probe().

Reported-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Aleksey Makarov <aleksey.makarov@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: thunderbolt: Clear finished Tx frame bus address in tbnet_tx_callback()
Mika Westerberg [Thu, 9 Nov 2017 10:46:28 +0000 (13:46 +0300)]
net: thunderbolt: Clear finished Tx frame bus address in tbnet_tx_callback()

When Thunderbolt network interface is disabled or when the cable is
unplugged the driver releases all allocated buffers by calling
tbnet_free_buffers() for each ring. This function then calls
dma_unmap_page() for each buffer it finds where bus address is non-zero.
Now, we only clear this bus address when the Tx buffer is sent to the
hardware so it is possible that the function finds an entry that has
already been unmapped.

Enabling DMA-API debugging catches this as well:

  thunderbolt 0000:06:00.0: DMA-API: device driver tries to free DMA
    memory it has not allocated [device address=0x0000000068321000] [size=4096 bytes]

Fix this by clearing the bus address of a Tx frame right after we have
unmapped the buffer.

Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agosock: Remove the global prot_inuse counter.
Tonghao Zhang [Thu, 9 Nov 2017 08:03:15 +0000 (00:03 -0800)]
sock: Remove the global prot_inuse counter.

The per-cpu counter for init_net is prepared in core_initcall.
The patch 7d720c3e ("percpu: add __percpu sparse annotations to net")
and d6d9ca0fe ("net: this_cpu_xxx conversions") optimize the
routines. Then remove the old counter.

Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Tonghao Zhang <zhangtonghao@didichuxing.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: sfc: remove redundant variable start
Colin Ian King [Thu, 9 Nov 2017 08:01:22 +0000 (08:01 +0000)]
net: sfc: remove redundant variable start

Variable start is assigned but never read hence it is redundant
and can be removed. Cleans up clang warning:

drivers/net/ethernet/sfc/ptp.c:655:2: warning: Value stored to 'start'
is never read

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Bert Kenward <bkenward@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoqlge: remove duplicated assignment to mbcp
Colin Ian King [Thu, 9 Nov 2017 07:52:15 +0000 (07:52 +0000)]
qlge: remove duplicated assignment to mbcp

The assignment to mbcp is identical to the initiatialized value assigned
to mbcp at declaration time a few lines earlier, hence we can remove the
second redundant assignment.  Cleans up clang warning:

drivers/net/ethernet/qlogic/qlge/qlge_mpi.c:209:22: warning:
Value stored to 'mbcp' during its initialization is never read

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: wan: x25_asy: mark expected switch fall-through
Gustavo A. R. Silva [Thu, 9 Nov 2017 04:25:08 +0000 (22:25 -0600)]
net: wan: x25_asy: mark expected switch fall-through

In preparation to enabling -Wimplicit-fallthrough, mark switch cases
where we are expecting to fall through.

Addresses-Coverity-ID: 114928
Signed-off-by: Gustavo A. R. Silva <garsilva@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: 3com: 3c574_cs: mark expected switch fall-through
Gustavo A. R. Silva [Thu, 9 Nov 2017 03:49:33 +0000 (21:49 -0600)]
net: 3com: 3c574_cs: mark expected switch fall-through

In preparation to enabling -Wimplicit-fallthrough, mark switch cases
where we are expecting to fall through.

Addresses-Coverity-ID: 114888
Signed-off-by: Gustavo A. R. Silva <garsilva@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: 8390: pcnet_cs: mark expected switch fall-through
Gustavo A. R. Silva [Thu, 9 Nov 2017 03:44:38 +0000 (21:44 -0600)]
net: 8390: pcnet_cs: mark expected switch fall-through

In preparation to enabling -Wimplicit-fallthrough, mark switch cases
where we are expecting to fall through.

Addresses-Coverity-ID: 114891
Signed-off-by: Gustavo A. R. Silva <garsilva@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: decnet: dn_table: mark expected switch fall-through
Gustavo A. R. Silva [Thu, 9 Nov 2017 03:38:28 +0000 (21:38 -0600)]
net: decnet: dn_table: mark expected switch fall-through

In preparation to enabling -Wimplicit-fallthrough, mark switch cases
where we are expecting to fall through.

Addresses-Coverity-ID: 115106
Signed-off-by: Gustavo A. R. Silva <garsilva@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agol2tp: don't close sessions in l2tp_tunnel_destruct()
Guillaume Nault [Wed, 8 Nov 2017 23:29:52 +0000 (08:29 +0900)]
l2tp: don't close sessions in l2tp_tunnel_destruct()

Sessions are already removed by the proto ->destroy() handlers, and
since commit f3c66d4e144a ("l2tp: prevent creation of sessions on terminated tunnels"),
we're guaranteed that no new session can be created afterwards.

Furthermore, l2tp_tunnel_closeall() can sleep when there are sessions
left to close. So we really shouldn't call it in a ->sk_destruct()
handler, as it can be used from atomic context.

Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'FACK-loss-recovery-remove'
David S. Miller [Sat, 11 Nov 2017 09:53:17 +0000 (18:53 +0900)]
Merge branch 'FACK-loss-recovery-remove'

Yuchung Cheng says:

====================
remove FACK loss recovery

This patch set removes the forward-acknowledgment (FACK)
packet-based loss and reordering detection. This simplifies TCP
loss recovery since the SACK scoreboard no longer needs to track
the number of pending packets under highest SACKed sequence. FACK
is subsumed by the time-based RACK loss detection which is more
robust under reordering and second order losses.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotcp: use sequence distance to detect reordering
Yuchung Cheng [Wed, 8 Nov 2017 21:01:27 +0000 (13:01 -0800)]
tcp: use sequence distance to detect reordering

Replace the reordering distance measurement in packet unit with
sequence based approach. Previously it trackes the number of "packets"
toward the forward ACK (i.e.  highest sacked sequence)in a state
variable "fackets_out".

Precisely measuring reordering degree on packet distance has not much
benefit, as the degree constantly changes by factors like path, load,
and congestion window. It is also complicated and prone to arcane bugs.
This patch replaces with sequence-based approach that's much simpler.

Signed-off-by: Yuchung Cheng <ycheng@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Neal Cardwell <ncardwell@google.com>
Reviewed-by: Soheil Hassas Yeganeh <soheil@google.com>
Reviewed-by: Priyaranjan Jha <priyarjha@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotcp: retire FACK loss detection
Yuchung Cheng [Wed, 8 Nov 2017 21:01:26 +0000 (13:01 -0800)]
tcp: retire FACK loss detection

FACK loss detection has been disabled by default and the
successor RACK subsumed FACK and can handle reordering better.
This patch removes FACK to simplify TCP loss recovery.

Signed-off-by: Yuchung Cheng <ycheng@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Neal Cardwell <ncardwell@google.com>
Reviewed-by: Soheil Hassas Yeganeh <soheil@google.com>
Reviewed-by: Priyaranjan Jha <priyarjha@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agofsl/fman_port: mark expected switch fall-throughs
Gustavo A. R. Silva [Wed, 8 Nov 2017 17:57:13 +0000 (11:57 -0600)]
fsl/fman_port: mark expected switch fall-throughs

In preparation to enabling -Wimplicit-fallthrough, mark switch cases
where we are expecting to fall through.

Addresses-Coverity-ID: 1397960
Signed-off-by: Gustavo A. R. Silva <garsilva@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: ethernet: bgmac: mark expected switch fall-through
Gustavo A. R. Silva [Wed, 8 Nov 2017 17:24:57 +0000 (11:24 -0600)]
net: ethernet: bgmac: mark expected switch fall-through

In preparation to enabling -Wimplicit-fallthrough, mark switch cases
where we are expecting to fall through.

Addresses-Coverity-ID: 1397972
Signed-off-by: Gustavo A. R. Silva <garsilva@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoibmvnic: Add vnic client data to login buffer
Nathan Fontenot [Wed, 8 Nov 2017 17:23:56 +0000 (11:23 -0600)]
ibmvnic: Add vnic client data to login buffer

Update the login buffer to include client data for the vnic driver,
this includes the OS name, LPAR name, and device name. This update
allows this information to be available in the VIOS.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agobpf: Revert bpf_overrid_function() helper changes.
David S. Miller [Sat, 11 Nov 2017 09:24:55 +0000 (18:24 +0900)]
bpf: Revert bpf_overrid_function() helper changes.

NACK'd by x86 maintainer.

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'bpf-Fix-bugs-in-sock_ops-samples'
David S. Miller [Sat, 11 Nov 2017 06:52:41 +0000 (15:52 +0900)]
Merge branch 'bpf-Fix-bugs-in-sock_ops-samples'

Lawrence Brakmo says:

====================
bpf: Fix bugs in sock_ops samples

The programs were returning -1 in some cases when they should
only return 0 or 1. Changes in the verifier now catch this
issue and the programs fail to load. This is now fixed.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agobpf: Fix tcp_clamp_kern.c sample program
Lawrence Brakmo [Sat, 11 Nov 2017 06:19:55 +0000 (22:19 -0800)]
bpf: Fix tcp_clamp_kern.c sample program

The program was returning -1 in some cases which is not allowed
by the verifier any longer.

Fixes: 390ee7e29fc8 ("bpf: enforce return code for cgroup-bpf programs")
Signed-off-by: Lawrence Brakmo <brakmo@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agobpf: Fix tcp_iw_kern.c sample program
Lawrence Brakmo [Sat, 11 Nov 2017 06:19:54 +0000 (22:19 -0800)]
bpf: Fix tcp_iw_kern.c sample program

The program was returning -1 in some cases which is not allowed
by the verifier any longer.

Fixes: 390ee7e29fc8 ("bpf: enforce return code for cgroup-bpf programs")
Signed-off-by: Lawrence Brakmo <brakmo@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agobpf: Fix tcp_cong_kern.c sample program
Lawrence Brakmo [Sat, 11 Nov 2017 06:19:53 +0000 (22:19 -0800)]
bpf: Fix tcp_cong_kern.c sample program

The program was returning -1 in some cases which is not allowed
by the verifier any longer.

Fixes: 390ee7e29fc8 ("bpf: enforce return code for cgroup-bpf programs")
Signed-off-by: Lawrence Brakmo <brakmo@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agobpf: Fix tcp_bufs_kern.c sample program
Lawrence Brakmo [Sat, 11 Nov 2017 06:19:52 +0000 (22:19 -0800)]
bpf: Fix tcp_bufs_kern.c sample program

The program was returning -1 in some cases which is not allowed
by the verifier any longer.

Fixes: 390ee7e29fc8 ("bpf: enforce return code for cgroup-bpf programs")
Signed-off-by: Lawrence Brakmo <brakmo@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agobpf: Fix tcp_rwnd_kern.c sample program
Lawrence Brakmo [Sat, 11 Nov 2017 06:19:51 +0000 (22:19 -0800)]
bpf: Fix tcp_rwnd_kern.c sample program

The program was returning -1 in some cases which is not allowed
by the verifier any longer.

Fixes: 390ee7e29fc8 ("bpf: enforce return code for cgroup-bpf programs")
Signed-off-by: Lawrence Brakmo <brakmo@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agobpf: Fix tcp_synrto_kern.c sample program
Lawrence Brakmo [Sat, 11 Nov 2017 06:19:50 +0000 (22:19 -0800)]
bpf: Fix tcp_synrto_kern.c sample program

The program was returning -1 in some cases which is not allowed
by the verifier any longer.

Fixes: 390ee7e29fc8 ("bpf: enforce return code for cgroup-bpf programs")
Signed-off-by: Lawrence Brakmo <brakmo@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: return after vlan prepare phase
Vivien Didelot [Wed, 8 Nov 2017 15:50:10 +0000 (10:50 -0500)]
net: dsa: return after vlan prepare phase

The current code does not return after successfully preparing the VLAN
addition on every ports member of a it. Fix this.

Fixes: 1ca4aa9cd4cc ("net: dsa: check VLAN capability of every switch")
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: return after mdb prepare phase
Vivien Didelot [Wed, 8 Nov 2017 15:49:56 +0000 (10:49 -0500)]
net: dsa: return after mdb prepare phase

The current code does not return after successfully preparing the MDB
addition on every ports member of a multicast group. Fix this.

Fixes: a1a6b7ea7f2d ("net: dsa: add cross-chip multicast support")
Reported-by: Egil Hjelmeland <privat@egil-hjelmeland.no>
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotipc: improve link resiliency when rps is activated
Jon Maloy [Wed, 8 Nov 2017 08:59:26 +0000 (09:59 +0100)]
tipc: improve link resiliency when rps is activated

Currently, the TIPC RPS dissector is based only on the incoming packets'
source node address, hence steering all traffic from a node to the same
core. We have seen that this makes the links vulnerable to starvation
and unnecessary resets when we turn down the link tolerance to very low
values.

To reduce the risk of this happening, we exempt probe and probe replies
packets from the convergence to one core per source node. Instead, we do
the opposite, - we try to diverge those packets across as many cores as
possible, by randomizing the flow selector key.

To make such packets identifiable to the dissector, we add a new
'is_keepalive' bit to word 0 of the LINK_PROTOCOL header. This bit is
set both for PROBE and PROBE_REPLY messages, and only for those.

It should be noted that these packets are not part of any flow anyway,
and only constitute a minuscule fraction of all packets sent across a
link. Hence, there is no risk that this will affect overall performance.

Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'macb-next'
David S. Miller [Sat, 11 Nov 2017 06:27:45 +0000 (15:27 +0900)]
Merge branch 'macb-next'

Michael Grzeschik says:

====================
net: macb: add error handling on probe and

This series adds more error handling to the macb driver.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: macb: add of_node_put to error paths
Michael Grzeschik [Wed, 8 Nov 2017 08:56:35 +0000 (09:56 +0100)]
net: macb: add of_node_put to error paths

We add the call of_node_put(bp->phy_node) to all associated error
paths for memory clean up.

Signed-off-by: Michael Grzeschik <m.grzeschik@pengutronix.de>
Acked-by: Nicolas Ferre <nicolas.ferre@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: macb: add of_phy_deregister_fixed_link to error paths
Michael Grzeschik [Wed, 8 Nov 2017 08:56:34 +0000 (09:56 +0100)]
net: macb: add of_phy_deregister_fixed_link to error paths

We add the call of_phy_deregister_fixed_link to all associated
error paths for memory clean up.

Signed-off-by: Michael Grzeschik <m.grzeschik@pengutronix.de>
Acked-by: Nicolas Ferre <nicolas.ferre@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: mvpp2: fix GOP statistics loop start and stop conditions
Miquel Raynal [Wed, 8 Nov 2017 07:59:40 +0000 (08:59 +0100)]
net: mvpp2: fix GOP statistics loop start and stop conditions

GOP statistics from all ports of one instance of the driver are gathered
with one work recalled in loop in a workqueue. The loop is started when
a port is up, and stopped when a port is down. This last condition is
obviously wrong.

Fix this by having a work per port. This way, starting and stoping it
when the port is up or down will be fine, while minimizing unnecessary
CPU usage.

Fixes: 118d6298f6f0 ("net: mvpp2: add ethtool GOP statistics")
Reported-by: Stefan Chulski <stefanc@marvell.com>
Signed-off-by: Miquel Raynal <miquel.raynal@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'hns3-bug-fixes'
David S. Miller [Sat, 11 Nov 2017 06:17:56 +0000 (15:17 +0900)]
Merge branch 'hns3-bug-fixes'

Lipeng says:

====================
net: hns3: Bug fixes & Code improvements in HNS3 driver

This patch-set introduces some bug fixes and code improvements.
As [patch 1/2] depends on the patch {5392902 net: hns3: Consistently using
GENMASK in hns3 driver}, which exists in net-next, not exists in net, so
push this serise to nex-next.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: hns3: cleanup mac auto-negotiation state query in hclge_update_speed_duplex
Fuyun Liang [Wed, 8 Nov 2017 07:52:23 +0000 (15:52 +0800)]
net: hns3: cleanup mac auto-negotiation state query in hclge_update_speed_duplex

When checking whether auto-negotiation is on, driver only needs to
check the value of mac.autoneg(SW) directly, and does not need to
query it from hardware. Because this value is always synchronized
with the auto-negotiation state of hardware.

This patch removes mac auto-negotiation state query in
hclge_update_speed_duplex().

Fixes: 46a3df9f9718 (net: hns3: Add HNS3 Acceleration Engine & Compatibility Layer Support)
Signed-off-by: Fuyun Liang <liangfuyun1@huawei.com>
Signed-off-by: Lipeng <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: hns3: fix a bug when getting phy address from NCL_config file
Fuyun Liang [Wed, 8 Nov 2017 07:52:22 +0000 (15:52 +0800)]
net: hns3: fix a bug when getting phy address from NCL_config file

Driver gets phy address from NCL_config file and uses the phy address
to initialize phydev. There are 5 bits for phy address. And C22 phy
address has 5 bits. So 0-31 are all valid address for phy. If there
is no phy, it will crash. Because driver always get a valid phy address.

This patch fixes the phy address to 8 bits, and use 0xff to indicate
invalid phy address.

Fixes: 46a3df9f9718 (net: hns3: Add HNS3 Acceleration Engine & Compatibility Layer Support)
Signed-off-by: Fuyun Liang <liangfuyun1@huawei.com>
Signed-off-by: Lipeng <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: netlink: Update attr validation to require exact length for some types
David Ahern [Wed, 8 Nov 2017 05:59:40 +0000 (21:59 -0800)]
net: netlink: Update attr validation to require exact length for some types

Attributes using NLA_U* and NLA_S* (where * is 8, 16,32 and 64) are
expected to be an exact length. Split these data types from
nla_attr_minlen into nla_attr_len and update validate_nla to require
the attribute to have exact length for them.

Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: ipv6: sysctl to specify IPv6 ND traffic class
Maciej Żenczykowski [Wed, 8 Nov 2017 05:52:09 +0000 (21:52 -0800)]
net: ipv6: sysctl to specify IPv6 ND traffic class

Add a per-device sysctl to specify the default traffic class to use for
kernel originated IPv6 Neighbour Discovery packets.

Currently this includes:

  - Router Solicitation (ICMPv6 type 133)
    ndisc_send_rs() -> ndisc_send_skb() -> ip6_nd_hdr()

  - Neighbour Solicitation (ICMPv6 type 135)
    ndisc_send_ns() -> ndisc_send_skb() -> ip6_nd_hdr()

  - Neighbour Advertisement (ICMPv6 type 136)
    ndisc_send_na() -> ndisc_send_skb() -> ip6_nd_hdr()

  - Redirect (ICMPv6 type 137)
    ndisc_send_redirect() -> ndisc_send_skb() -> ip6_nd_hdr()

and if the kernel ever gets around to generating RA's,
it would presumably also include:

  - Router Advertisement (ICMPv6 type 134)
    (radvd daemon could pick up on the kernel setting and use it)

Interface drivers may examine the Traffic Class value and translate
the DiffServ Code Point into a link-layer appropriate traffic
prioritization scheme.  An example of mapping IETF DSCP values to
IEEE 802.11 User Priority values can be found here:

    https://tools.ietf.org/html/draft-ietf-tsvwg-ieee-802-11

The expected primary use case is to properly prioritize ND over wifi.

Testing:
  jzem22:~# cat /proc/sys/net/ipv6/conf/eth0/ndisc_tclass
  0
  jzem22:~# echo -1 > /proc/sys/net/ipv6/conf/eth0/ndisc_tclass
  -bash: echo: write error: Invalid argument
  jzem22:~# echo 256 > /proc/sys/net/ipv6/conf/eth0/ndisc_tclass
  -bash: echo: write error: Invalid argument
  jzem22:~# echo 0 > /proc/sys/net/ipv6/conf/eth0/ndisc_tclass
  jzem22:~# echo 255 > /proc/sys/net/ipv6/conf/eth0/ndisc_tclass
  jzem22:~# cat /proc/sys/net/ipv6/conf/eth0/ndisc_tclass
  255
  jzem22:~# echo 34 > /proc/sys/net/ipv6/conf/eth0/ndisc_tclass
  jzem22:~# cat /proc/sys/net/ipv6/conf/eth0/ndisc_tclass
  34

  jzem22:~# echo $[0xDC] > /proc/sys/net/ipv6/conf/eth0/ndisc_tclass
  jzem22:~# tcpdump -v -i eth0 icmp6 and src host jzem22.pgc and dst host fe80::1
  tcpdump: listening on eth0, link-type EN10MB (Ethernet), capture size 262144 bytes
  IP6 (class 0xdc, hlim 255, next-header ICMPv6 (58) payload length: 24)
  jzem22.pgc > fe80::1: [icmp6 sum ok] ICMP6, neighbor advertisement,
  length 24, tgt is jzem22.pgc, Flags [solicited]

(based on original change written by Erik Kline, with minor changes)

v2: fix 'suspicious rcu_dereference_check() usage'
    by explicitly grabbing the rcu_read_lock.

Cc: Lorenzo Colitti <lorenzo@google.com>
Signed-off-by: Erik Kline <ek@google.com>
Signed-off-by: Maciej Żenczykowski <maze@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/ncsi: Don't return error on normal response
Samuel Mendoza-Jonas [Wed, 8 Nov 2017 05:30:45 +0000 (16:30 +1100)]
net/ncsi: Don't return error on normal response

Several response handlers return EBUSY if the data corresponding to the
command/response pair is already set. There is no reason to return an
error here; the channel is advertising something as enabled because we
told it to enable it, and it's possible that the feature has been
enabled previously.

Signed-off-by: Samuel Mendoza-Jonas <sam@mendozajonas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/ncsi: Improve general state logging
Samuel Mendoza-Jonas [Wed, 8 Nov 2017 05:30:44 +0000 (16:30 +1100)]
net/ncsi: Improve general state logging

The NCSI driver is mostly silent which becomes a headache when trying to
determine what has occurred on the NCSI connection. This adds additional
logging in a few key areas such as state transitions and calling out
certain errors more visibly.

Signed-off-by: Samuel Mendoza-Jonas <sam@mendozajonas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'bpftool-show-filenames-of-pinned-objects'
David S. Miller [Sat, 11 Nov 2017 03:35:49 +0000 (12:35 +0900)]
Merge branch 'bpftool-show-filenames-of-pinned-objects'

Prashant Bhole says:

====================
tools: bpftool: show filenames of pinned objects

This patchset adds support to show pinned objects in object details.

Patch1 adds a funtionality to open a path in bpf-fs regardless of its object
type.

Patch2 adds actual functionality by scanning the bpf-fs once and adding
object information in hash table, with object id as a key. One object may be
associated with multiple paths because an object can be pinned multiple times

Patch3 adds command line option to enable this functionality. Making it optional
because scanning bpf-fs can be costly.
====================

Acked-by: Jakub Kicinski <jakub.kicinski@netronome.com>
6 years agotools: bpftool: optionally show filenames of pinned objects
Prashant Bhole [Wed, 8 Nov 2017 04:55:49 +0000 (13:55 +0900)]
tools: bpftool: optionally show filenames of pinned objects

Making it optional to show file names of pinned objects because
it scans complete bpf-fs filesystem which is costly.
Added option -f|--bpffs. Documentation updated.

Signed-off-by: Prashant Bhole <bhole_prashant_q7@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotools: bpftool: show filenames of pinned objects
Prashant Bhole [Wed, 8 Nov 2017 04:55:48 +0000 (13:55 +0900)]
tools: bpftool: show filenames of pinned objects

Added support to show filenames of pinned objects.

For example:

root@test# ./bpftool prog
3: tracepoint  name tracepoint__irq  tag f677a7dd722299a3
    loaded_at Oct 26/11:39  uid 0
    xlated 160B  not jited  memlock 4096B  map_ids 4
    pinned /sys/fs/bpf/softirq_prog

4: tracepoint  name tracepoint__irq  tag ea5dc530d00b92b6
    loaded_at Oct 26/11:39  uid 0
    xlated 392B  not jited  memlock 4096B  map_ids 4,6

root@test# ./bpftool --json --pretty prog
[{
        "id": 3,
        "type": "tracepoint",
        "name": "tracepoint__irq",
        "tag": "f677a7dd722299a3",
        "loaded_at": "Oct 26/11:39",
        "uid": 0,
        "bytes_xlated": 160,
        "jited": false,
        "bytes_memlock": 4096,
        "map_ids": [4
        ],
        "pinned": ["/sys/fs/bpf/softirq_prog"
        ]
    },{
        "id": 4,
        "type": "tracepoint",
        "name": "tracepoint__irq",
        "tag": "ea5dc530d00b92b6",
        "loaded_at": "Oct 26/11:39",
        "uid": 0,
        "bytes_xlated": 392,
        "jited": false,
        "bytes_memlock": 4096,
        "map_ids": [4,6
        ],
        "pinned": []
    }
]

root@test# ./bpftool map
4: hash  name start  flags 0x0
    key 4B  value 16B  max_entries 10240  memlock 1003520B
    pinned /sys/fs/bpf/softirq_map1
5: hash  name iptr  flags 0x0
    key 4B  value 8B  max_entries 10240  memlock 921600B

root@test# ./bpftool --json --pretty map
[{
        "id": 4,
        "type": "hash",
        "name": "start",
        "flags": 0,
        "bytes_key": 4,
        "bytes_value": 16,
        "max_entries": 10240,
        "bytes_memlock": 1003520,
        "pinned": ["/sys/fs/bpf/softirq_map1"
        ]
    },{
        "id": 5,
        "type": "hash",
        "name": "iptr",
        "flags": 0,
        "bytes_key": 4,
        "bytes_value": 8,
        "max_entries": 10240,
        "bytes_memlock": 921600,
        "pinned": []
    }
]

Signed-off-by: Prashant Bhole <bhole_prashant_q7@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotools: bpftool: open pinned object without type check
Prashant Bhole [Wed, 8 Nov 2017 04:55:47 +0000 (13:55 +0900)]
tools: bpftool: open pinned object without type check

This was needed for opening any file in bpf-fs without knowing
its object type

Signed-off-by: Prashant Bhole <bhole_prashant_q7@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'BPF-directed-error-injection'
David S. Miller [Sat, 11 Nov 2017 03:18:06 +0000 (12:18 +0900)]
Merge branch 'BPF-directed-error-injection'

Josef Bacik says:

====================
Add the ability to do BPF directed error injection

I'm sending this through Dave since it'll conflict with other BPF changes in his
tree, but since it touches tracing as well Dave would like a review from
somebody on the tracing side.

v4->v5:
- disallow kprobe_override programs from being put in the prog map array so we
  don't tail call into something we didn't check.  This allows us to make the
  normal path still fast without a bunch of percpu operations.

v3->v4:
- fix a build error found by kbuild test bot (I didn't wait long enough
  apparently.)
- Added a warning message as per Daniels suggestion.

v2->v3:
- added a ->kprobe_override flag to bpf_prog.
- added some sanity checks to disallow attaching bpf progs that have
  ->kprobe_override set that aren't for ftrace kprobes.
- added the trace_kprobe_ftrace helper to check if the trace_event_call is a
  ftrace kprobe.
- renamed bpf_kprobe_state to bpf_kprobe_override, fixed it so we only read this
  value in the kprobe path, and thus only write to it if we're overriding or
  clearing the override.

v1->v2:
- moved things around to make sure that bpf_override_return could really only be
  used for an ftrace kprobe.
- killed the special return values from trace_call_bpf.
- renamed pc_modified to bpf_kprobe_state so bpf_override_return could tell if
  it was being called from an ftrace kprobe context.
- reworked the logic in kprobe_perf_func to take advantage of bpf_kprobe_state.
- updated the test as per Alexei's review.

- Original message -

A lot of our error paths are not well tested because we have no good way of
injecting errors generically.  Some subystems (block, memory) have ways to
inject errors, but they are random so it's hard to get reproduceable results.

With BPF we can add determinism to our error injection.  We can use kprobes and
other things to verify we are injecting errors at the exact case we are trying
to test.  This patch gives us the tool to actual do the error injection part.
It is very simple, we just set the return value of the pt_regs we're given to
whatever we provide, and then override the PC with a dummy function that simply
returns.

Right now this only works on x86, but it would be simple enough to expand to
other architectures.  Thanks,
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agosamples/bpf: add a test for bpf_override_return
Josef Bacik [Tue, 7 Nov 2017 20:28:43 +0000 (15:28 -0500)]
samples/bpf: add a test for bpf_override_return

This adds a basic test for bpf_override_return to verify it works.  We
override the main function for mounting a btrfs fs so it'll return
-ENOMEM and then make sure that trying to mount a btrfs fs will fail.

Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Josef Bacik <jbacik@fb.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agobpf: add a bpf_override_function helper
Josef Bacik [Tue, 7 Nov 2017 20:28:42 +0000 (15:28 -0500)]
bpf: add a bpf_override_function helper

Error injection is sloppy and very ad-hoc.  BPF could fill this niche
perfectly with it's kprobe functionality.  We could make sure errors are
only triggered in specific call chains that we care about with very
specific situations.  Accomplish this with the bpf_override_funciton
helper.  This will modify the probe'd callers return value to the
specified value and set the PC to an override function that simply
returns, bypassing the originally probed function.  This gives us a nice
clean way to implement systematic error injection for all of our code
paths.

Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Josef Bacik <jbacik@fb.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agobrcmfmac: add CLM download support
Chung-Hsien Hsu [Fri, 10 Nov 2017 09:27:15 +0000 (17:27 +0800)]
brcmfmac: add CLM download support

The firmware for brcmfmac devices includes information regarding
regulatory constraints. For certain devices this information is kept
separately in a binary form that needs to be downloaded to the device.
This patch adds support to download this so-called CLM blob file. It
uses the same naming scheme as the other firmware files with extension
of .clm_blob.

The CLM blob file is optional. If the file does not exist, the download
process will be bypassed. It will not affect the driver loading.

Reviewed-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Signed-off-by: Chung-Hsien Hsu <stanley.hsu@cypress.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
6 years agoMerge tag 'ceph-for-4.14-rc9' of git://github.com/ceph/ceph-client
Linus Torvalds [Fri, 10 Nov 2017 22:18:24 +0000 (14:18 -0800)]
Merge tag 'ceph-for-4.14-rc9' of git://github.com/ceph/ceph-client

Pull ceph gix from Ilya Dryomov:
 "Memory allocation flags fix, marked for stable"

* tag 'ceph-for-4.14-rc9' of git://github.com/ceph/ceph-client:
  rbd: use GFP_NOIO for parent stat and data requests

6 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Fri, 10 Nov 2017 22:14:23 +0000 (14:14 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

Pull input layer updates from Dmitry Torokhov:

 - a new ACPI ID for Elan touchpad found in yet another Ideapad model

 - Synaptics RMI4 will allow binding to controllers reporting SMB
   version 3 (note that we are not adding any new ACPI IDs to the
   Synaptics PS/2 drover so unless user explicitly enables intertouch
   support there is no user-visible change)

 - a fixup to TSC 2004/5 touchscreen driver to mark input devices as
   "direct" to help userspace identify the type of device they are
   dealing with

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: synaptics-rmi4 - RMI4 can also use SMBUS version 3
  Input: tsc200x-core - set INPUT_PROP_DIRECT
  Input: elan_i2c - add ELAN060C to the ACPI table

6 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Fri, 10 Nov 2017 20:24:42 +0000 (12:24 -0800)]
Merge tag 'for-linus' of git://git./virt/kvm/kvm

Pull KVM fix from Radim Krčmář:
 "Fix PPC HV host crash that can occur as a result of resizing the guest
  hashed page table"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: PPC: Book3S HV: Fix exclusion between HPT resizing and other HPT updates

6 years agoMerge tag 'mips_fixes_4.14_2' of git://git.kernel.org/pub/scm/linux/kernel/git/jhogan...
Linus Torvalds [Fri, 10 Nov 2017 20:21:15 +0000 (12:21 -0800)]
Merge tag 'mips_fixes_4.14_2' of git://git./linux/kernel/git/jhogan/mips

Pull MIPS fixes from James Hogan:
 "A final few MIPS fixes for 4.14:

   - fix BMIPS NULL pointer dereference (4.7)

   - fix AR7 early GPIO init allocation failure (3.19)

   - fix dead serial output on certain AR7 platforms (2.6.35)"

* tag 'mips_fixes_4.14_2' of git://git.kernel.org/pub/scm/linux/kernel/git/jhogan/mips:
  MIPS: AR7: Ensure that serial ports are properly set up
  MIPS: AR7: Defer registration of GPIO
  MIPS: BMIPS: Fix missing cbr address

6 years ago.mailmap: Add Maciej W. Rozycki's Imagination e-mail address
Maciej W. Rozycki [Fri, 10 Nov 2017 20:05:24 +0000 (20:05 +0000)]
.mailmap: Add Maciej W. Rozycki's Imagination e-mail address

Following my recent transition from Imagination Technologies to the=20
reincarnated MIPS company add a .mailmap mapping for my work address,
so that `scripts/get_maintainer.pl' gets it right for past commits.

Signed-off-by: Maciej W. Rozycki <macro@mips.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agoRevert "x86: CPU: Fix up "cpu MHz" in /proc/cpuinfo"
Linus Torvalds [Fri, 10 Nov 2017 19:19:11 +0000 (11:19 -0800)]
Revert "x86: CPU: Fix up "cpu MHz" in /proc/cpuinfo"

This reverts commit 941f5f0f6ef5338814145cf2b813cf1f98873e2f.

Sadly, it turns out that we really can't just do the cross-CPU IPI to
all CPU's to get their proper frequencies, because it's much too
expensive on systems with lots of cores.

So we'll have to revert this for now, and revisit it using a smarter
model (probably doing one system-wide IPI at open time, and doing all
the frequency calculations in parallel).

Reported-by: WANG Chao <chao.wang@ucloud.cn>
Reported-by: Ingo Molnar <mingo@kernel.org>
Cc: Rafael J Wysocki <rafael.j.wysocki@intel.com>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agoMerge tag 'drm-fixes-for-v4.14-rc9' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Fri, 10 Nov 2017 17:59:41 +0000 (09:59 -0800)]
Merge tag 'drm-fixes-for-v4.14-rc9' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
 "Last few patches to wrap up.

  Two i915 fixes that are on their way to stable, one vmware black
  screen bug, and one const patch that I was going to drop, but it was
  clearly a pretty safe one liner"

* tag 'drm-fixes-for-v4.14-rc9' of git://people.freedesktop.org/~airlied/linux:
  drm/i915: Deconstruct struct sgt_dma initialiser
  drm/i915: Reject unknown syncobj flags
  drm/vmwgfx: Fix Ubuntu 17.10 Wayland black screen issue
  drm/vmwgfx: constify vmw_fence_ops

6 years agocan: ifi: Fix transmitter delay calculation
Marek Vasut [Fri, 10 Nov 2017 10:22:39 +0000 (11:22 +0100)]
can: ifi: Fix transmitter delay calculation

The CANFD transmitter delay calculation formula was updated in the
latest software drop from IFI and improves the behavior of the IFI
CANFD core during bitrate switching. Use the new formula to improve
stability of the CANFD operation.

Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Markus Marb <markus@marb.org>
Cc: linux-stable <stable@vger.kernel.org>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
6 years agotcp: fix tcp_fastretrans_alert warning
Yuchung Cheng [Tue, 7 Nov 2017 23:33:43 +0000 (15:33 -0800)]
tcp: fix tcp_fastretrans_alert warning

This patch fixes the cause of an WARNING indicatng TCP has pending
retransmission in Open state in tcp_fastretrans_alert().

The root cause is a bad interaction between path mtu probing,
if enabled, and the RACK loss detection. Upong receiving a SACK
above the sequence of the MTU probing packet, RACK could mark the
probe packet lost in tcp_fastretrans_alert(), prior to calling
tcp_simple_retransmit().

tcp_simple_retransmit() only enters Loss state if it newly marks
the probe packet lost. If the probe packet is already identified as
lost by RACK, the sender remains in Open state with some packets
marked lost and retransmitted. Then the next SACK would trigger
the warning. The likely scenario is that the probe packet was
lost due to its size or network congestion. The actual impact of
this warning is small by potentially entering fast recovery an
ACK later.

The simple fix is always entering recovery (Loss) state if some
packet is marked lost during path MTU probing.

Fixes: a0370b3f3f2c ("tcp: enable RACK loss detection to trigger recovery")
Reported-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Reported-by: Alexei Starovoitov <alexei.starovoitov@gmail.com>
Reported-by: Roman Gushchin <guro@fb.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotcp: gso: avoid refcount_t warning from tcp_gso_segment()
Eric Dumazet [Tue, 7 Nov 2017 23:15:04 +0000 (15:15 -0800)]
tcp: gso: avoid refcount_t warning from tcp_gso_segment()

When a GSO skb of truesize O is segmented into 2 new skbs of truesize N1
and N2, we want to transfer socket ownership to the new fresh skbs.

In order to avoid expensive atomic operations on a cache line subject to
cache bouncing, we replace the sequence :

refcount_add(N1, &sk->sk_wmem_alloc);
refcount_add(N2, &sk->sk_wmem_alloc); // repeated by number of segments

refcount_sub(O, &sk->sk_wmem_alloc);

by a single

refcount_add(sum_of(N) - O, &sk->sk_wmem_alloc);

Problem is :

In some pathological cases, sum(N) - O might be a negative number, and
syzkaller bot was apparently able to trigger this trace [1]

atomic_t was ok with this construct, but we need to take care of the
negative delta with refcount_t

[1]
refcount_t: saturated; leaking memory.
------------[ cut here ]------------
WARNING: CPU: 0 PID: 8404 at lib/refcount.c:77 refcount_add_not_zero+0x198/0x200 lib/refcount.c:77
Kernel panic - not syncing: panic_on_warn set ...

CPU: 0 PID: 8404 Comm: syz-executor2 Not tainted 4.14.0-rc5-mm1+ #20
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Call Trace:
 __dump_stack lib/dump_stack.c:16 [inline]
 dump_stack+0x194/0x257 lib/dump_stack.c:52
 panic+0x1e4/0x41c kernel/panic.c:183
 __warn+0x1c4/0x1e0 kernel/panic.c:546
 report_bug+0x211/0x2d0 lib/bug.c:183
 fixup_bug+0x40/0x90 arch/x86/kernel/traps.c:177
 do_trap_no_signal arch/x86/kernel/traps.c:211 [inline]
 do_trap+0x260/0x390 arch/x86/kernel/traps.c:260
 do_error_trap+0x120/0x390 arch/x86/kernel/traps.c:297
 do_invalid_op+0x1b/0x20 arch/x86/kernel/traps.c:310
 invalid_op+0x18/0x20 arch/x86/entry/entry_64.S:905
RIP: 0010:refcount_add_not_zero+0x198/0x200 lib/refcount.c:77
RSP: 0018:ffff8801c606e3a0 EFLAGS: 00010282
RAX: 0000000000000026 RBX: 0000000000001401 RCX: 0000000000000000
RDX: 0000000000000026 RSI: ffffc900036fc000 RDI: ffffed0038c0dc68
RBP: ffff8801c606e430 R08: 0000000000000001 R09: 0000000000000000
R10: ffff8801d97f5eba R11: 0000000000000000 R12: ffff8801d5acf73c
R13: 1ffff10038c0dc75 R14: 00000000ffffffff R15: 00000000fffff72f
 refcount_add+0x1b/0x60 lib/refcount.c:101
 tcp_gso_segment+0x10d0/0x16b0 net/ipv4/tcp_offload.c:155
 tcp4_gso_segment+0xd4/0x310 net/ipv4/tcp_offload.c:51
 inet_gso_segment+0x60c/0x11c0 net/ipv4/af_inet.c:1271
 skb_mac_gso_segment+0x33f/0x660 net/core/dev.c:2749
 __skb_gso_segment+0x35f/0x7f0 net/core/dev.c:2821
 skb_gso_segment include/linux/netdevice.h:3971 [inline]
 validate_xmit_skb+0x4ba/0xb20 net/core/dev.c:3074
 __dev_queue_xmit+0xe49/0x2070 net/core/dev.c:3497
 dev_queue_xmit+0x17/0x20 net/core/dev.c:3538
 neigh_hh_output include/net/neighbour.h:471 [inline]
 neigh_output include/net/neighbour.h:479 [inline]
 ip_finish_output2+0xece/0x1460 net/ipv4/ip_output.c:229
 ip_finish_output+0x85e/0xd10 net/ipv4/ip_output.c:317
 NF_HOOK_COND include/linux/netfilter.h:238 [inline]
 ip_output+0x1cc/0x860 net/ipv4/ip_output.c:405
 dst_output include/net/dst.h:459 [inline]
 ip_local_out+0x95/0x160 net/ipv4/ip_output.c:124
 ip_queue_xmit+0x8c6/0x18e0 net/ipv4/ip_output.c:504
 tcp_transmit_skb+0x1ab7/0x3840 net/ipv4/tcp_output.c:1137
 tcp_write_xmit+0x663/0x4de0 net/ipv4/tcp_output.c:2341
 __tcp_push_pending_frames+0xa0/0x250 net/ipv4/tcp_output.c:2513
 tcp_push_pending_frames include/net/tcp.h:1722 [inline]
 tcp_data_snd_check net/ipv4/tcp_input.c:5050 [inline]
 tcp_rcv_established+0x8c7/0x18a0 net/ipv4/tcp_input.c:5497
 tcp_v4_do_rcv+0x2ab/0x7d0 net/ipv4/tcp_ipv4.c:1460
 sk_backlog_rcv include/net/sock.h:909 [inline]
 __release_sock+0x124/0x360 net/core/sock.c:2264
 release_sock+0xa4/0x2a0 net/core/sock.c:2776
 tcp_sendmsg+0x3a/0x50 net/ipv4/tcp.c:1462
 inet_sendmsg+0x11f/0x5e0 net/ipv4/af_inet.c:763
 sock_sendmsg_nosec net/socket.c:632 [inline]
 sock_sendmsg+0xca/0x110 net/socket.c:642
 ___sys_sendmsg+0x31c/0x890 net/socket.c:2048
 __sys_sendmmsg+0x1e6/0x5f0 net/socket.c:2138

Fixes: 14afee4b6092 ("net: convert sock.sk_wmem_alloc from atomic_t to refcount_t")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: fix incorrect comment with regard to VLAN packet handling
Girish Moodalbail [Tue, 7 Nov 2017 19:32:11 +0000 (11:32 -0800)]
net: fix incorrect comment with regard to VLAN packet handling

The commit bcc6d4790361 ("net: vlan: make non-hw-accel rx path similar
to hw-accel") unified accel and non-accel path for VLAN RX. With that
fix we do not register any packet_type handler for VLANs anymore, so fix
the incorrect comment.

Signed-off-by: Girish Moodalbail <girish.moodalbail@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agocan: peak: Add support for new PCIe/M2 CAN FD interfaces
Stephane Grosjean [Thu, 9 Nov 2017 13:42:14 +0000 (14:42 +0100)]
can: peak: Add support for new PCIe/M2 CAN FD interfaces

This adds support for the following PEAK-System CAN FD interfaces:

PCAN-cPCIe FD         CAN FD Interface for cPCI Serial (2 or 4 channels)
PCAN-PCIe/104-Express CAN FD Interface for PCIe/104-Express (1, 2 or 4 ch.)
PCAN-miniPCIe FD      CAN FD Interface for PCIe Mini (1, 2 or 4 channels)
PCAN-PCIe FD OEM      CAN FD Interface for PCIe OEM version (1, 2 or 4 ch.)
PCAN-M.2              CAN FD Interface for M.2 (1 or 2 channels)

Like the PCAN-PCIe FD interface, all of these boards run the same IP Core
that is able to handle CAN FD (see also http://www.peak-system.com).

Signed-off-by: Stephane Grosjean <s.grosjean@peak-system.com>
Cc: linux-stable <stable@vger.kernel.org>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
6 years agocan: sun4i: handle overrun in RX FIFO
Gerhard Bertelsmann [Mon, 6 Nov 2017 17:16:56 +0000 (18:16 +0100)]
can: sun4i: handle overrun in RX FIFO

SUN4Is CAN IP has a 64 byte deep FIFO buffer. If the buffer is not
drained fast enough (overrun) it's getting mangled. Already received
frames are dropped - the data can't be restored.

Signed-off-by: Gerhard Bertelsmann <info@gerhard-bertelsmann.de>
Cc: linux-stable <stable@vger.kernel.org>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
6 years agocan: c_can: don't indicate triple sampling support for D_CAN
Richard Schütz [Sun, 29 Oct 2017 12:03:22 +0000 (13:03 +0100)]
can: c_can: don't indicate triple sampling support for D_CAN

The D_CAN controller doesn't provide a triple sampling mode, so don't set
the CAN_CTRLMODE_3_SAMPLES flag in ctrlmode_supported. Currently enabling
triple sampling is a no-op.

Signed-off-by: Richard Schütz <rschuetz@uni-koblenz.de>
Cc: linux-stable <stable@vger.kernel.org> # >= v3.6
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
6 years agonet/mlx5e: Increase Striding RQ minimum size limit to 4 multi-packet WQEs
Eugenia Emantayev [Thu, 12 Jan 2017 15:11:45 +0000 (17:11 +0200)]
net/mlx5e: Increase Striding RQ minimum size limit to 4 multi-packet WQEs

This is to prevent the case of working with a single MPWQE
(1 WQE is always reserved as RQ is linked-list).
When the WQE is fully consumed, HW should still have available buffer
in order not to drop packets.

Fixes: 461017cb006a ("net/mlx5e: Support RX multi-packet WQE (Striding RQ)")
Signed-off-by: Eugenia Emantayev <eugenia@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Cc: kernel-team@fb.com
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
6 years agonet/mlx5e: Set page to null in case dma mapping fails
Inbar Karmy [Sun, 15 Oct 2017 14:30:59 +0000 (17:30 +0300)]
net/mlx5e: Set page to null in case dma mapping fails

Currently, when dma mapping fails, put_page is called,
but the page is not set to null. Later, in the page_reuse treatment in
mlx5e_free_rx_descs(), mlx5e_page_release() is called for the second time,
improperly doing dma_unmap (for a non-mapped address) and an extra put_page.
Prevent this by nullifying the page pointer when dma_map fails.

Fixes: accd58833237 ("net/mlx5e: Introduce RX Page-Reuse")
Signed-off-by: Inbar Karmy <inbark@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Cc: kernel-team@fb.com
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
6 years agonet/mlx5e: Fix napi poll with zero budget
Saeed Mahameed [Tue, 31 Oct 2017 22:34:00 +0000 (15:34 -0700)]
net/mlx5e: Fix napi poll with zero budget

napi->poll can be called with budget 0, e.g. in netpoll scenarios
where the caller only wants to poll TX rings
(poll_one_napi@net/core/netpoll.c).

The below commit changed RX polling from "while" loop to "do {} while",
which caused to ignore the initial budget and handle at least one RX
packet.

This fixes the following warning:
[ 2852.049194] mlx5e_napi_poll+0x0/0x260 [mlx5_core] exceeded budget in poll
[ 2852.049195] ------------[ cut here ]------------
[ 2852.049195] WARNING: CPU: 0 PID: 25691 at net/core/netpoll.c:171 netpoll_poll_dev+0x18a/0x1a0

Fixes: 4b7dfc992514 ("net/mlx5e: Early-return on empty completion queues")
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Reported-by: Martin KaFai Lau <kafai@fb.com>
Tested-by: Martin KaFai Lau <kafai@fb.com>
Cc: kernel-team@fb.com
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
6 years agonet/mlx5: Cancel health poll before sending panic teardown command
Huy Nguyen [Tue, 26 Sep 2017 20:11:56 +0000 (15:11 -0500)]
net/mlx5: Cancel health poll before sending panic teardown command

After the panic teardown firmware command, health_care detects the error
in PCI bus and calls the mlx5_pci_err_detected. This health_care flow is
no longer needed because the panic teardown firmware command will bring
down the PCI bus communication with the HCA.

The solution is to cancel the health care timer and its pending
workqueue request before sending panic teardown firmware command.

Kernel trace:
mlx5_core 0033:01:00.0: Shutdown was called
mlx5_core 0033:01:00.0: health_care:154:(pid 9304): handling bad device here
mlx5_core 0033:01:00.0: mlx5_handle_bad_state:114:(pid 9304): NIC state 1
mlx5_core 0033:01:00.0: mlx5_pci_err_detected was called
mlx5_core 0033:01:00.0: mlx5_enter_error_state:96:(pid 9304): start
mlx5_3:mlx5_ib_event:3061:(pid 9304): warning: event on port 0
mlx5_core 0033:01:00.0: mlx5_enter_error_state:104:(pid 9304): end
Unable to handle kernel paging request for data at address 0x0000003f
Faulting instruction address: 0xc0080000434b8c80

Fixes: 8812c24d28f4 ('net/mlx5: Add fast unload support in shutdown flow')
Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Moshe Shemesh <moshe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>