openwrt/staging/blogic.git
7 years agonet/mlx5e: Add dcbnl dscp to priority support
Huy Nguyen [Tue, 18 Jul 2017 21:23:36 +0000 (16:23 -0500)]
net/mlx5e: Add dcbnl dscp to priority support

This patch implements dcbnl hooks to set and delete DSCP to priority map
as defined by the DCB subsystem. Device maintains internal trust state
which needs to be set to DSCP state for performing DSCP to priority mapping.

When the first dscp to priority APP entry is added by the user, the
trust state is changed to dscp.

When the last dscp to priority APP entry is deleted by the user, the
trust state is changed to pcp.

If user sends multiple dscp to priority APP entries on the same dscp,
the last sent one will take effect. All the previous sent will be
deleted.

The dscp to priority APP entries are added and deleted in the net/dcb
APP database using dcb_ieee_setapp/getapp.

Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Parav Pandit <parav@mellanox.com>
Reviewed-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
7 years agonet/mlx5: QPTS and QPDPM register firmware command support
Huy Nguyen [Tue, 18 Jul 2017 21:08:46 +0000 (16:08 -0500)]
net/mlx5: QPTS and QPDPM register firmware command support

The QPTS register allows changing the priority trust state between pcp and
dscp. Add support to get/set trust state from device. When the port is
in pcp/dscp trust state, packet is routed by hardware to matching priority
based on its pcp/dscp value respectively.

The QPDPM register allow channing the dscp to priority mapping. Add support
to get/set dscp to priority mapping from device.
Note that to change a dscp mapping, the "e" bit of this dscp structure
must be set in the QPDPM firmware command.

Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Parav Pandit <parav@mellanox.com>
Reviewed-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
7 years agonet/mlx5: Add MLX5_SET16 and MLX5_GET16
Huy Nguyen [Thu, 3 Aug 2017 02:36:23 +0000 (21:36 -0500)]
net/mlx5: Add MLX5_SET16 and MLX5_GET16

Add MLX5_SET16 and MLX5_GET16 for 16bit structure field in firmware
command.

Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Parav Pandit <parav@mellanox.com>
Reviewed-by: Eli Cohen <eli@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
7 years agonet/mlx5: QCAM register firmware command support
Huy Nguyen [Tue, 18 Jul 2017 21:03:17 +0000 (16:03 -0500)]
net/mlx5: QCAM register firmware command support

The QCAM register provides capability bit for all the QoS registers
using ACCESS_REG command.

Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Parav Pandit <parav@mellanox.com>
Reviewed-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
7 years agonet/dcb: Add dscp to priority selector type
Huy Nguyen [Tue, 18 Jul 2017 20:42:15 +0000 (15:42 -0500)]
net/dcb: Add dscp to priority selector type

IEEE specification P802.1Qcd/D2.1 defines priority selector 5.
This APP TLV selector defines DSCP to priority map.
This patch defines such DSCP selector.

Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Parav Pandit <parav@mellanox.com>
Reviewed-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
7 years agoipv6: remove IN6_ADDR_HSIZE from addrconf.h
Eric Dumazet [Sat, 4 Nov 2017 15:53:27 +0000 (08:53 -0700)]
ipv6: remove IN6_ADDR_HSIZE from addrconf.h

IN6_ADDR_HSIZE is private to addrconf.c, move it here to avoid
confusion.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agopktgen: do not abuse IN6_ADDR_HSIZE
Eric Dumazet [Sat, 4 Nov 2017 15:27:14 +0000 (08:27 -0700)]
pktgen: do not abuse IN6_ADDR_HSIZE

pktgen accidentally used IN6_ADDR_HSIZE, instead of using the size of an
IPv6 address.

Since IN6_ADDR_HSIZE recently was increased from 16 to 256, this old
bug is hitting us.

Fixes: 3f27fb23219e ("ipv6: addrconf: add per netns perturbation in inet6_addr_hash()")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: sched: cls_u32: use bitwise & rather than logical && on n->flags
Colin Ian King [Fri, 3 Nov 2017 08:09:45 +0000 (08:09 +0000)]
net: sched: cls_u32: use bitwise & rather than logical && on n->flags

Currently n->flags is being operated on by a logical && operator rather
than a bitwise & operator. This looks incorrect as these should be bit
flag operations. Fix this.

Detected by CoverityScan, CID#1460398 ("Logical vs. bitwise operator")

Fixes: 245dc5121a9b ("net: sched: cls_u32: call block callbacks for offload")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agotcp_nv: use do_div() instead of expensive div64_u64()
Konstantin Khlebnikov [Thu, 2 Nov 2017 14:07:05 +0000 (17:07 +0300)]
tcp_nv: use do_div() instead of expensive div64_u64()

Average RTT is 32-bit thus full 64-bit division is redundant.

Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Suggested-by: Stephen Hemminger <stephen@networkplumber.org>
Suggested-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoadd support of IFF_XMIT_DST_RELEASE bit in vlan
Vadim Fedorenko [Thu, 2 Nov 2017 12:49:08 +0000 (15:49 +0300)]
add support of IFF_XMIT_DST_RELEASE bit in vlan

Some time ago Eric Dumazet suggested a "hack the IFF_XMIT_DST_RELEASE
flag on the vlan netdev". But the last comment was "does not support
properly bonding/team.(If the real_dev->privflags IFF_XMIT_DST_RELEASE
bit changes, we want to update all the vlans at the same time )"

I've extended that patch to support changes of IFF_XMIT_DST_RELEASE in
bonding/team.
Both bonding and team call netdev_change_features() after recalculation
of features including priv_flags IFF_XMIT_DST_RELEASE bit. So the only
thing needed to support is to recheck this bit in
vlan_transfer_features().

Suggested-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Vadim Fedorenko <vfedorenko@yandex-team.ru>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agophylink: make local function phylink_phy_change() static
Wei Yongjun [Thu, 2 Nov 2017 11:14:48 +0000 (11:14 +0000)]
phylink: make local function phylink_phy_change() static

Fixes the following sparse warnings:

drivers/net/phy/phylink.c:570:6: warning:
 symbol 'phylink_phy_change' was not declared. Should it be static?

Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoMerge tag 'wireless-drivers-next-for-davem-2017-11-03' of git://git.kernel.org/pub...
David S. Miller [Sat, 4 Nov 2017 09:07:50 +0000 (18:07 +0900)]
Merge tag 'wireless-drivers-next-for-davem-2017-11-03' of git://git./linux/kernel/git/kvalo/wireless-drivers-next

Kalle Valo says:

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

Mostly fixes this time, but also few new features.

Major changes:

wil6210

* remove ssid debugfs file

rsi

* add WOWLAN support for suspend, hibernate and shutdown states

ath10k

* add support for CCMP-256, GCMP and GCMP-256 ciphers on hardware
  where it's supported (QCA99x0 and QCA4019)
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
David S. Miller [Sat, 4 Nov 2017 00:26:51 +0000 (09:26 +0900)]
Merge git://git./linux/kernel/git/davem/net

Files removed in 'net-next' had their license header updated
in 'net'.  We take the remove from 'net-next'.

Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoliquidio: Fix an issue with multiple switchdev enable disables
Vijaya Mohan Guvva [Fri, 3 Nov 2017 19:17:44 +0000 (12:17 -0700)]
liquidio: Fix an issue with multiple switchdev enable disables

Return success if the same dispatch function is being registered for
a given opcode and subcode, there by allow multiple switchdev enable
and disables.

Signed-off-by: Vijaya Mohan Guvva <vijaya.guvva@cavium.com>
Signed-off-by: Satanand Burla <satananda.burla@cavium.com>
Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoMerge branch 'mlxsw-Handle-changes-in-GRE-configuration'
David S. Miller [Sat, 4 Nov 2017 00:15:18 +0000 (09:15 +0900)]
Merge branch 'mlxsw-Handle-changes-in-GRE-configuration'

Jiri Pirko says:

====================
mlxsw: Handle changes in GRE configuration

Petr says:

Until now, when an IP tunnel was offloaded by the mlxsw driver, the
offload was pretty much static, and changes in Linux configuration were
not reflected in the hardware. That led to discrepancies between traffic
flows in slow path and fast path. The work-around used to be to remove
all routes that forward to the netdevice and re-add them. This is
clearly suboptimal, but actually, as of the decap-only patchset, it's
not even enough anymore, and one needs to go all the way and simply drop
the tunnel and recreate it correctly.

With this patchset, the NETDEV_CHANGE events that are generated for
changes of up'd tunnel netdevices are captured and interpreted to
correctly reconfigure the HW in accordance with changes requested at the
software layer. In addition, NETDEV_CHANGEUPPER, NETDEV_UP and
NETDEV_DOWN are now handled not only for tunnel devices themselves, but
also for their bound devices. Each change is then translated to one or
more of the following updates to the HW configuration:

- refresh of offload of local route that corresponds to tunnel's local
  address
- refresh of the loopback RIF
- refresh of offloads of routes that forward to the changed tunnel
- removal of tunnel offloads

These tools are used to implement the following configuration changes:

- addition of a new offloadable tunnel with local address that conflicts
  with that of an already-offloaded tunnel (the existing tunnel is
  onloaded, the new one isn't offloaded)
- changes to TTL, TOS that make tunnel unsuitable for offloading
- changes to ikey, okey, remote
- changes to local, which when they cause conflict with another
  tunnel, lead to onloading of both newly-conflicting tunnels
- migration of a bound device of an offloaded tunnel device to a
  different VRF
- changes to what device is bound to a tunnel device (i.e. like what
  "ip tunnel change name g dev another" does)
- changes to up / down state of a bound device. A down bound device
  doesn't forward encapsulated traffic anymore, but decap still works.

This patchset starts with a suite of patches that adapt the existing
code base step by step to facilitate introduction of the offloading
code. The five substantial patches at the end then implement the changes
mentioned above.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agomlxsw: spectrum_router: Handle down of tunnel underlay
Petr Machata [Fri, 3 Nov 2017 09:03:44 +0000 (10:03 +0100)]
mlxsw: spectrum_router: Handle down of tunnel underlay

When the bound device of a tunnel device is down, encapsulated packets
are not egressed anymore, but tunnel decap still works. Extend
mlxsw_sp_nexthop_rif_update() to take IFF_UP into consideration when
deciding whether a given next hop should be offloaded.

Because the new logic was added to mlxsw_sp_nexthop_rif_update(), this
fixes the case where a newly-added tunnel has a down bound device, which
would previously be fully offloaded. Now the down state of the bound
device is noted and next hops forwarding to such tunnel are not
offloaded.

In addition to that, notice NETDEV_UP and NETDEV_DOWN of a bound device
to force refresh of tunnel encap route offloads.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agomlxsw: spectrum_ipip: Handle underlay device change
Petr Machata [Fri, 3 Nov 2017 09:03:43 +0000 (10:03 +0100)]
mlxsw: spectrum_ipip: Handle underlay device change

When a bound device of an IP-in-IP tunnel changes, such as through
'ip tunnel change name $name dev $dev', the loopback backing the tunnel
needs to be recreated.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agomlxsw: spectrum: Handle NETDEV_CHANGE on L3 tunnels
Petr Machata [Fri, 3 Nov 2017 09:03:42 +0000 (10:03 +0100)]
mlxsw: spectrum: Handle NETDEV_CHANGE on L3 tunnels

Changes to L3 tunnel netdevices (through `ip tunnel change' as well as
`ip link set') lead to NETDEV_CHANGE being generated on the tunnel
device. Because what is relevant for the tunnel in question depends on
the tunnel type, handling of the event is dispatched to the IPIP module
through a newly-added interface mlxsw_sp_ipip_ops.ol_netdev_change().

IPIP tunnels now remember the last set of tunnel parameters in struct
mlxsw_sp_ipip_entry.parms, and use it to figure out what exactly has
changed.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agomlxsw: spectrum: Support IPIP underlay VRF migration
Petr Machata [Fri, 3 Nov 2017 09:03:41 +0000 (10:03 +0100)]
mlxsw: spectrum: Support IPIP underlay VRF migration

When a bound device of a tunnel netdevice changes VRF, the loopback RIF
that backs the tunnel needs to be updated and existing encapsulating
routes need to be refreshed.

Note that several tunnels can share the same bound device, in which case
all the impacted tunnels need to be updated.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agomlxsw: spectrum_router: Onload conflicting tunnels
Petr Machata [Fri, 3 Nov 2017 09:03:40 +0000 (10:03 +0100)]
mlxsw: spectrum_router: Onload conflicting tunnels

The approach for offloading IP tunnels implemented currently by mlxsw
doesn't allow two tunnels that have the same local IP address in the
same (underlay) VRF. Previously, offloads were introduced on demand as
encap routes were formed. When such a route was created that would cause
offload of a conflicting tunnel, mlxsw_sp_ipip_entry_create() would
detect it and return -EEXIST, which would propagate up and cause FIB
abort.

Now however IPIP entries are created as soon as an offloadable netdevice
is created, and the failure prevents creation of such device.
Furthermore, if the driver is installed at the point where such
conflicting tunnels exist, the failure actually prevents successful
modprobe.

Furthermore, follow-up patches implement handling of NETDEV_CHANGE due
to the local address change. However, NETDEV_CHANGE can't be vetoed. The
failure merely means that the offloads weren't updated, but the change
in Linux configuration is not rolled back. It is thus desirable to have
a robust way of handling these conflicts, which can later be reused for
handling NETDEV_CHANGE as well.

To fix this, when a conflicting tunnel is created, instead of failing,
simply pull the old tunnel to slow path and reject offloading the
new one.

Introduce two functions: mlxsw_sp_ipip_entry_demote_tunnel() and
mlxsw_sp_ipip_demote_tunnel_by_saddr() to handle this. Make them both
public, because they will be useful later on in this patchset.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agomlxsw: spectrum_router: Fix saddr deduction in mlxsw_sp_ipip_entry_create()
Petr Machata [Fri, 3 Nov 2017 09:03:39 +0000 (10:03 +0100)]
mlxsw: spectrum_router: Fix saddr deduction in mlxsw_sp_ipip_entry_create()

When trying to determine whether there are other offloaded tunnels with
the same local address, mlxsw_sp_ipip_entry_create() should look for a
tunnel with matching UL protocol, matching saddr, in the same VRF.
However instead of taking into account the UL protocol of the tunnel
netdevice (which mlxsw_sp_ipip_entry_saddr_matches() then compares to
the UL protocol of inspected IPIP entry), it deduces the UL protocol
from the inspected IPIP entry (and that's compared to itself).

This is currently immaterial, because only one tunnel type is offloaded,
and therefore the UL protocol always matches, but introducing support
for a tunnel with IPv6 underlay would uncover this error.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agomlxsw: spectrum_router: Generalize __mlxsw_sp_ipip_entry_update_tunnel()
Petr Machata [Fri, 3 Nov 2017 09:03:38 +0000 (10:03 +0100)]
mlxsw: spectrum_router: Generalize __mlxsw_sp_ipip_entry_update_tunnel()

The work that needs to be done to update HW configuration in response to
changes is similar to what __mlxsw_sp_ipip_entry_update_tunnel() already
does, but with a number of twists: each change requires a different
subset of things to happen. Extend the function to support all these
uses, and allow finely-grained configuration of what should happen at
each call through a suite of function arguments.

Publish the updated function to allow use from the spectrum_ipip module.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agomlxsw: spectrum_router: Extract __mlxsw_sp_ipip_entry_update_tunnel()
Petr Machata [Fri, 3 Nov 2017 09:03:37 +0000 (10:03 +0100)]
mlxsw: spectrum_router: Extract __mlxsw_sp_ipip_entry_update_tunnel()

The work that's done by mlxsw_sp_netdevice_ipip_ol_vrf_event() is a good
basis for a more versatile function that would take care of all sorts of
tunnel updates requests: __mlxsw_sp_ipip_entry_update_tunnel(). Extract
that function. Factor out a helper mlxsw_sp_ipip_entry_ol_lb_update() as
well.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agomlxsw: spectrum: Propagate extack for tunnel events
Petr Machata [Fri, 3 Nov 2017 09:03:36 +0000 (10:03 +0100)]
mlxsw: spectrum: Propagate extack for tunnel events

The function mlxsw_sp_rif_create() takes an extack parameter. So far,
for creation of loopback interfaces, NULL was passed. For some events
however the extack can be extracted and passed along. So do that for
NETDEV_CHANGEUPPER handler.

Use the opportunity to update the type of info argument that
mlxsw_sp_netdevice_ipip_ol_event() takes. Follow-up patches will
introduce handling of more changes, and some of them carry an extack as
well, but in an info structure of a different type. Though not strictly
erroneous (the pointer could be cast whichever way), it makes no sense
to pretend the value is always of a certain type, when in fact it isn't.
So change the prototype of the above-mentioned function as well.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agomlxsw: spectrum_router: Extract mlxsw_sp_ipip_entry_ol_up_event()
Petr Machata [Fri, 3 Nov 2017 09:03:35 +0000 (10:03 +0100)]
mlxsw: spectrum_router: Extract mlxsw_sp_ipip_entry_ol_up_event()

The piece of logic to promote decap route, if any, is useful for generic
tunnel updates, not just for handling of NETDEV_UP events on tunnel
interfaces. Extract it to a separate function.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agomlxsw: spectrum_router: Make mlxsw_sp_netdevice_ipip_ol_up_event() void
Petr Machata [Fri, 3 Nov 2017 09:03:34 +0000 (10:03 +0100)]
mlxsw: spectrum_router: Make mlxsw_sp_netdevice_ipip_ol_up_event() void

This function only ever returns 0, so don't pretend it returns anything
useful and just make it void.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agomlxsw: spectrum_router: Extract mlxsw_sp_ipip_entry_ol_down_event()
Petr Machata [Fri, 3 Nov 2017 09:03:33 +0000 (10:03 +0100)]
mlxsw: spectrum_router: Extract mlxsw_sp_ipip_entry_ol_down_event()

Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agomlxsw: spectrum_ipip: Split accessor functions
Petr Machata [Fri, 3 Nov 2017 09:03:32 +0000 (10:03 +0100)]
mlxsw: spectrum_ipip: Split accessor functions

To implement NETDEV_CHANGE notifications on IP-in-IP tunnels, the
handler needs to figure out what actually changed, to understand how
exactly to update the offloads. It will do so by storing struct
ip_tunnel_parm with previous configuration, and comparing that to the
new version.

To facilitate these comparisons, extract the code that operates on
struct ip_tunnel_parm from the existing accessor functions, and make
those a thin wrapper that extracts tunnel parameters and dispatches.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agomlxsw: spectrum: Move mlxsw_sp_ipip_netdev_{s, d}addr{, 4}()
Petr Machata [Fri, 3 Nov 2017 09:03:31 +0000 (10:03 +0100)]
mlxsw: spectrum: Move mlxsw_sp_ipip_netdev_{s, d}addr{, 4}()

These functions ideologically belong to the IPIP module, and some
follow-up work will benefit from their presence there.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agomlxsw: spectrum_router: Extract mlxsw_sp_netdevice_ipip_can_offload()
Petr Machata [Fri, 3 Nov 2017 09:03:30 +0000 (10:03 +0100)]
mlxsw: spectrum_router: Extract mlxsw_sp_netdevice_ipip_can_offload()

Some of the code down the road needs this logic as well.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agomlxsw: spectrum: Rename IPIP-related netdevice handlers
Petr Machata [Fri, 3 Nov 2017 09:03:29 +0000 (10:03 +0100)]
mlxsw: spectrum: Rename IPIP-related netdevice handlers

To distinguish between events related to tunnel device itself and its
bound device, rename a number of functions related to handling tunneling
netdevice events to include _ol_ (for "overlay") in the name. That
leaves room in the namespace for underlay-related functions, which would
have _ul_ in the name.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoMerge tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 3 Nov 2017 20:56:15 +0000 (13:56 -0700)]
Merge tag 'clk-fixes-for-linus' of git://git./linux/kernel/git/clk/linux

Pull clk fix from Stephen Boyd:
 "One fix for USB clks on Uniphier PXs3 SoCs"

* tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/clk/linux:
  clk: uniphier: fix clock data for PXs3

7 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux-tile
Linus Torvalds [Fri, 3 Nov 2017 17:36:43 +0000 (10:36 -0700)]
Merge git://git./linux/kernel/git/cmetcalf/linux-tile

Pull arch/tile fixes from Chris Metcalf:
 "Two one-line bug fixes"

* git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux-tile:
  arch/tile: Implement ->set_state_oneshot_stopped()
  tile: pass machine size to sparse

7 years agoMerge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Linus Torvalds [Fri, 3 Nov 2017 17:26:40 +0000 (10:26 -0700)]
Merge tag 'scsi-fixes' of git://git./linux/kernel/git/jejb/scsi

Pull SCSI fix from James Bottomley:
 "One minor fix in the error leg of the qla2xxx driver (it oopses the
  system if we get an error trying to start the internal kernel thread).

  The fix is minor because the problem isn't often encountered in the
  field (although it can be induced by inserting the module in a low
  memory environment)"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: qla2xxx: Fix oops in qla2x00_probe_one error path

7 years agoarch/tile: Implement ->set_state_oneshot_stopped()
Chris Metcalf [Fri, 3 Nov 2017 17:20:54 +0000 (13:20 -0400)]
arch/tile: Implement ->set_state_oneshot_stopped()

set_state_oneshot_stopped() is called by the clkevt core, when the
next event is required at an expiry time of 'KTIME_MAX'. This normally
happens with NO_HZ_{IDLE|FULL} in both LOWRES/HIGHRES modes.

This patch makes the clockevent device to stop on such an event, to
avoid spurious interrupts, as explained by: commit 8fff52fd5093
("clockevents: Introduce CLOCK_EVT_STATE_ONESHOT_STOPPED state").

Signed-off-by: Chris Metcalf <cmetcalf@mellanox.com>
7 years agoMerge tag 'powerpc-4.14-6' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc...
Linus Torvalds [Fri, 3 Nov 2017 16:25:53 +0000 (09:25 -0700)]
Merge tag 'powerpc-4.14-6' of git://git./linux/kernel/git/powerpc/linux

Pull powerpc fixes from Michael Ellerman:
 "Some more powerpc fixes for 4.14.

  This is bigger than I like to send at rc7, but that's at least partly
  because I didn't send any fixes last week. If it wasn't for the IMC
  driver, which is new and getting heavy testing, the diffstat would
  look a bit better. I've also added ftrace on big endian to my test
  suite, so we shouldn't break that again in future.

   - A fix to the handling of misaligned paste instructions (P9 only),
     where a change to a #define has caused the check for the
     instruction to always fail.

   - The preempt handling was unbalanced in the radix THP flush (P9
     only). Though we don't generally use preempt we want to keep it
     working as much as possible.

   - Two fixes for IMC (P9 only), one when booting with restricted
     number of CPUs and one in the error handling when initialisation
     fails due to firmware etc.

   - A revert to fix function_graph on big endian machines, and then a
     rework of the reverted patch to fix kprobes blacklist handling on
     big endian machines.

  Thanks to: Anju T Sudhakar, Guilherme G. Piccoli, Madhavan Srinivasan,
  Naveen N. Rao, Nicholas Piggin, Paul Mackerras"

* tag 'powerpc-4.14-6' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  powerpc/perf: Fix core-imc hotplug callback failure during imc initialization
  powerpc/kprobes: Dereference function pointers only if the address does not belong to kernel text
  Revert "powerpc64/elfv1: Only dereference function descriptor for non-text symbols"
  powerpc/64s/radix: Fix preempt imbalance in TLB flush
  powerpc: Fix check for copy/paste instructions in alignment handler
  powerpc/perf: Fix IMC allocation routine

7 years agoMerge tag 'mmc-v4.14-rc4-3' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc
Linus Torvalds [Fri, 3 Nov 2017 16:19:20 +0000 (09:19 -0700)]
Merge tag 'mmc-v4.14-rc4-3' of git://git./linux/kernel/git/ulfh/mmc

Pull MMC fixes from Ulf Hansson:
 "Fix dw_mmc request timeout issues"

* tag 'mmc-v4.14-rc4-3' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc:
  mmc: dw_mmc: Fix the DTO timeout calculation
  mmc: dw_mmc: Add locking to the CTO timer
  mmc: dw_mmc: Fix the CTO timeout calculation
  mmc: dw_mmc: cancel the CTO timer after a voltage switch

7 years agoMerge tag 'drm-fixes-for-v4.14-rc8' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Fri, 3 Nov 2017 16:14:22 +0000 (09:14 -0700)]
Merge tag 'drm-fixes-for-v4.14-rc8' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:

 - one nouveau regression fix

 - some amdgpu fixes for stable to fix hangs on some harvested Polaris
   GPUs

 - a set of KASAN and regression fixes for i915, their CI system seems
   to be working pretty well now.

* tag 'drm-fixes-for-v4.14-rc8' of git://people.freedesktop.org/~airlied/linux:
  drm/amdgpu: allow harvesting check for Polaris VCE
  drm/amdgpu: return -ENOENT from uvd 6.0 early init for harvesting
  drm/i915: Check incoming alignment for unfenced buffers (on i915gm)
  drm/nouveau/kms/nv50: use the correct state for base channel notifier setup
  drm/i915: Hold rcu_read_lock when iterating over the radixtree (vma idr)
  drm/i915: Hold rcu_read_lock when iterating over the radixtree (objects)
  drm/i915/edp: read edp display control registers unconditionally
  drm/i915: Do not rely on wm preservation for ILK watermarks
  drm/i915: Cancel the modeset retry work during modeset cleanup

7 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Fri, 3 Nov 2017 16:09:21 +0000 (09:09 -0700)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:
 "Hopefully this is the last batch of networking fixes for 4.14

  Fingers crossed...

   1) Fix stmmac to use the proper sized OF property read, from Bhadram
      Varka.

   2) Fix use after free in net scheduler tc action code, from Cong
      Wang.

   3) Fix SKB control block mangling in tcp_make_synack().

   4) Use proper locking in fib_dump_info(), from Florian Westphal.

   5) Fix IPG encodings in systemport driver, from Florian Fainelli.

   6) Fix division by zero in NV TCP congestion control module, from
      Konstantin Khlebnikov.

   7) Fix use after free in nf_reject_ipv4, from Tejaswi Tanikella"

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net:
  net: systemport: Correct IPG length settings
  tcp: do not mangle skb->cb[] in tcp_make_synack()
  fib: fib_dump_info can no longer use __in_dev_get_rtnl
  stmmac: use of_property_read_u32 instead of read_u8
  net_sched: hold netns refcnt for each action
  net_sched: acquire RTNL in tc_action_net_exit()
  net: vrf: correct FRA_L3MDEV encode type
  tcp_nv: fix division by zero in tcpnv_acked()
  netfilter: nf_reject_ipv4: Fix use-after-free in send_reset
  netfilter: nft_set_hash: disable fast_ops for 2-len keys

7 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Fri, 3 Nov 2017 16:03:50 +0000 (09:03 -0700)]
Merge branch 'akpm' (patches from Andrew)

Merge misc fixes from Andrew Morton:
 "7 fixes"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
  mm, swap: fix race between swap count continuation operations
  mm/huge_memory.c: deposit page table when copying a PMD migration entry
  initramfs: fix initramfs rebuilds w/ compression after disabling
  fs/hugetlbfs/inode.c: fix hwpoison reserve accounting
  ocfs2: fstrim: Fix start offset of first cluster group during fstrim
  mm, /proc/pid/pagemap: fix soft dirty marking for PMD migration entry
  userfaultfd: hugetlbfs: prevent UFFDIO_COPY to fill beyond the end of i_size

7 years agoUpdate MIPS email addresses
Paul Burton [Thu, 26 Oct 2017 00:04:33 +0000 (17:04 -0700)]
Update MIPS email addresses

MIPS will soon not be a part of Imagination Technologies, and as such
many @imgtec.com email addresses will no longer be valid. This patch
updates the addresses for those who:

 - Have 10 or more patches in mainline authored using an @imgtec.com
   email address, or any patches dated within the past year.

 - Are still with Imagination but leaving as part of the MIPS business
   unit, as determined from an internal email address list.

 - Haven't already updated their email address (ie. JamesH) or expressed
   a desire to be excluded (ie. Maciej).

 - Acked v2 or earlier of this patch, which leaves Deng-Cheng, Matt &
   myself.

New addresses are of the form firstname.lastname@mips.com, and all
verified against an internal email address list.  An entry is added to
.mailmap for each person such that get_maintainer.pl will report the new
addresses rather than @imgtec.com addresses which will soon be dead.

Instances of the affected addresses throughout the tree are then
mechanically replaced with the new @mips.com address.

Signed-off-by: Paul Burton <paul.burton@mips.com>
Cc: Deng-Cheng Zhu <dengcheng.zhu@imgtec.com>
Cc: Deng-Cheng Zhu <dengcheng.zhu@mips.com>
Acked-by: Dengcheng Zhu <dengcheng.zhu@mips.com>
Cc: Matt Redfearn <matt.redfearn@imgtec.com>
Cc: Matt Redfearn <matt.redfearn@mips.com>
Acked-by: Matt Redfearn <matt.redfearn@mips.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: linux-kernel@vger.kernel.org
Cc: linux-mips@linux-mips.org
Cc: trivial@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agox86: CPU: Fix up "cpu MHz" in /proc/cpuinfo
Rafael J. Wysocki [Fri, 3 Nov 2017 15:35:49 +0000 (16:35 +0100)]
x86: CPU: Fix up "cpu MHz" in /proc/cpuinfo

Commit 890da9cf0983 (Revert "x86: do not use cpufreq_quick_get() for
/proc/cpuinfo "cpu MHz"") is not sufficient to restore the previous
behavior of "cpu MHz" in /proc/cpuinfo on x86 due to some changes
made after the commit it has reverted.

To address this, make the code in question use arch_freq_get_on_cpu()
which also is used by cpufreq for reporting the current frequency of
CPUs and since that function doesn't really depend on cpufreq in any
way, drop the CONFIG_CPU_FREQ dependency for the object file
containing it.

Also refactor arch_freq_get_on_cpu() somewhat to avoid IPIs and
return cached values right away if it is called very often over a
short time (to prevent user space from triggering IPI storms through
it).

Fixes: 890da9cf0983 (Revert "x86: do not use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz"")
Cc: stable@kernel.org # 4.13 - together with 890da9cf0983
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm, swap: fix race between swap count continuation operations
Huang Ying [Thu, 2 Nov 2017 22:59:50 +0000 (15:59 -0700)]
mm, swap: fix race between swap count continuation operations

One page may store a set of entries of the sis->swap_map
(swap_info_struct->swap_map) in multiple swap clusters.

If some of the entries has sis->swap_map[offset] > SWAP_MAP_MAX,
multiple pages will be used to store the set of entries of the
sis->swap_map.  And the pages are linked with page->lru.  This is called
swap count continuation.  To access the pages which store the set of
entries of the sis->swap_map simultaneously, previously, sis->lock is
used.  But to improve the scalability of __swap_duplicate(), swap
cluster lock may be used in swap_count_continued() now.  This may race
with add_swap_count_continuation() which operates on a nearby swap
cluster, in which the sis->swap_map entries are stored in the same page.

The race can cause wrong swap count in practice, thus cause unfreeable
swap entries or software lockup, etc.

To fix the race, a new spin lock called cont_lock is added to struct
swap_info_struct to protect the swap count continuation page list.  This
is a lock at the swap device level, so the scalability isn't very well.
But it is still much better than the original sis->lock, because it is
only acquired/released when swap count continuation is used.  Which is
considered rare in practice.  If it turns out that the scalability
becomes an issue for some workloads, we can split the lock into some
more fine grained locks.

Link: http://lkml.kernel.org/r/20171017081320.28133-1-ying.huang@intel.com
Fixes: 235b62176712 ("mm/swap: add cluster lock")
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Shaohua Li <shli@kernel.org>
Cc: Tim Chen <tim.c.chen@intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Aaron Lu <aaron.lu@intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: <stable@vger.kernel.org> [4.11+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm/huge_memory.c: deposit page table when copying a PMD migration entry
Zi Yan [Thu, 2 Nov 2017 22:59:47 +0000 (15:59 -0700)]
mm/huge_memory.c: deposit page table when copying a PMD migration entry

We need to deposit pre-allocated PTE page table when a PMD migration
entry is copied in copy_huge_pmd().  Otherwise, we will leak the
pre-allocated page and cause a NULL pointer dereference later in
zap_huge_pmd().

The missing counters during PMD migration entry copy process are added
as well.

The bug report is here: https://lkml.org/lkml/2017/10/29/214

Link: http://lkml.kernel.org/r/20171030144636.4836-1-zi.yan@sent.com
Fixes: 84c3fc4e9c563 ("mm: thp: check pmd migration entry in common path")
Signed-off-by: Zi Yan <zi.yan@cs.rutgers.edu>
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agoinitramfs: fix initramfs rebuilds w/ compression after disabling
Florian Fainelli [Thu, 2 Nov 2017 22:59:44 +0000 (15:59 -0700)]
initramfs: fix initramfs rebuilds w/ compression after disabling

This is a follow-up to commit 57ddfdaa9a72 ("initramfs: fix disabling of
initramfs (and its compression)").  This particular commit fixed the use
case where we build the kernel with an initramfs with no compression,
and then we build the kernel with no initramfs.

Now this still left us with the same case as described here:

  http://lkml.kernel.org/r/20170521033337.6197-1-f.fainelli@gmail.com

not working with initramfs compression.  This can be seen by the
following steps/timestamps:

  https://www.spinics.net/lists/kernel/msg2598153.html

.initramfs_data.cpio.gz.cmd is correct:

  cmd_usr/initramfs_data.cpio.gz := /bin/bash
  ./scripts/gen_initramfs_list.sh -o usr/initramfs_data.cpio.gz  -u 1000 -g 1000  /home/fainelli/work/uclinux-rootfs/romfs /home/fainelli/work/uclinux-rootfs/misc/initramfs.dev

and was generated the first time we did generate the gzip initramfs, so
the command has not changed, nor its arguments, so we just don't call
it, no initramfs cpio is re-generated as a consequence.

The fix for this problem is just to properly keep track of the
.initramfs_cpio_data.d file by suffixing it with the compression
extension.  This takes care of properly tracking dependencies such that
the initramfs get (re)generated any time files are added/deleted etc.

Link: http://lkml.kernel.org/r/20170930033936.6722-1-f.fainelli@gmail.com
Fixes: db2aa7fd15e8 ("initramfs: allow again choice of the embedded initramfs compression algorithm")
Fixes: 9e3596b0c653 ("kbuild: initramfs cleanup, set target from Kconfig")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Cc: "Francisco Blas Izquierdo Riera (klondike)" <klondike@xiscosoft.net>
Cc: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agofs/hugetlbfs/inode.c: fix hwpoison reserve accounting
Mike Kravetz [Thu, 2 Nov 2017 22:59:41 +0000 (15:59 -0700)]
fs/hugetlbfs/inode.c: fix hwpoison reserve accounting

Calling madvise(MADV_HWPOISON) on a hugetlbfs page will result in bad
(negative) reserved huge page counts.  This may not happen immediately,
but may happen later when the underlying file is removed or filesystem
unmounted.  For example:

  AnonHugePages:         0 kB
  ShmemHugePages:        0 kB
  HugePages_Total:       1
  HugePages_Free:        0
  HugePages_Rsvd:    18446744073709551615
  HugePages_Surp:        0
  Hugepagesize:       2048 kB

In routine hugetlbfs_error_remove_page(), hugetlb_fix_reserve_counts is
called after remove_huge_page.  hugetlb_fix_reserve_counts is designed
to only be called/used only if a failure is returned from
hugetlb_unreserve_pages.  Therefore, call hugetlb_unreserve_pages as
required and only call hugetlb_fix_reserve_counts in the unlikely event
that hugetlb_unreserve_pages returns an error.

Link: http://lkml.kernel.org/r/20171019230007.17043-2-mike.kravetz@oracle.com
Fixes: 78bb920344b8 ("mm: hwpoison: dissolve in-use hugepage in unrecoverable memory error")
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Acked-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Aneesh Kumar <aneesh.kumar@linux.vnet.ibm.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agoocfs2: fstrim: Fix start offset of first cluster group during fstrim
Ashish Samant [Thu, 2 Nov 2017 22:59:37 +0000 (15:59 -0700)]
ocfs2: fstrim: Fix start offset of first cluster group during fstrim

The first cluster group descriptor is not stored at the start of the
group but at an offset from the start.  We need to take this into
account while doing fstrim on the first cluster group.  Otherwise we
will wrongly start fstrim a few blocks after the desired start block and
the range can cross over into the next cluster group and zero out the
group descriptor there.  This can cause filesytem corruption that cannot
be fixed by fsck.

Link: http://lkml.kernel.org/r/1507835579-7308-1-git-send-email-ashish.samant@oracle.com
Signed-off-by: Ashish Samant <ashish.samant@oracle.com>
Reviewed-by: Junxiao Bi <junxiao.bi@oracle.com>
Reviewed-by: Joseph Qi <jiangqi903@gmail.com>
Cc: Mark Fasheh <mfasheh@versity.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm, /proc/pid/pagemap: fix soft dirty marking for PMD migration entry
Huang Ying [Thu, 2 Nov 2017 22:59:34 +0000 (15:59 -0700)]
mm, /proc/pid/pagemap: fix soft dirty marking for PMD migration entry

When the pagetable is walked in the implementation of /proc/<pid>/pagemap,
pmd_soft_dirty() is used for both the PMD huge page map and the PMD
migration entries.  That is wrong, pmd_swp_soft_dirty() should be used
for the PMD migration entries instead because the different page table
entry flag is used.

As a result, /proc/pid/pagemap may report incorrect soft dirty information
for PMD migration entries.

Link: http://lkml.kernel.org/r/20171017081818.31795-1-ying.huang@intel.com
Fixes: 84c3fc4e9c56 ("mm: thp: check pmd migration entry in common path")
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: "Jérôme Glisse" <jglisse@redhat.com>
Cc: Daniel Colascione <dancol@google.com>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agouserfaultfd: hugetlbfs: prevent UFFDIO_COPY to fill beyond the end of i_size
Andrea Arcangeli [Thu, 2 Nov 2017 22:59:29 +0000 (15:59 -0700)]
userfaultfd: hugetlbfs: prevent UFFDIO_COPY to fill beyond the end of i_size

This oops:

  kernel BUG at fs/hugetlbfs/inode.c:484!
  RIP: remove_inode_hugepages+0x3d0/0x410
  Call Trace:
    hugetlbfs_setattr+0xd9/0x130
    notify_change+0x292/0x410
    do_truncate+0x65/0xa0
    do_sys_ftruncate.constprop.3+0x11a/0x180
    SyS_ftruncate+0xe/0x10
    tracesys+0xd9/0xde

was caused by the lack of i_size check in hugetlb_mcopy_atomic_pte.

mmap() can still succeed beyond the end of the i_size after vmtruncate
zapped vmas in those ranges, but the faults must not succeed, and that
includes UFFDIO_COPY.

We could differentiate the retval to userland to represent a SIGBUS like
a page fault would do (vs SIGSEGV), but it doesn't seem very useful and
we'd need to pick a random retval as there's no meaningful syscall
retval that would differentiate from SIGSEGV and SIGBUS, there's just
-EFAULT.

Link: http://lkml.kernel.org/r/20171016223914.2421-2-aarcange@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agoMerge branch 'net-mini_Qdisc'
David S. Miller [Fri, 3 Nov 2017 12:57:35 +0000 (21:57 +0900)]
Merge branch 'net-mini_Qdisc'

Jiri Pirko says:

====================
net: core: introduce mini_Qdisc and eliminate usage of tp->q for clsact fastpath

This patchset's main patch is patch number 2. It carries the
description. Patch 1 is just a dependency.

---
v3->v4:
- rebased to be applicable on top of the current net-next
v2->v3:
- Using head change callback to replace miniq pointer every time tp head
  changes. This eliminates one rcu dereference and makes the claim "without
  added overhead" valid.
v1->v2:
- Use dev instead of skb->dev in sch_handle_egress as pointed out by Daniel
- Fixed synchronize_rcu_bh() in mini_qdisc_disable and commented
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: core: introduce mini_Qdisc and eliminate usage of tp->q for clsact fastpath
Jiri Pirko [Fri, 3 Nov 2017 10:46:25 +0000 (11:46 +0100)]
net: core: introduce mini_Qdisc and eliminate usage of tp->q for clsact fastpath

In sch_handle_egress and sch_handle_ingress tp->q is used only in order
to update stats. So stats and filter list are the only things that are
needed in clsact qdisc fastpath processing. Introduce new mini_Qdisc
struct to hold those items. Also, introduce a helper to swap the
mini_Qdisc structures in case filter list head changes.

This removes need for tp->q usage without added overhead.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: sched: introduce chain_head_change callback
Jiri Pirko [Fri, 3 Nov 2017 10:46:24 +0000 (11:46 +0100)]
net: sched: introduce chain_head_change callback

Add a callback that is to be called whenever head of the chain changes.
Also provide a callback for the default case when the caller gets a
block using non-extended getter.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoMerge branch 'hns3-ethtool-ksettings'
David S. Miller [Fri, 3 Nov 2017 12:36:50 +0000 (21:36 +0900)]
Merge branch 'hns3-ethtool-ksettings'

Lipeng says:

====================
net: hns3: support set_link_ksettings and for nway_reset ethtool command

This patch-set adds support for set_link_ksettings && for nway_resets
ethtool command and fixes some related ethtool bugs.
1, patch[4/6] adds support for ethtool_ops.set_link_ksettings.
2, patch[5/6] adds support ethtool_ops.for nway_reset.
3, patch[1/6,2/6,3/6,6/6] fix some bugs for getting port information by
   ethtool command(ethtool ethx).
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: hns3: fix a bug for phy supported feature initialization
Fuyun Liang [Fri, 3 Nov 2017 04:18:30 +0000 (12:18 +0800)]
net: hns3: fix a bug for phy supported feature initialization

This patch fixes a bug for phy supported feature initialization.
Currently, the value of phydev->supported is initialized by kernel.
So it includes many features that we do not support, such as
SUPPORTED_FIBRE and SUPPORTED_BNC. This patch fixes it.

Fixes: 256727d (net: hns3: Add MDIO support to HNS3 Ethernet driver for hip08 SoC)
Signed-off-by: Fuyun Liang <liangfuyun1@huawei.com>
Signed-off-by: Lipeng <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: hns3: add support for nway_reset
Fuyun Liang [Fri, 3 Nov 2017 04:18:29 +0000 (12:18 +0800)]
net: hns3: add support for nway_reset

This patch adds nway_reset support for ethtool cmd.

Signed-off-by: Fuyun Liang <liangfuyun1@huawei.com>
Signed-off-by: Lipeng <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: hns3: add support for set_link_ksettings
Fuyun Liang [Fri, 3 Nov 2017 04:18:28 +0000 (12:18 +0800)]
net: hns3: add support for set_link_ksettings

This patch adds set_link_ksettings support for ethtool cmd.

Signed-off-by: Fuyun Liang <liangfuyun1@huawei.com>
Signed-off-by: Lipeng <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: hns3: fix a bug in hns3_driv_to_eth_caps
Fuyun Liang [Fri, 3 Nov 2017 04:18:27 +0000 (12:18 +0800)]
net: hns3: fix a bug in hns3_driv_to_eth_caps

The value of link_modes.advertising and the value of link_modes.supported
is initialized to zero every time in for loop in hns3_driv_to_eth_caps().
But we just want to set specified bit for them. Initialization is
unnecessary. This patch fixes it.

Fixes: 496d03e (net: hns3: Add Ethtool support to HNS3 driver)
Signed-off-by: Fuyun Liang <liangfuyun1@huawei.com>
Signed-off-by: Lipeng <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: hns3: fix for getting advertised_caps in hns3_get_link_ksettings
Fuyun Liang [Fri, 3 Nov 2017 04:18:26 +0000 (12:18 +0800)]
net: hns3: fix for getting advertised_caps in hns3_get_link_ksettings

This patch fixes a bug for ethtool's get_link_ksettings().
The advertising for autoneg is always added to advertised_caps
whether autoneg is enable or disable. This patch fixes it.

Fixes: 496d03e (net: hns3: Add Ethtool support to HNS3 driver)
Signed-off-by: Fuyun Liang <liangfuyun1@huawei.com>
Signed-off-by: Lipeng <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: hns3: fix for getting autoneg in hns3_get_link_ksettings
Fuyun Liang [Fri, 3 Nov 2017 04:18:25 +0000 (12:18 +0800)]
net: hns3: fix for getting autoneg in hns3_get_link_ksettings

This patch fixes a bug for ethtool's get_link_ksettings().
When phy exists, we should get autoneg from phy rather than from mac.
Because the value of mac.autoneg is invalid when phy exists.

Fixes: 496d03e (net: hns3: Add Ethtool support to HNS3 driver)
Signed-off-by: Fuyun Liang <liangfuyun1@huawei.com>
Signed-off-by: Lipeng <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoMerge branch 'bnxt_en-next'
David S. Miller [Fri, 3 Nov 2017 12:33:31 +0000 (21:33 +0900)]
Merge branch 'bnxt_en-next'

Michael Chan says:

====================
bnxt_en: Fix IRQ coalescing regressions.

There was a typo and missing guard-rail against illegal values in the
recent code clean up.  All reported by Andy Gospodarek.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agobnxt_en: Fix IRQ coalescing regression.
Michael Chan [Fri, 3 Nov 2017 07:32:39 +0000 (03:32 -0400)]
bnxt_en: Fix IRQ coalescing regression.

Recent IRQ coalescing clean up has removed a guard-rail for the max DMA
buffer coalescing value.  This is a 6-bit value and must not be 0.  We
already have a check for 0 but 64 is equivalent to 0 and will cause
non-stop interrupts.  Fix it by adding the proper check.

Fixes: f8503969d27b ("bnxt_en: Refactor and simplify coalescing code.")
Reported-by: Andy Gospodarek <gospo@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agobnxt_en: fix typo in bnxt_set_coalesce
Andy Gospodarek [Fri, 3 Nov 2017 07:32:38 +0000 (03:32 -0400)]
bnxt_en: fix typo in bnxt_set_coalesce

Recent refactoring of coalesce settings contained a typo that prevents
receive settings from being set properly.

Fixes: 18775aa8a91f ("bnxt_en: Reorganize the coalescing parameters.")
Signed-off-by: Andy Gospodarek <gospo@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet_sched: check NULL in tcf_block_put()
Cong Wang [Fri, 3 Nov 2017 00:32:08 +0000 (17:32 -0700)]
net_sched: check NULL in tcf_block_put()

Callers of tcf_block_put() could pass NULL so
we can't use block->q before checking if block is
NULL or not.

tcf_block_put_ext() callers are fine, it is always
non-NULL.

Fixes: 8c4083b30e56 ("net: sched: add block bind/unbind notif. and extended block_get/put")
Reported-by: Dave Taht <dave.taht@gmail.com>
Cc: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agotcp: tcp_fragment() should not assume rtx skbs
Eric Dumazet [Fri, 3 Nov 2017 01:10:03 +0000 (18:10 -0700)]
tcp: tcp_fragment() should not assume rtx skbs

While stress testing MTU probing, we had crashes in list_del() that we root-caused
to the fact that tcp_fragment() is unconditionally inserting the freshly allocated
skb into tsorted_sent_queue list.

But this list is supposed to contain skbs that were sent.
This was mostly harmless until MTU probing was enabled.

Fortunately we can use the tcp_queue enum added later (but in same linux version)
for rtx-rb-tree to fix the bug.

Fixes: e2080072ed2d ("tcp: new list for sent but unacked skbs for RACK recovery")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Soheil Hassas Yeganeh <soheil@google.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Priyaranjan Jha <priyarjha@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agomISDN: hfcpci: Convert timers to use timer_setup()
Kees Cook [Thu, 2 Nov 2017 23:18:07 +0000 (16:18 -0700)]
mISDN: hfcpci: Convert timers to use timer_setup()

In preparation for unconditionally passing the struct timer_list pointer to
all timer callbacks, switch to using the new timer_setup() and from_timer()
to pass the timer pointer explicitly.

Cc: Karsten Keil <isdn@linux-pingi.de>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Arvind Yadav <arvind.yadav.cs@gmail.com>
Cc: Geliang Tang <geliangtang@gmail.com>
Cc: netdev@vger.kernel.org
Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoISDN: eicon: message: mark expected switch fall-throughs
Gustavo A. R. Silva [Thu, 2 Nov 2017 21:14:14 +0000 (16:14 -0500)]
ISDN: eicon: message: mark expected switch fall-throughs

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

Addresses-Coverity-ID: 114780
Addresses-Coverity-ID: 114781
Addresses-Coverity-ID: 114782
Addresses-Coverity-ID: 114783
Addresses-Coverity-ID: 114784
Addresses-Coverity-ID: 114785
Addresses-Coverity-ID: 114786
Addresses-Coverity-ID: 114787
Addresses-Coverity-ID: 114788
Addresses-Coverity-ID: 114789
Addresses-Coverity-ID: 114790
Addresses-Coverity-ID: 114791
Addresses-Coverity-ID: 114792
Addresses-Coverity-ID: 114793
Addresses-Coverity-ID: 114794
Addresses-Coverity-ID: 114795
Addresses-Coverity-ID: 200521
Signed-off-by: Gustavo A. R. Silva <garsilva@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: sched: cls_bpf: use bitwise & rather than logical && on gen_flags
Colin Ian King [Thu, 2 Nov 2017 20:04:12 +0000 (20:04 +0000)]
net: sched: cls_bpf: use bitwise & rather than logical && on gen_flags

Currently gen_flags is being operated on by a logical && operator rather
than a bitwise & operator. This looks incorrect as these should be bit
flag operations. Fix this.

Detected by CoverityScan, CID#1460305 ("Logical vs. bitwise operator")

Fixes: 3f7889c4c79b ("net: sched: cls_bpf: call block callbacks for offload)
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agotcp: fix a lockdep issue in tcp_fastopen_reset_cipher()
Eric Dumazet [Thu, 2 Nov 2017 18:53:04 +0000 (11:53 -0700)]
tcp: fix a lockdep issue in tcp_fastopen_reset_cipher()

icsk_accept_queue.fastopenq.lock is only fully initialized at listen()
time.

LOCKDEP is not happy if we attempt a spin_lock_bh() on it, because
of missing annotation. (Although kernel runs just fine)

Lets use net->ipv4.tcp_fastopen_ctx_lock to protect ctx access.

Fixes: 1fba70e5b6be ("tcp: socket option to set TCP fast open key")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Christoph Paasch <cpaasch@apple.com>
Reviewed-by: Christoph Paasch <cpaasch@apple.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agobpf: fix verifier NULL pointer dereference
Craig Gallek [Thu, 2 Nov 2017 15:18:01 +0000 (11:18 -0400)]
bpf: fix verifier NULL pointer dereference

do_check() can fail early without allocating env->cur_state under
memory pressure.  Syzkaller found the stack below on the linux-next
tree because of this.

  kasan: CONFIG_KASAN_INLINE enabled
  kasan: GPF could be caused by NULL-ptr deref or user memory access
  general protection fault: 0000 [#1] SMP KASAN
  Dumping ftrace buffer:
     (ftrace buffer empty)
  Modules linked in:
  CPU: 1 PID: 27062 Comm: syz-executor5 Not tainted 4.14.0-rc7+ #106
  Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
  task: ffff8801c2c74700 task.stack: ffff8801c3e28000
  RIP: 0010:free_verifier_state kernel/bpf/verifier.c:347 [inline]
  RIP: 0010:bpf_check+0xcf4/0x19c0 kernel/bpf/verifier.c:4533
  RSP: 0018:ffff8801c3e2f5c8 EFLAGS: 00010202
  RAX: dffffc0000000000 RBX: 00000000fffffff4 RCX: 0000000000000000
  RDX: 0000000000000070 RSI: ffffffff817d5aa9 RDI: 0000000000000380
  RBP: ffff8801c3e2f668 R08: 0000000000000000 R09: 1ffff100387c5d9f
  R10: 00000000218c4e80 R11: ffffffff85b34380 R12: ffff8801c4dc6a28
  R13: 0000000000000000 R14: ffff8801c4dc6a00 R15: ffff8801c4dc6a20
  FS:  00007f311079b700(0000) GS:ffff8801db300000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: 00000000004d4a24 CR3: 00000001cbcd0000 CR4: 00000000001406e0
  DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
  DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
  Call Trace:
   bpf_prog_load+0xcbb/0x18e0 kernel/bpf/syscall.c:1166
   SYSC_bpf kernel/bpf/syscall.c:1690 [inline]
   SyS_bpf+0xae9/0x4620 kernel/bpf/syscall.c:1652
   entry_SYSCALL_64_fastpath+0x1f/0xbe
  RIP: 0033:0x452869
  RSP: 002b:00007f311079abe8 EFLAGS: 00000212 ORIG_RAX: 0000000000000141
  RAX: ffffffffffffffda RBX: 0000000000758020 RCX: 0000000000452869
  RDX: 0000000000000030 RSI: 0000000020168000 RDI: 0000000000000005
  RBP: 00007f311079aa20 R08: 0000000000000000 R09: 0000000000000000
  R10: 0000000000000000 R11: 0000000000000212 R12: 00000000004b7550
  R13: 00007f311079ab58 R14: 00000000004b7560 R15: 0000000000000000
  Code: df 48 c1 ea 03 80 3c 02 00 0f 85 e6 0b 00 00 4d 8b 6e 20 48 b8 00 00 00 00 00 fc ff df 49 8d bd 80 03 00 00 48 89 fa 48 c1 ea 03 <80> 3c 02 00 0f 85 b6 0b 00 00 49 8b bd 80 03 00 00 e8 d6 0c 26
  RIP: free_verifier_state kernel/bpf/verifier.c:347 [inline] RSP: ffff8801c3e2f5c8
  RIP: bpf_check+0xcf4/0x19c0 kernel/bpf/verifier.c:4533 RSP: ffff8801c3e2f5c8
  ---[ end trace c8d37f339dc64004 ]---

Fixes: 638f5b90d460 ("bpf: reduce verifier memory consumption")
Fixes: 1969db47f8d0 ("bpf: fix verifier memory leaks")
Signed-off-by: Craig Gallek <kraig@google.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agotipc: eliminate unnecessary probing
Jon Maloy [Thu, 2 Nov 2017 14:44:15 +0000 (15:44 +0100)]
tipc: eliminate unnecessary probing

The neighbor monitor employs a threshold, default set to 32 peer nodes,
where it activates the "Overlapping Neighbor Monitoring" algorithm.
Below that threshold, monitoring is full-mesh, and no "domain records"
are passed between the nodes.

Because of this, a node never received a peer's ack that it has received
the most recent update of the own domain. Hence, the field 'acked_gen'
in struct tipc_monitor_state remains permamently at zero, whereas the
own domain generation is incremented for each added or removed peer.

This has the effect that the function tipc_mon_get_state() always sets
the field 'probing' in struct tipc_monitor_state true, again leading the
tipc_link_timeout() of the link in question to always send out a probe,
even when link->silent_intv_count is zero.

This is functionally harmless, but leads to some unncessary probing,
which can easily be eliminated by setting the 'probing' field of the
said struct correctly in such cases.

At the same time, we explictly invalidate the sent domain records when
the algorithm is not activated. This will eliminate any risk that an
invalid domain record might be inadverently accepted by the peer.

Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: sched: move block offload unbind after all chains are flushed
Jiri Pirko [Thu, 2 Nov 2017 14:07:01 +0000 (15:07 +0100)]
net: sched: move block offload unbind after all chains are flushed

Currently, the offload unbind is done before the chains are flushed.
That causes driver to unregister block callback before it can get all
the callback calls done during flush, leaving the offloaded tps inside
the HW. So fix the order to prevent this situation and restore the
original behaviour.

Reported-by: Alexander Duyck <alexander.duyck@gmail.com>
Reported-by: Jakub Kicinski <kubakici@wp.pl>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agocxgb4vf: define get_fecparam ethtool callback
Ganesh Goudar [Thu, 2 Nov 2017 13:58:20 +0000 (19:28 +0530)]
cxgb4vf: define get_fecparam ethtool callback

Add support to new ethtool operation get_fecparam to
fetch FEC parameters.

Original Work by: Casey Leedom <leedom@chelsio.com>
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agocxgb4: add new T6 pci device id's
Ganesh Goudar [Thu, 2 Nov 2017 13:56:58 +0000 (19:26 +0530)]
cxgb4: add new T6 pci device id's

Add 0x6086 T6 device id.

Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet/ncsi: Make local function ncsi_get_filter() static
Wei Yongjun [Thu, 2 Nov 2017 11:15:28 +0000 (11:15 +0000)]
net/ncsi: Make local function ncsi_get_filter() static

Fixes the following sparse warnings:

net/ncsi/ncsi-manage.c:41:5: warning:
 symbol 'ncsi_get_filter' was not declared. Should it be static?

Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: bridge: Convert timers to use timer_setup()
Allen Pais [Fri, 3 Nov 2017 06:21:11 +0000 (11:51 +0530)]
net: bridge: Convert timers to use timer_setup()

switch to using the new timer_setup() and from_timer() api's.

Signed-off-by: Allen Pais <allen.pais@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: bridge: Convert timers to use timer_setup()
Allen Pais [Fri, 3 Nov 2017 06:21:10 +0000 (11:51 +0530)]
net: bridge: Convert timers to use timer_setup()

switch to using the new timer_setup() and from_timer() api's.

Signed-off-by: Allen Pais <allen.pais@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoMerge branch 'mlxsw-Align-multipath-hash-parameters-with-kernels'
David S. Miller [Fri, 3 Nov 2017 06:40:42 +0000 (15:40 +0900)]
Merge branch 'mlxsw-Align-multipath-hash-parameters-with-kernels'

Jiri Pirko says:

====================
mlxsw: Align multipath hash parameters with kernel's

Ido says:

This set makes sure the device is using the same parameters as the
kernel when it computes the multipath hash during IP forwarding.

First patch adds a new netevent to let interested listeners know that
the multipath hash policy has changed.

Next two patches do small and non-functional changes in the mlxsw
driver.

Last patches configure the multipath hash policy upon driver
initialization and as a response to netevents.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agomlxsw: spectrum_router: Update multipath hash parameters upon netevents
Ido Schimmel [Thu, 2 Nov 2017 16:14:10 +0000 (17:14 +0100)]
mlxsw: spectrum_router: Update multipath hash parameters upon netevents

Make sure the device and the kernel are performing the multipath hash
according to the same parameters by updating the device whenever the
relevant netevent is generated.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reviewed-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agomlxsw: spectrum_router: Align multipath hash parameters with kernel's
Ido Schimmel [Thu, 2 Nov 2017 16:14:09 +0000 (17:14 +0100)]
mlxsw: spectrum_router: Align multipath hash parameters with kernel's

Up until now we used the hardware's defaults for multipath hash
computation. This patch aligns the hardware's multipath parameters with
the kernel's.

For IPv4 packets, the parameters are determined according to the
'fib_multipath_hash_policy' sysctl during module initialization. In case
L3-mode is requested, only the source and destination IP addresses are
used. There is no special handling of ICMP error packets.

In case L4-mode is requested, a 5-tuple is used: source and destination
IP addresses, source and destination ports and IP protocol. Note that
the layer 4 fields are not considered for fragmented packets.

For IPv6 packets, the source and destination IP addresses are used, as
well as the flow label and the next header fields.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reviewed-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agomlxsw: reg: Add Router ECMP Configuration Register Version 2
Ido Schimmel [Thu, 2 Nov 2017 16:14:08 +0000 (17:14 +0100)]
mlxsw: reg: Add Router ECMP Configuration Register Version 2

The RECRv2 register is used for setting up the router's ECMP hash
configuration.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reviewed-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agomlxsw: spectrum_router: Properly name netevent work struct
Ido Schimmel [Thu, 2 Nov 2017 16:14:07 +0000 (17:14 +0100)]
mlxsw: spectrum_router: Properly name netevent work struct

The struct containing the work item queued from the netevent handler is
named after the only event it is currently used for, which is neighbour
updates.

Use a more appropriate name for the struct, as we are going to use it
for more events.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reviewed-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agomlxsw: spectrum_router: Embed netevent notifier block in router struct
Ido Schimmel [Thu, 2 Nov 2017 16:14:06 +0000 (17:14 +0100)]
mlxsw: spectrum_router: Embed netevent notifier block in router struct

We are going to need to respond to netevents notifying us about
multipath hash updates by configuring the device's hash parameters.

Embed the netevent notifier in the router struct so that we could
retrieve it upon notifications and use it to configure the device.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reviewed-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoipv4: Send a netevent whenever multipath hash policy is changed
Ido Schimmel [Thu, 2 Nov 2017 16:14:05 +0000 (17:14 +0100)]
ipv4: Send a netevent whenever multipath hash policy is changed

Devices performing IPv4 forwarding need to update their multipath hash
policy whenever it is changed.

Inform these devices by generating a netevent.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reviewed-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Acked-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: systemport: Correct IPG length settings
Florian Fainelli [Thu, 2 Nov 2017 23:08:40 +0000 (16:08 -0700)]
net: systemport: Correct IPG length settings

Due to a documentation mistake, the IPG length was set to 0x12 while it
should have been 12 (decimal). This would affect short packet (64B
typically) performance since the IPG was bigger than necessary.

Fixes: 44a4524c54af ("net: systemport: Add support for SYSTEMPORT Lite")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agotcp: do not mangle skb->cb[] in tcp_make_synack()
Eric Dumazet [Thu, 2 Nov 2017 19:30:25 +0000 (12:30 -0700)]
tcp: do not mangle skb->cb[] in tcp_make_synack()

Christoph Paasch sent a patch to address the following issue :

tcp_make_synack() is leaving some TCP private info in skb->cb[],
then send the packet by other means than tcp_transmit_skb()

tcp_transmit_skb() makes sure to clear skb->cb[] to not confuse
IPv4/IPV6 stacks, but we have no such cleanup for SYNACK.

tcp_make_synack() should not use tcp_init_nondata_skb() :

tcp_init_nondata_skb() really should be limited to skbs put in write/rtx
queues (the ones that are only sent via tcp_transmit_skb())

This patch fixes the issue and should even save few cpu cycles ;)

Fixes: 971f10eca186 ("tcp: better TCP_SKB_CB layout to reduce cache line misses")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Christoph Paasch <cpaasch@apple.com>
Reviewed-by: Christoph Paasch <cpaasch@apple.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agofib: fib_dump_info can no longer use __in_dev_get_rtnl
Florian Westphal [Thu, 2 Nov 2017 15:02:20 +0000 (16:02 +0100)]
fib: fib_dump_info can no longer use __in_dev_get_rtnl

syzbot reported yet another regression added with DOIT_UNLOCKED.
When nexthop is marked as dead, fib_dump_info uses __in_dev_get_rtnl():

./include/linux/inetdevice.h:230 suspicious rcu_dereference_protected() usage!
rcu_scheduler_active = 2, debug_locks = 1
1 lock held by syz-executor2/23859:
 #0:  (rcu_read_lock){....}, at: [<ffffffff840283f0>]
inet_rtm_getroute+0xaa0/0x2d70 net/ipv4/route.c:2738
[..]
  lockdep_rcu_suspicious+0x123/0x170 kernel/locking/lockdep.c:4665
  __in_dev_get_rtnl include/linux/inetdevice.h:230 [inline]
  fib_dump_info+0x1136/0x13d0 net/ipv4/fib_semantics.c:1377
  inet_rtm_getroute+0xf97/0x2d70 net/ipv4/route.c:2785
..

This isn't safe anymore, callers either hold RTNL mutex or rcu read lock,
so these spots must use rcu_dereference_rtnl() or plain rcu_derefence()
(plus unconditional rcu read lock).

This does the latter.

Fixes: 394f51abb3d04f ("ipv4: route: set ipv4 RTM_GETROUTE to not use rtnl")
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agocxgb4: fix error return code in cxgb4_set_hash_filter()
Wei Yongjun [Thu, 2 Nov 2017 11:15:07 +0000 (11:15 +0000)]
cxgb4: fix error return code in cxgb4_set_hash_filter()

Fix to return a negative error code from thecxgb4_alloc_atid()
error handling case instead of 0.

Fixes: 12b276fbf6e0 ("cxgb4: add support to create hash filters")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Acked-By: Kumar Sanghvi <kumaras@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agobpf: fix out-of-bounds access warning in bpf_check
Arnd Bergmann [Thu, 2 Nov 2017 11:05:52 +0000 (12:05 +0100)]
bpf: fix out-of-bounds access warning in bpf_check

The bpf_verifer_ops array is generated dynamically and may be
empty depending on configuration, which then causes an out
of bounds access:

kernel/bpf/verifier.c: In function 'bpf_check':
kernel/bpf/verifier.c:4320:29: error: array subscript is above array bounds [-Werror=array-bounds]

This adds a check to the start of the function as a workaround.
I would assume that the function is never called in that configuration,
so the warning is probably harmless.

Fixes: 00176a34d9e2 ("bpf: remove the verifier ops from program structure")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agobpf: fix link error without CONFIG_NET
Arnd Bergmann [Thu, 2 Nov 2017 11:05:51 +0000 (12:05 +0100)]
bpf: fix link error without CONFIG_NET

I ran into this link error with the latest net-next plus linux-next
trees when networking is disabled:

kernel/bpf/verifier.o:(.rodata+0x2958): undefined reference to `tc_cls_act_analyzer_ops'
kernel/bpf/verifier.o:(.rodata+0x2970): undefined reference to `xdp_analyzer_ops'

It seems that the code was written to deal with varying contents of
the arrray, but the actual #ifdef was missing. Both tc_cls_act_analyzer_ops
and xdp_analyzer_ops are defined in the core networking code, so adding
a check for CONFIG_NET seems appropriate here, and I've verified this with
many randconfig builds

Fixes: 4f9218aaf8a4 ("bpf: move knowledge about post-translation offsets out of verifier")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: Define eth_stp_addr in linux/etherdevice.h
Egil Hjelmeland [Thu, 2 Nov 2017 09:36:48 +0000 (10:36 +0100)]
net: Define eth_stp_addr in linux/etherdevice.h

The lan9303 driver defines eth_stp_addr as a synonym to
eth_reserved_addr_base to get the STP ethernet address 01:80:c2:00:00:00.

eth_reserved_addr_base is also used to define the start of Bridge Reserved
ethernet address range, which happen to be the STP address.

br_dev_setup refer to eth_reserved_addr_base as a definition of STP
address.

Clean up by:
 - Move the eth_stp_addr definition to linux/etherdevice.h
 - Use eth_stp_addr instead of eth_reserved_addr_base in br_dev_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>
7 years agostmmac: use of_property_read_u32 instead of read_u8
Bhadram Varka [Thu, 2 Nov 2017 07:22:13 +0000 (12:52 +0530)]
stmmac: use of_property_read_u32 instead of read_u8

Numbers in DT are stored in “cells” which are 32-bits
in size. of_property_read_u8 does not work properly
because of endianness problem.

This causes it to always return 0 with little-endian
architectures.

Fix it by using of_property_read_u32() OF API.

Signed-off-by: Bhadram Varka <vbhadram@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoliquidio: bump up driver version to 1.7.0 to match newer NIC firmware
Felix Manlunas [Thu, 2 Nov 2017 01:14:49 +0000 (18:14 -0700)]
liquidio: bump up driver version to 1.7.0 to match newer NIC firmware

Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Acked-by: Derek Chickles <derek.chickles@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoMerge branch 'net-sched-use-after-free'
David S. Miller [Fri, 3 Nov 2017 01:30:39 +0000 (10:30 +0900)]
Merge branch 'net-sched-use-after-free'

Cong Wang says:

====================
net_sched: fix a use-after-free for tc actions

This patchset fixes a use-after-free reported by Lucas
and closes potential races too.

Please see each patch for details.
====================

Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet_sched: hold netns refcnt for each action
Cong Wang [Wed, 1 Nov 2017 17:23:50 +0000 (10:23 -0700)]
net_sched: hold netns refcnt for each action

TC actions have been destroyed asynchronously for a long time,
previously in a RCU callback and now in a workqueue. If we
don't hold a refcnt for its netns, we could use the per netns
data structure, struct tcf_idrinfo, after it has been freed by
netns workqueue.

Hold refcnt to ensure netns destroy happens after all actions
are gone.

Fixes: ddf97ccdd7cb ("net_sched: add network namespace support for tc actions")
Reported-by: Lucas Bates <lucasb@mojatatu.com>
Tested-by: Lucas Bates <lucasb@mojatatu.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet_sched: acquire RTNL in tc_action_net_exit()
Cong Wang [Wed, 1 Nov 2017 17:23:49 +0000 (10:23 -0700)]
net_sched: acquire RTNL in tc_action_net_exit()

I forgot to acquire RTNL in tc_action_net_exit()
which leads that action ops->cleanup() is not always
called with RTNL. This usually is not a big deal because
this function is called after all netns refcnt are gone,
but given RTNL protects more than just actions, add it
for safety and consistency.

Also add an assertion to catch other potential bugs.

Fixes: ddf97ccdd7cb ("net_sched: add network namespace support for tc actions")
Reported-by: Lucas Bates <lucasb@mojatatu.com>
Tested-by: Lucas Bates <lucasb@mojatatu.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agotcp: add tracepoint trace_tcp_retransmit_synack()
Song Liu [Mon, 30 Oct 2017 21:41:35 +0000 (14:41 -0700)]
tcp: add tracepoint trace_tcp_retransmit_synack()

This tracepoint can be used to trace synack retransmits. It maintains
pointer to struct request_sock.

We cannot simply reuse trace_tcp_retransmit_skb() here, because the
sk here is the LISTEN socket. The IP addresses and ports should be
extracted from struct request_sock.

Note that, like many other tracepoints, this patch uses IS_ENABLED
in TP_fast_assign macro, which triggers sparse warning like:

./include/trace/events/tcp.h:274:1: error: directive in argument list
./include/trace/events/tcp.h:281:1: error: directive in argument list

However, there is no good solution to avoid these warnings. To the
best of our knowledge, these warnings are harmless.

Signed-off-by: Song Liu <songliubraving@fb.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoipv6: Implement limits on Hop-by-Hop and Destination options
Tom Herbert [Mon, 30 Oct 2017 21:16:00 +0000 (14:16 -0700)]
ipv6: Implement limits on Hop-by-Hop and Destination options

RFC 8200 (IPv6) defines Hop-by-Hop options and Destination options
extension headers. Both of these carry a list of TLVs which is
only limited by the maximum length of the extension header (2048
bytes). By the spec a host must process all the TLVs in these
options, however these could be used as a fairly obvious
denial of service attack. I think this could in fact be
a significant DOS vector on the Internet, one mitigating
factor might be that many FWs drop all packets with EH (and
obviously this is only IPv6) so an Internet wide attack might not
be so effective (yet!).

By my calculation, the worse case packet with TLVs in a standard
1500 byte MTU packet that would be processed by the stack contains
1282 invidual TLVs (including pad TLVS) or 724 two byte TLVs. I
wrote a quick test program that floods a whole bunch of these
packets to a host and sure enough there is substantial time spent
in ip6_parse_tlv. These packets contain nothing but unknown TLVS
(that are ignored), TLV padding, and bogus UDP header with zero
payload length.

  25.38%  [kernel]                    [k] __fib6_clean_all
  21.63%  [kernel]                    [k] ip6_parse_tlv
   4.21%  [kernel]                    [k] __local_bh_enable_ip
   2.18%  [kernel]                    [k] ip6_pol_route.isra.39
   1.98%  [kernel]                    [k] fib6_walk_continue
   1.88%  [kernel]                    [k] _raw_write_lock_bh
   1.65%  [kernel]                    [k] dst_release

This patch adds configurable limits to Destination and Hop-by-Hop
options. There are three limits that may be set:
  - Limit the number of options in a Hop-by-Hop or Destination options
    extension header.
  - Limit the byte length of a Hop-by-Hop or Destination options
    extension header.
  - Disallow unrecognized options in a Hop-by-Hop or Destination
    options extension header.

The limits are set in corresponding sysctls:

  ipv6.sysctl.max_dst_opts_cnt
  ipv6.sysctl.max_hbh_opts_cnt
  ipv6.sysctl.max_dst_opts_len
  ipv6.sysctl.max_hbh_opts_len

If a max_*_opts_cnt is less than zero then unknown TLVs are disallowed.
The number of known TLVs that are allowed is the absolute value of
this number.

If a limit is exceeded when processing an extension header the packet is
dropped.

Default values are set to 8 for options counts, and set to INT_MAX
for maximum length. Note the choice to limit options to 8 is an
arbitrary guess (roughly based on the fact that the stack supports
three HBH options and just one destination option).

These limits have being proposed in draft-ietf-6man-rfc6434-bis.

Tested (by Martin Lau)

I tested out 1 thread (i.e. one raw_udp process).

I changed the net.ipv6.max_dst_(opts|hbh)_number between 8 to 2048.
With sysctls setting to 2048, the softirq% is packed to 100%.
With 8, the softirq% is almost unnoticable from mpstat.

v2;
  - Code and documention cleanup.
  - Change references of RFC2460 to be RFC8200.
  - Add reference to RFC6434-bis where the limits will be in standard.

Signed-off-by: Tom Herbert <tom@quantonium.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agopowerpc/perf: Fix core-imc hotplug callback failure during imc initialization
Madhavan Srinivasan [Thu, 2 Nov 2017 12:12:26 +0000 (17:42 +0530)]
powerpc/perf: Fix core-imc hotplug callback failure during imc initialization

Call trace observed during boot:

  nest_capp0_imc performance monitor hardware support registered
  nest_capp1_imc performance monitor hardware support registered
  core_imc memory allocation for cpu 56 failed
  Unable to handle kernel paging request for data at address 0xffa400010
  Faulting instruction address: 0xc000000000bf3294
  0:mon> e
  cpu 0x0: Vector: 300 (Data Access) at [c000000ff38ff8d0]
      pc: c000000000bf3294: mutex_lock+0x34/0x90
      lr: c000000000bf3288: mutex_lock+0x28/0x90
      sp: c000000ff38ffb50
     msr: 9000000002009033
     dar: ffa400010
   dsisr: 80000
    current = 0xc000000ff383de00
    paca    = 0xc000000007ae0000  softe: 0  irq_happened: 0x01
      pid   = 13, comm = cpuhp/0
  Linux version 4.11.0-39.el7a.ppc64le (mockbuild@ppc-058.build.eng.bos.redhat.com) (gcc version 4.8.5 20150623 (Red Hat 4.8.5-16) (GCC) ) #1 SMP Tue Oct 3 07:42:44 EDT 2017
  0:mon> t
  [c000000ff38ffb80c0000000002ddfac perf_pmu_migrate_context+0xac/0x470
  [c000000ff38ffc40c00000000011385c ppc_core_imc_cpu_offline+0x1ac/0x1e0
  [c000000ff38ffc90c000000000125758 cpuhp_invoke_callback+0x198/0x5d0
  [c000000ff38ffd00c00000000012782c cpuhp_thread_fun+0x8c/0x3d0
  [c000000ff38ffd60c0000000001678d0 smpboot_thread_fn+0x290/0x2a0
  [c000000ff38ffdc0c00000000015ee78 kthread+0x168/0x1b0
  [c000000ff38ffe30c00000000000b368 ret_from_kernel_thread+0x5c/0x74

While registering the cpuhoplug callbacks for core-imc, if we fails
in the cpuhotplug online path for any random core (either because opal call to
initialize the core-imc counters fails or because memory allocation fails for
that core), ppc_core_imc_cpu_offline() will get invoked for other cpus who
successfully returned from cpuhotplug online path.

But in the ppc_core_imc_cpu_offline() path we are trying to migrate the event
context, when core-imc counters are not even initialized. Thus creating the
above stack dump.

Add a check to see if core-imc counters are enabled or not in the cpuhotplug
offline path before migrating the context to handle this failing scenario.

Fixes: 885dcd709ba9 ("powerpc/perf: Add nest IMC PMU support")
Signed-off-by: Madhavan Srinivasan <maddy@linux.vnet.ibm.com>
Signed-off-by: Anju T Sudhakar <anju@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
7 years agoKbuild: don't pass "-C" to preprocessor when processing linker scripts
Linus Torvalds [Thu, 2 Nov 2017 21:10:37 +0000 (14:10 -0700)]
Kbuild: don't pass "-C" to preprocessor when processing linker scripts

For some odd historical reason, we preprocessed the linker scripts with
"-C", which keeps comments around.  That makes no sense, since the
comments are not meaningful for the build anyway.

And it actually breaks things, since linker scripts can't have C++ style
"//" comments in them, so keeping comments after preprocessing now
limits us in odd and surprising ways in our header files for no good
reason.

The -C option goes back to pre-git and pre-bitkeeper times, but seems to
have been historically used (along with "-traditional") for some
odd-ball architectures (ia64, MIPS and SH).  It probably didn't matter
back then either, but might possibly have been used to minimize the
difference between the original file and the pre-processed result.

The reason for this may be lost in time, but let's not perpetuate it
only because we can't remember why we did this crazy thing.

This was triggered by the recent addition of SPDX lines to the source
tree, where people apparently were confused about why header files
couldn't use the C++ comment format.

Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Greg KH <gregkh@linuxfoundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agoRevert "x86: do not use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz""
Linus Torvalds [Thu, 2 Nov 2017 21:06:32 +0000 (14:06 -0700)]
Revert "x86: do not use cpufreq_quick_get() for /proc/cpuinfo "cpu MHz""

This reverts commit 51204e0639c49ada02fd823782ad673b6326d748.

There wasn't really any good reason for it, and people are complaining
(rightly) that it broke existing practice.

Cc: Len Brown <len.brown@intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>