brcm80211: smac: mute transmit on ops_start

Monitor mode functionality (not functional yet) requires transmit to be
muted after ops_start() is called, transmit is unmuted when the first
interface is added.

Reviewed-by: Alwin Beukers <alwin@broadcom.com>
Reviewed-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Roland Vossen <rvossen@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcm80211: smac: modified Mac80211 callback interface

Upon ops_start(), a Mac80211 driver should enable receive functionality to
support monitor mode. Also, upon ops_stop(), it should disable rx.

Driver did not follow this rule so code has been changed.

Reported-by: Johannes Berg <johannes@sipsolutions.net>
Reviewed-by: Alwin Beukers <alwin@broadcom.com>
Reviewed-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Roland Vossen <rvossen@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcm80211: smac: fixed inconsistency in transmit mute

Transmit was muted in two ways: full mute and a partial mute called
'pre ism cac time' mute. But, this 'pre ism cac time' mute was done at
one place in the code (when tx_mute == false), and overridden later
on in another place in code.

To fix this, the 'pre ism cac time' mute has been replaced by a non
mute.

Reviewed-by: Alwin Beukers <alwin@broadcom.com>
Reviewed-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Roland Vossen <rvossen@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcm80211: smac: bugfix for tx mute in brcms_b_init()

Transmit can only be muted if the mac core is enabled. When brcms_b_init()
is called, the mac core is suspended. Brcms_b_init() calls a transmit mute
function that requires an enabled mac core. This code path is never taken,
but would have been taken in subsequent patches.

Reviewed-by: Alwin Beukers <alwin@broadcom.com>
Reviewed-by: Arend van Spriel <arend@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Roland Vossen <rvossen@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcm80211: smac: removed down-on-rf-kill functionality

Softmac would bring its interface down on an RF kill switch condition,
without Mac80211 intervention. Because Mac80211 should be the only party
initiating interfaces going up and down, this functionality has been
removed.

Reviewed-by: Alwin Beukers <alwin@broadcom.com>
Reviewed-by: Arend van Spriel <arend@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Roland Vossen <rvossen@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcm80211: smac: removed down-on-watchdog MPC functionality

Softmac would bring its interface down on a certain Minimum Power Save
related condition, without Mac80211 intervention. Because Mac80211 should
be the only party initiating interfaces going up and down, this functionality
has been removed. All notions of 'MPC' have been removed in the code as
well.

Reviewed-by: Alwin Beukers <alwin@broadcom.com>
Reviewed-by: Arend van Spriel <arend@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Roland Vossen <rvossen@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcm80211: smac: removed MPC related variables

Several member variables were never read.

Reviewed-by: Alwin Beukers <alwin@broadcom.com>
Reviewed-by: Arend van Spriel <arend@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Roland Vossen <rvossen@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcm80211: smac: removed MPC related code

The chip init sequence enables MPC (Minimum Power Consumption), but the
driver disables it after that. As there are no interfaces to enable this
mode the related code is unused (member variable wlc->mpc is false).

Reviewed-by: Alwin Beukers <alwin@broadcom.com>
Reviewed-by: Arend van Spriel <arend@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Roland Vossen <rvossen@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcm80211: fmac: use brcmf_del_if for all net devices

Use brcmf_del_if for primary and virtual net device interfaces. This
is part of the net device interface clean up for fullmac.

Reviewed-by: Roland Vossen <rvossen@broadcom.com>
Reviewed-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Franky (Zhenhui) Lin <frankyl@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcm80211: fmac: allow wd timer to be disabled when bus down

Watchdog timer should be able to be stopped even firmware is not
loaded.

Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Franky (Zhenhui) Lin <frankyl@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: exit cooked monitor RX early if there are none

If there are no cooked monitor interfaces, there's
no point in building the radiotap RX header for the
frame and iterating the interface list.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcm80211: fmac: use sk_buff list for handling frames in receive path

The functions in the receive patch of the fullmac now use sk_buff
list and skb_queue_xx() functions instead of dealing with list pointers
in the sk_buff directly.

Reported-by: Johannes Berg <johannes@sipsolutions.net>
Reviewed-by: Franky (Zhenhui) Lin <frankyl@broadcom.com>
Reviewed-by: Roland Vossen <rvossen@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rtl8192cu: Add new device IDs

The latest vendor (non-mac80211) driver of 9/22/2011 shows some new
device IDs for rtl8192cu. In addition, some typos in the table are
fixed and one duplicate is removed.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcm80211: util: remove function brcmu_format_hex() from brcmutil

The function brcmu_format_hex() filled a string buffer with byte
values from a data buffer. The calling function used this string
buffer in a printk. Now the calling function uses the kernel
function print_hex_dump_bytes().

Reported-by: Johannes Berg <johannes@sipsolutions.net>
Reviewed-by: Alwin Beukers <alwin@broadcom.com>
Reviewed-by: Roland Vossen <rvossen@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcm80211: util: move brcmu_pktfrombuf() function to brcmfmac

The function brcmu_pktfrombuf was only used in the brcmfmac source
and has been moved there. It has been refactored to match its use.

Reported-by: Johannes Berg <johannes@sipsolutions.net>
Reviewed-by: Roland Vossen <rvossen@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: Franky (Zhenhui) Lin <frankyl@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcm80211: smac: remove phy api bypass in rate.h

Obviously the phy api should be used to interface with the phy. In
rate.h a table within phy was accessed directly by declaring the
table extern in rate.h itself. This patch fixes this using the
provided api function to obtain the table reference. This bypass
was found by a sparse warning on the table not being defined static.

Reviewed-by: Alwin Beukers <alwin@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: Roland Vossen <rvossen@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcm80211: smac: some local function made static in main.c

In main.c a couple of functions were not static although they
were only locally used. Sparse gave warnings on them and these
functions have been made static.

Reviewed-by: Alwin Beukers <alwin@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: Roland Vossen <rvossen@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcm80211: smac: avoid sprom endianess conversions for crc8 check

The data from the sprom consists of u16 values stored in little
endian notation over which a crc8 was determined. To validate this
the buffer needed to be converted for big-endian systems. Reading
the sprom data is now done per byte so conversion is only done
after a successful crc8 check.

Reviewed-by: Alwin Beukers <alwin@broadcom.com>
Reviewed-by: Roland Vossen <rvossen@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcm80211: smac: remove obsolete srom variables from n-phy

The n-phy requested some srom variables that are no longer needed
and consequently not present in the srom revision 8 and higher that
this driver support. This code has been removed from the n-phy.

Reviewed-by: Alwin Beukers <alwin@broadcom.com>
Reviewed-by: Roland Vossen <rvossen@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcm80211: smac: indicate severe problems to Mac80211

In case the hardware crashes, a reinitialization internal to the driver
was performed. Since Mac80211 must be in the know of such an event as
well, ieee80211_restart_hw() is now called.

Reviewed-by: Alwin Beukers <alwin@broadcom.com>
Reviewed-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Roland Vossen <rvossen@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcm80211: fmac: changed two scan related structures

struct brcmf_scan_results contained a 1 element array, but in reality
the number of scan results can be 0 or more, as indicated by the
count field in the same struct. Array has be redefined to be 0 elements
length to indicate the array is purely for reference.

Reported-by: Johannes Berg <johannes@sipsolutions.net>
Reviewed-by: Arend van Spriel <arend@broadcom.com>
Reviewed-by: Franky (Zhenhui) Lin <frankyl@broadcom.com>
Signed-off-by: Roland Vossen <rvossen@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brmc80211: fmac: reworked next_bss()

Moved function to where it is called and made it more readable.

Reviewed-by: Arend van Spriel <arend@broadcom.com>
Reviewed-by: Franky (Zhenhui) Lin <frankyl@broadcom.com>
Signed-off-by: Roland Vossen <rvossen@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcm80211: fmac: annotated little endian struct with _le

Made code more readable.

Reviewed-by: Arend van Spriel <arend@broadcom.com>
Reviewed-by: Franky (Zhenhui) Lin <frankyl@broadcom.com>
Signed-off-by: Roland Vossen <rvossen@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcm80211: smac: removed support for SROM rev < 8

Supported chips contain SROM rev 8 and upwards.

Reported-by: Hauke Mehrtens <hauke@hauke-m.de>
Reviewed-by: Alwin Beukers <alwin@broadcom.com>
Reviewed-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Roland Vossen <rvossen@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcm80211: smac: drop "40MHz intolerant" flag from HT capability info

The brcmsmac driver registered with mac80211 with HT capability info
set to 40MHz intolerant. This cause any other station on the channel
to be forced to use 20MHz. This flag has been removed.

Reported-by: Johannes Berg <johannes@sipsolutions.net>
Reviewed-by: Alwin Beukers <alwin@broadcom.com>
Reviewed-by: Roland Vossen <rvossen@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcm80211: removed duplicate defines

Removed defines from aiutils.h also present in soc.h.

Reported-by: Hauke Mehrtens <hauke@hauke-m.de>
Reviewed-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Alwin Beukers <alwin@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcm80211: cleanup defines in main.c

Signed-off-by: Alwin Beukers <alwin@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: fix confusing parentheses

There's an extra pair of parentheses here that
is simply confusing because it implies a nesting
that doesn't actually exist. Just remove it.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwmc3200wifi: add some more range checks

My previous patch added a check to get_key() but missed a couple
other places which need range checks.

The problem here is that wifi drivers have different numbers of keys.
The lower levels assume that they can have up to 4 default keys and
2 management keys but this driver only has the default keys so we
could go past the end of the ->keys[] array.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Samuel Ortiz <sameo@linux.intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rndis_wlan: add range check in del_key()

Wifi drivers can have up to 6 keys but the rndis_wlan only has 4 so
it needs to have its own checks to make sure we don't go out of
bounds.  The add_key() function already checks but I added some
checks to del_key() and set_default_key().

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k_hw: min_t() casts u32 to int

The code here treats very large values of "limit" as less than
MAX_POWER_RATE because of the cast to int.  We should do the compare
as u32 instead.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Pavel Roskin <proski@gnu.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: call set_wmm_default only for valid vifs

mac80211 calls ieee80211_set_wmm_default (which in turn
calls drv_conf_tx()) for every new interface, including
"internal" ones (e.g. monitor interface, which the low-level
driver doesn't know about).

Limit this call only to valid interfaces.

Reported-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlagn: use 6 Mbps rate for no-CCK scans

When userspace requested that a scan not be
done with CCK rates, use 6 Mbps. This is used
for example for P2P scanning.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlagn: simplify iwl_alloc_all

The iwl_alloc_all routine is only called once.  Delete the argument
and print an error in the calling routine if needed.

Signed-off-by: Don Fry <donald.h.fry@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlwifi: HW rev for 105 and 135 series

Set the HW rev. for both 105 and 135 series

Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlagn: remove unnecessary type for tracing operations

The device tracing routines only use the priv pointer as an opaque
value.  Change from a typed iwl_priv pointer to a null pointer and
eliminate the need to include iwl_priv.h.  CMD_ASYNC is defined in
iwl_shared.h which is the only reason it is included.

Signed-off-by: Don Fry <donald.h.fry@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlagn: update wowlan API

The WoWLAN API changed due to netdetect and
we now have a more generic "D3 configuration"
command that enables the sysassert & rfkill
wakeup triggers.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

dp83640: free packet queues on remove

If the PHY should disappear (for example, on an USB Ethernet MAC), then
the driver would leak any undelivered time stamp packets. This commit
fixes the issue by calling the appropriate functions to free any packets
left in the transmit and receive queues.

The driver first appeared in v3.0.

Signed-off-by: Richard Cochran <richard.cochran@omicron.at>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

dp83640: use proper function to free transmit time stamping packets

The previous commit enforces a new rule for handling the cloned packets
for transmit time stamping. These packets must not be freed using any other
function than skb_complete_tx_timestamp. This commit fixes the one and only
driver using this API.

The driver first appeared in v3.0.

Signed-off-by: Richard Cochran <richard.cochran@omicron.at>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv6: Do not use routes from locally generated RAs

When hybrid mode is enabled (accept_ra == 2), the kernel also sees RAs
generated locally. This is useful since it allows the kernel to auto-configure
its own interface addresses.

However, if 'accept_ra_defrtr' and/or 'accept_ra_rtr_pref' are set and the
locally generated RAs announce the default route and/or other route information,
the kernel happily inserts bogus routes with its own address as gateway.

With this patch, adding routes from an RA will be skiped when the RAs source
address matches any local address, just as if 'accept_ra_defrtr' and
'accept_ra_rtr_pref' were set to 0.

Signed-off-by: Andreas Hofmeister <andi@collax.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

|PATCH net-next] tg3: add tx_dropped counter

If a frame cant be transmitted, it is silently discarded.

Add a counter to report these errors to user.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: don't create multiple RX/TX rings in multi channel mode

When the HW is in multi-channel mode based on the skew/IPL, there are
4 functions per port and so not enough resources to create multiple
RX/TX rings for each function.

Signed-off-by: Suresh Reddy <suresh.reddy@emulex.com>
Signed-off-by: Sathya Perla <sathya.perla@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: don't create multiple TXQs in BE2

Multiple TXQ support is partially broken in BE2. It is fully
supported BE3 onwards and in Lancer.

Signed-off-by: Vasundhara Volam <vasundhara.volam@emulex.com>
Signed-off-by: Sathya Perla <sathya.perla@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: refactor VF setup/teardown code into be_vf_setup/clear()

Currently the code for VF setup/teardown done by a PF (if_create,
mac_add_config, link_status_query etc) is scattered; this patch
refactors this code into be_vf_setup() and be_vf_clear().  The
if_create/if_destroy/mac_addr_query cmds are now called after the MCCQ
is created; so these cmds are now modified to use the MCCQ instead of
MBOX.

Signed-off-by: Sathya Perla <sathya.perla@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: add vlan/rx-mode/flow-control config to be_setup()

When a card is reset due to EEH error recovery or due to a suspend,
rx-mode config (promisc/mc) is not being sent to the FW. be_setup() is
called in these flows and is the best place for such config/re-config
cmds. Hence include rx-mode, vlan and flow-control config in
be_setup().

Signed-off-by: Sathya Perla <sathya.perla@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net_sched: cls_flow: use skb_header_pointer()

Dan Siemon would like to add tunnelling support to cls_flow

This preliminary patch introduces use of skb_header_pointer() to help
this task, while avoiding skb head reallocation because of deep packet
inspection.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: avoid useless call of the function check_peer_pmtu

In func ipv4_dst_check,check_peer_pmtu should be called only when peer is updated.
So,if the peer is not updated in ip_rt_frag_needed,we can not inc __rt_peer_genid.

Signed-off-by: Gao feng <gaofeng@cn.fujitsu.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'master' of ra./linux/kernel/git/davem/net

TCP: remove TCP_DEBUG

It was enabled by default and the messages guarded
by the define are useful.

Signed-off-by: Flavio Leitner <fbl@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Fix driver name for mdio-gpio.c

Since commit
"7488876... dt/net: Eliminate users of of_platform_{,un}register_driver"
there are two platform drivers named "mdio-gpio" registered.
I renamed the of variant to "mdio-ofgpio".

Signed-off-by: Dirk Eibach <eibach@gdsys.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge git://git./linux/kernel/git/jkirsher/net-next

ipv4: tcp: fix TOS value in ACK messages sent from TIME_WAIT

There is a long standing bug in linux tcp stack, about ACK messages sent
on behalf of TIME_WAIT sockets.

In the IP header of the ACK message, we choose to reflect TOS field of
incoming message, and this might break some setups.

Example of things that were broken :
  - Routing using TOS as a selector
  - Firewalls
  - Trafic classification / shaping

We now remember in timewait structure the inet tos field and use it in
ACK generation, and route lookup.

Notes :
 - We still reflect incoming TOS in RST messages.
 - We could extend MuraliRaja Muniraju patch to report TOS value in
netlink messages for TIME_WAIT sockets.
 - A patch is needed for IPv6

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

rtnetlink: Add missing manual netlink notification in dev_change_net_namespaces

Renato Westphal noticed that since commit a2835763e130c343ace5320c20d33c281e7097b7
"rtnetlink: handle rtnl_link netlink notifications manually" was merged
we no longer send a netlink message when a networking device is moved
from one network namespace to another.

Fix this by adding the missing manual notification in dev_change_net_namespaces.

Since all network devices that are processed by dev_change_net_namspaces are
in the initialized state the complicated tests that guard the manual
rtmsg_ifinfo calls in rollback_registered and register_netdevice are
unnecessary and we can just perform a plain notification.

Cc: stable@kernel.org
Tested-by: Renato Westphal <renatowestphal@gmail.com>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: fix ipsec forward performance regression

There is bug in commit 5e2b61f(ipv4: Remove flowi from struct rtable).
It makes xfrm4_fill_dst() modify wrong data structure.

Signed-off-by: Zheng Yan <zheng.z.yan@intel.com>
Reported-by: Kim Phillips <kim.phillips@freescale.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

jme: fix irq storm after suspend/resume

If the device is down during suspend/resume, interrupts are enabled
without a registered interrupt handler, causing a storm of
unhandled interrupts until the IRQ is disabled because "nobody
cared".

Instead, check that the device is up before touching it in the
suspend/resume code.

Fixes https://bugzilla.kernel.org/show_bug.cgi?id=39112

Helped-by: Adrian Chadd <adrian@freebsd.org>
Helped-by: Mohammed Shafi <shafi.wireless@gmail.com>
Signed-off-by: Clemens Buchacher <drizzd@aon.at>
Signed-off-by: David S. Miller <davem@davemloft.net>

route: fix ICMP redirect validation

The commit f39925dbde7788cfb96419c0f092b086aa325c0f
(ipv4: Cache learned redirect information in inetpeer.)
removed some ICMP packet validations which are required by
RFC 1122, section 3.2.2.2:
...
  A Redirect message SHOULD be silently discarded if the new
  gateway address it specifies is not on the same connected
  (sub-) net through which the Redirect arrived [INTRO:2,
  Appendix A], or if the source of the Redirect is not the
  current first-hop gateway for the specified destination (see
  Section 3.3.1).

Signed-off-by: Flavio Leitner <fbl@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: hold sock reference while processing tx timestamps

The pair of functions,

 * skb_clone_tx_timestamp()
 * skb_complete_tx_timestamp()

were designed to allow timestamping in PHY devices. The first
function, called during the MAC driver's hard_xmit method, identifies
PTP protocol packets, clones them, and gives them to the PHY device
driver. The PHY driver may hold onto the packet and deliver it at a
later time using the second function, which adds the packet to the
socket's error queue.

As pointed out by Johannes, nothing prevents the socket from
disappearing while the cloned packet is sitting in the PHY driver
awaiting a timestamp. This patch fixes the issue by taking a reference
on the socket for each such packet. In addition, the comments
regarding the usage of these function are expanded to highlight the
rule that PHY drivers must use skb_complete_tx_timestamp() to release
the packet, in order to release the socket reference, too.

These functions first appeared in v2.6.36.

Reported-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Richard Cochran <richard.cochran@omicron.at>
Cc: <stable@vger.kernel.org>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Reviewed-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: md5: add more const attributes

Now tcp_md5_hash_header() has a const tcphdr argument, we can add more
const attributes to callers.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add ethtool -g support to virtio_net

Add support for reporting ring sizes via ethtool -g to the virtio_net
driver.

Signed-off-by: Rick Jones <rick.jones2@hp.com>
Acked-by: Rusty Russell <rusty@rustcorp.com.au>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: md5: dont write skb head in tcp_md5_hash_header()

tcp_md5_hash_header() writes into skb header a temporary zero value,
this might confuse other users of this area.

Since tcphdr is small (20 bytes), copy it in a temporary variable and
make the change in the copy.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bonding: Add a forgetten sysfs_attr_init on class_attr_bonding_masters

When I made class_attr_bonding_matters per network namespace and dynamically
allocated I overlooked the need for calling sysfs_attr_init.  Oops.

This fixes the following lockdep splat:

[    5.749651] bonding: Ethernet Channel Bonding Driver: v3.7.1 (April 27, 2011)
[    5.749655] bonding: MII link monitoring set to 100 ms
[    5.749676] BUG: key f49a831c not in .data!
[    5.749677] ------------[ cut here ]------------
[    5.749752] WARNING: at kernel/lockdep.c:2897 lockdep_init_map+0x1c3/0x460()
[    5.749809] Hardware name: ProLiant BL460c G1
[    5.749862] Modules linked in: bonding(+)
[    5.749978] Pid: 3177, comm: modprobe Not tainted 3.1.0-rc9-02177-gf2d1a4e-dirty #1157
[    5.750066] Call Trace:
[    5.750120]  [<c1352c2f>] ? printk+0x18/0x21
[    5.750176]  [<c103112d>] warn_slowpath_common+0x6d/0xa0
[    5.750231]  [<c1060133>] ? lockdep_init_map+0x1c3/0x460
[    5.750287]  [<c1060133>] ? lockdep_init_map+0x1c3/0x460
[    5.750342]  [<c103117d>] warn_slowpath_null+0x1d/0x20
[    5.750398]  [<c1060133>] lockdep_init_map+0x1c3/0x460
[    5.750453]  [<c1355ddd>] ? _raw_spin_unlock+0x1d/0x20
[    5.750510]  [<c11255c8>] ? sysfs_new_dirent+0x68/0x110
[    5.750565]  [<c1124d4b>] sysfs_add_file_mode+0x8b/0xe0
[    5.750621]  [<c1124db3>] sysfs_add_file+0x13/0x20
[    5.750675]  [<c1124e7c>] sysfs_create_file+0x1c/0x20
[    5.750737]  [<c1208f09>] class_create_file+0x19/0x20
[    5.750794]  [<c12c186f>] netdev_class_create_file+0xf/0x20
[    5.750853]  [<f85deaf4>] bond_create_sysfs+0x44/0x90 [bonding]
[    5.750911]  [<f8410947>] ? bond_create_proc_dir+0x1e/0x3e [bonding]
[    5.750970]  [<f841007e>] bond_net_init+0x7e/0x87 [bonding]
[    5.751026]  [<f8410000>] ? 0xf840ffff
[    5.751080]  [<c12abc7a>] ops_init.clone.4+0xba/0x100
[    5.751135]  [<c12abdb2>] ? register_pernet_subsys+0x12/0x30
[    5.751191]  [<c12abd03>] register_pernet_operations.clone.3+0x43/0x80
[    5.751249]  [<c12abdb9>] register_pernet_subsys+0x19/0x30
[    5.751306]  [<f84108b9>] bonding_init+0x832/0x8a2 [bonding]
[    5.751363]  [<c10011f0>] do_one_initcall+0x30/0x160
[    5.751420]  [<f8410087>] ? bond_net_init+0x87/0x87 [bonding]
[    5.751477]  [<c106d5cf>] sys_init_module+0xef/0x1890
[    5.751533]  [<c1356490>] sysenter_do_call+0x12/0x36
[    5.751588] ---[ end trace 89f492d83a7f5006 ]---

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Reported-by: Eric Dumazet <eric.dumazet@gmail.com>
Tested-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: fix tigon3_dma_hwbug_workaround()

Ari got kernel panics using tg3 NIC, and bisected to 2669069aacc9 "tg3:
enable transmit time stamping."

This is because tigon3_dma_hwbug_workaround() might alloc a new skb and
free the original. We panic when skb_tx_timestamp() is called on freed
skb.

Reported-by: Ari Savolainen <ari.m.savolainen@gmail.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

inet: add rfc 3168 extract in front of INET_ECN_encapsulate()

INET_ECN_encapsulate() is better understood if we can read the official
statement.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: use INET_ECN_MASK instead of hardcoded 3

Signed-off-by: Maciej Żenczykowski <maze@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

igb: VFTA Table Fix for i350 devices

Due to a hardware problem, writes to the VFTA register can
theoretically fail. Although the likelihood of this is very low.
This patch adds a shadow vfta in the adapter struct for reading
and adds new write functions for these devices to work around the problem.

Signed-off-by: Carolyn Wyborny <carolyn.wyborny@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

igb: Move DMA Coalescing init code to separate function.

This patch moves the DMA Coalescing feature initialization code from
igb_reset to a new function and replaces it with a call to the new
function.

Signed-off-by: Carolyn Wyborny <carolyn.wyborny@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

igb: Fix for Alt MAC Address feature on 82580 and later devices

In 82580 and later devices, the alternate MAC address feature is
completely handled by the option ROM and software does not handle
it anymore.  This patch changes the check_alt_mac_addr function to
exit immediately if device is 82580 or later.

Signed-off-by: Carolyn Wyborny <carolyn.wyborny@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

igbvf: Bump version number

Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Tested-by: Sibai Li <sibai.li@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

igbvf: Update module identification strings

Update adapter identification strings to properly indicate i350 VF devices
in the VF driver. Change the driver ID string to remove 82576-specific
wording. Update copyright date.

Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Tested-by: Sibai Li <sibai.li@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

tcp: add const qualifiers where possible

Adding const qualifiers to pointers can ease code review, and spot some
bugs. It might allow compiler to optimize code further.

For example, is it legal to temporary write a null cksum into tcphdr
in tcp_md5_hash_header() ? I am afraid a sniffer could catch the
temporary null value...

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

dev: use name hash for dev_seq_ops

Instead of using the dev->next chain and trying to resync at each call to
dev_seq_start, use the name hash, keeping the bucket and the offset in
seq->private field.

Tests revealed the following results for ifconfig > /dev/null
* 1000 interfaces:
* 0.114s without patch
* 0.089s with patch
* 3000 interfaces:
* 0.489s without patch
* 0.110s with patch
* 5000 interfaces:
* 1.363s without patch
* 0.250s with patch
* 128000 interfaces (other setup):
* ~100s without patch
* ~30s with patch

Signed-off-by: Mihai Maruseac <mmaruseac@ixiacom.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

macvtap: Fix the minor device number allocation

On systems that create and delete lots of dynamic devices the
31bit linux ifindex fails to fit in the 16bit macvtap minor,
resulting in unusable macvtap devices.  I have systems running
automated tests that that hit this condition in just a few days.

Use a linux idr allocator to track which mavtap minor numbers
are available and and to track the association between macvtap
minor numbers and macvtap network devices.

Remove the unnecessary unneccessary check to see if the network
device we have found is indeed a macvtap device.  With macvtap
specific data structures it is impossible to find any other
kind of networking device.

Increase the macvtap minor range from 65536 to the full 20 bits
that is supported by linux device numbers.  It doesn't solve the
original problem but there is no penalty for a larger minor
device range.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

macvtap: Rewrite macvtap_newlink so the error handling works.

Place macvlan_common_newlink at the end of macvtap_newlink because
failing in newlink after registering your network device is not
supported.

Move device_create into a netdevice creation notifier.   The network device
notifier is the only hook that is called after the network device has been
registered with the device layer and before register_network_device returns
success.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

macvtap: Don't leak unreceived packets when we delete a macvtap device.

To avoid leaking packets in the receive queue.  Add a socket destructor
that will run whenever destroy a macvtap socket.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

macvtap: Fix macvtap_open races in the zero copy enable code.

To see if it is appropriate to enable the macvtap zero copy feature
don't test the lowerdev network device flags.   Instead test the
macvtap network device flags which are a direct copy of the lowerdev
flags.  This is important because nothing holds a reference to lowerdev
and on a very bad day we lowerdev could be a pointer to stale memory.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

macvtap: Close a race between macvtap_open and macvtap_dellink.

There is a small window in macvtap_open between looking up a
networking device and calling macvtap_set_queue in which
macvtap_del_queues called from macvtap_dellink.   After
calling macvtap_del_queues it is totally incorrect to
allow macvtap_set_queue to proceed so prevent success by
reporting that all of the available queues are in use.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

virtio_net: fix truesize underestimation

We must account in skb->truesize, the size of the fragments, not the
used part of them.

Doing this work is important to avoid unexpected OOM situations.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Rusty Russell <rusty@rustcorp.com.au>
CC: "Michael S. Tsirkin" <mst@redhat.com>
CC: virtualization@lists.linux-foundation.org
CC: Krishna Kumar <krkumar2@in.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: fix skb truesize underestimation

bnx2x allocates a full page per fragment.

We must account in skb->truesize, the size of the fragment, not the used
part of it.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: add opaque struct around skb frag page

I've split this bit out of the skb frag destructor patch since it helps enforce
the use of the fragment API.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgbi: convert to SKB paged frag API.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Cc: "James E.J. Bottomley" <JBottomley@parallels.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Mike Christie <michaelc@cs.wisc.edu>
Cc: James Bottomley <James.Bottomley@suse.de>
Cc: Karen Xie <kxie@chelsio.com>
Cc: linux-scsi@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4vf: convert to SKB paged frag API.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Cc: Casey Leedom <leedom@chelsio.com>
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: convert to SKB paged frag API.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Cc: Dimitris Michailidis <dm@chelsio.com>
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>

mlx4: convert to SKB paged frag API.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>

net: allow CAP_NET_RAW to set socket options IP{,V6}_TRANSPARENT

Up till now the IP{,V6}_TRANSPARENT socket options (which actually set
the same bit in the socket struct) have required CAP_NET_ADMIN
privileges to set or clear the option.

- we make clearing the bit not require any privileges.
- we allow CAP_NET_ADMIN to set the bit (as before this change)
- we allow CAP_NET_RAW to set this bit, because raw
  sockets already pretty much effectively allow you
  to emulate socket transparency.

Signed-off-by: Maciej Żenczykowski <maze@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: constify skbuff and Qdisc elements

Preliminary patch before tcp constification

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: remove unused tcp_fin() parameters

tcp_fin() only needs socket pointer, we can remove skb and th params.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'batman-adv/maint' of git://git.open-mesh.org/linux-merge

ll_temac: Add support for ethtool

This patch enables the ethtool interface. The implementation is done
using the libphy helper functions.

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

igb: fix a compile warning

control these three function declarations and
definitions with same macro CONFIG_PCI_IOV

drivers/net/ethernet/intel/igb/igb_main.c:165:
warning: ‘igb_vf_configure’ declared ‘static’ but never defined
drivers/net/ethernet/intel/igb/igb_main.c:166:
warning: ‘igb_find_enabled_vfs’ declared ‘static’ but never defined
drivers/net/ethernet/intel/igb/igb_main.c:167:
warning: ‘igb_check_vf_assignment’ declared ‘static’ but never defined

Signed-off-by: RongQing Li <roy.qing.li@gmail.com>
Acked-by: Greg Rose <gregory.v.rose@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

myri10ge: fix truesize underestimation

skb->truesize must account for allocated memory, not the used part of
it. Doing this work is important to avoid unexpected OOM situations.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Jon Mason <mason@myri.com>
Acked-by: Jon Mason <mason@myri.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

igbvf: fix truesize underestimation

igbvf allocates half a page per skb fragment. We must account
PAGE_SIZE/2 increments on skb->truesize, not the actual frag length.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pktgen: remove ndelay() call

Daniel Turull reported inaccuracies in pktgen when using low packet
rates, because we call ndelay(val) with values bigger than 20000.

Instead of calling ndelay() for delays < 100us, we can instead loop
calling ktime_now() only.

Reported-by: Daniel Turull <daniel.turull@gmail.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: use TCP_DEFAULT_INIT_RCVWND in tcp_fixup_rcvbuf()

Since commit 356f039822b (TCP: increase default initial receive
window.), we allow sender to send 10 (TCP_DEFAULT_INIT_RCVWND) segments.

Change tcp_fixup_rcvbuf() to reflect this change, even if no real change
is expected, since sysctl_tcp_rmem[1] = 87380 and this value
is bigger than tcp_fixup_rcvbuf() computed rcvmem (~23720)

Note: Since commit 356f039822b limited default window to maximum of
10*1460 and 2*MSS, we use same heuristic in this patch.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ip_gre: dont increase dev->needed_headroom on a live device

It seems ip_gre is able to change dev->needed_headroom on the fly.

Its is not legal unfortunately and triggers a BUG in raw_sendmsg()

skb = sock_alloc_send_skb(sk, ... + LL_ALLOCATED_SPACE(rt->dst.dev)

< another cpu change dev->needed_headromm (making it bigger)

...
skb_reserve(skb, LL_RESERVED_SPACE(rt->dst.dev));

We end with LL_RESERVED_SPACE() being bigger than LL_ALLOCATED_SPACE()
-> we crash later because skb head is exhausted.

Bug introduced in commit 243aad83 in 2.6.34 (ip_gre: include route
header_len in max_headroom calculation)

Reported-by: Elmar Vonlanthen <evonlanthen@gmail.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Timo Teräs <timo.teras@iki.fi>
CC: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge git://git./linux/kernel/git/davem/sparc

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc:
  sparc: Add alignment flag to PCI expansion resources
  sparc: Avoid calling sigprocmask()
  sparc: Use set_current_blocked()
  sparc32,leon: SRMMU MMU Table probe fix

Merge git://git./linux/kernel/git/davem/net

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net:
  fib_rules: fix unresolved_rules counting
  r8169: fix wrong eee setting for rlt8111evl
  r8169: fix driver shutdown WoL regression.
  ehea: Change maintainer to me
  pptp: pptp_rcv_core() misses pskb_may_pull() call
  tproxy: copy transparent flag when creating a time wait
  pptp: fix skb leak in pptp_xmit()
  bonding: use local function pointer of bond->recv_probe in bond_handle_frame
  smsc911x: Add support for SMSC LAN89218
  tg3: negate USE_PHYLIB flag check
  netconsole: enable netconsole can make net_device refcnt incorrent
  bluetooth: Properly clone LSM attributes to newly created child connections
  l2tp: fix a potential skb leak in l2tp_xmit_skb()
  bridge: fix hang on removal of bridge via netlink
  x25: Prevent skb overreads when checking call user data
  x25: Handle undersized/fragmented skbs
  x25: Validate incoming call user data lengths
  udplite: fast-path computation of checksum coverage
  IPVS netns shutdown/startup dead-lock
  netfilter: nf_conntrack: fix event flooding in GRE protocol tracker

mm: add a "struct page_frag" type containing a page, offset and length

A few network drivers currently use skb_frag_struct for this purpose but I have
patches which add additional fields and semantics there which these other uses
do not want.

A structure for reference sub-page regions seems like a generally useful thing
so do so instead of adding a network subsystem specific structure.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Jens Axboe <jaxboe@fusionio.com>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mlx4_en: fix skb truesize underestimation

skb->truesize must account for allocated memory, not the used part of
it. Doing this work is important to avoid unexpected OOM situations.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>

virtio_net: Clean up set_skb_frag()

Remove manual initialization in set_skb_frag, and instead
use __skb_fill_page_desc() to do the same. Patch tested
on net-next.

Signed-off-by: Krishna Kumar <krkumar2@in.ibm.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mm: fix race between mremap and removing migration entry

I don't usually pay much attention to the stale "? " addresses in
stack backtraces, but this lucky report from Pawel Sikora hints that
mremap's move_ptes() has inadequate locking against page migration.

 3.0 BUG_ON(!PageLocked(p)) in migration_entry_to_page():
 kernel BUG at include/linux/swapops.h:105!
 RIP: 0010:[<ffffffff81127b76>]  [<ffffffff81127b76>]
                       migration_entry_wait+0x156/0x160
  [<ffffffff811016a1>] handle_pte_fault+0xae1/0xaf0
  [<ffffffff810feee2>] ? __pte_alloc+0x42/0x120
  [<ffffffff8112c26b>] ? do_huge_pmd_anonymous_page+0xab/0x310
  [<ffffffff81102a31>] handle_mm_fault+0x181/0x310
  [<ffffffff81106097>] ? vma_adjust+0x537/0x570
  [<ffffffff81424bed>] do_page_fault+0x11d/0x4e0
  [<ffffffff81109a05>] ? do_mremap+0x2d5/0x570
  [<ffffffff81421d5f>] page_fault+0x1f/0x30

mremap's down_write of mmap_sem, together with i_mmap_mutex or lock,
and pagetable locks, were good enough before page migration (with its
requirement that every migration entry be found) came in, and enough
while migration always held mmap_sem; but not enough nowadays, when
there's memory hotremove and compaction.

The danger is that move_ptes() lets a migration entry dodge around
behind remove_migration_pte()'s back, so it's in the old location when
looking at the new, then in the new location when looking at the old.

Either mremap's move_ptes() must additionally take anon_vma lock(), or
migration's remove_migration_pte() must stop peeking for is_swap_entry()
before it takes pagetable lock.

Consensus chooses the latter: we prefer to add overhead to migration
than to mremapping, which gets used by JVMs and by exec stack setup.

Reported-and-tested-by: Paweł Sikora <pluto@agmk.net>
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: stable@vger.kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>