GitHub/LineageOS/G12/android_kernel_amlogic_linux-4.9.git
11 years agocgroup: cls: remove unnecessary task_cls_classid
Gao feng [Tue, 8 Oct 2013 03:05:20 +0000 (11:05 +0800)]
cgroup: cls: remove unnecessary task_cls_classid

We can get classid through cgroup_subsys_state,
this is directviewing and effective.

Signed-off-by: Gao feng <gaofeng@cn.fujitsu.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agocgroup: netprio: remove unnecessary task_netprioidx
Gao feng [Tue, 8 Oct 2013 03:05:19 +0000 (11:05 +0800)]
cgroup: netprio: remove unnecessary task_netprioidx

Since the tasks have been migrated to the cgroup,
there is no need to call task_netprioidx to get
task's cgroup id.

Signed-off-by: Gao feng <gaofeng@cn.fujitsu.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: ipv4 only populate IP_PKTINFO when needed
Shawn Bohrer [Mon, 7 Oct 2013 16:01:40 +0000 (11:01 -0500)]
net: ipv4 only populate IP_PKTINFO when needed

The since the removal of the routing cache computing
fib_compute_spec_dst() does a fib_table lookup for each UDP multicast
packet received.  This has introduced a performance regression for some
UDP workloads.

This change skips populating the packet info for sockets that do not have
IP_PKTINFO set.

Benchmark results from a netperf UDP_RR test:
Before 89789.68 transactions/s
After  90587.62 transactions/s

Benchmark results from a fio 1 byte UDP multicast pingpong test
(Multicast one way unicast response):
Before 12.63us RTT
After  12.48us RTT

Signed-off-by: Shawn Bohrer <sbohrer@rgmadvisors.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoudp: ipv4: Add udp early demux
Shawn Bohrer [Mon, 7 Oct 2013 16:01:39 +0000 (11:01 -0500)]
udp: ipv4: Add udp early demux

The removal of the routing cache introduced a performance regression for
some UDP workloads since a dst lookup must be done for each packet.
This change caches the dst per socket in a similar manner to what we do
for TCP by implementing early_demux.

For UDP multicast we can only cache the dst if there is only one
receiving socket on the host.  Since caching only works when there is
one receiving socket we do the multicast socket lookup using RCU.

For UDP unicast we only demux sockets with an exact match in order to
not break forwarding setups.  Additionally since the hash chains may be
long we only check the first socket to see if it is a match and not
waste extra time searching the whole chain when we might not find an
exact match.

Benchmark results from a netperf UDP_RR test:
Before 87961.22 transactions/s
After  89789.68 transactions/s

Benchmark results from a fio 1 byte UDP multicast pingpong test
(Multicast one way unicast response):
Before 12.97us RTT
After  12.63us RTT

Signed-off-by: Shawn Bohrer <sbohrer@rgmadvisors.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoudp: Only allow busy read/poll on connected sockets
Shawn Bohrer [Mon, 7 Oct 2013 16:01:38 +0000 (11:01 -0500)]
udp: Only allow busy read/poll on connected sockets

UDP sockets can receive packets from multiple endpoints and thus may be
received on multiple receive queues.  Since packets packets can arrive
on multiple receive queues we should not mark the napi_id for all
packets.  This makes busy read/poll only work for connected UDP sockets.

This additionally enables busy read/poll for UDP multicast packets as
long as the socket is connected by moving the check into
__udp_queue_rcv_skb().

Signed-off-by: Shawn Bohrer <sbohrer@rgmadvisors.com>
Suggested-by: Eric Dumazet <edumazet@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet_sched: increment drop counters in qdisc_tree_decrease_qlen()
Eric Dumazet [Mon, 7 Oct 2013 15:32:32 +0000 (08:32 -0700)]
net_sched: increment drop counters in qdisc_tree_decrease_qlen()

qdisc_tree_decrease_qlen() is called when some packets are dropped
on a qdisc, and we want to notify parents of qlen changes.

We also can increment parents qdisc qstats drop counters.

This permits more accurate drop counters up to root qdisc.

For example a graft operation typically resets a qdisc
(drops all packets) and call qdisc_tree_decrease_qlen()

Note that callers are responsible for their drop counters.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobonding: ensure that TLB mode's active slave has correct mac filter
Veaceslav Falico [Mon, 7 Oct 2013 07:17:20 +0000 (09:17 +0200)]
bonding: ensure that TLB mode's active slave has correct mac filter

Currently, in TLB mode we change mac addresses only by memcpy-ing the to
net_device->dev_addr, without actually setting them via
dev_set_mac_address(). This permits us to receive all the traffic always on
one mac address.

However, in case the interface flips, some drivers might enforce the
mac filtering for its FW/HW based on current ->dev_addr, and thus we won't
be able to receive traffic on that interface, in case it will be selected
as active in TLB mode.

Fix it by setting the mac address forcefully on every new active slave that
we select in TLB mode.

CC: Jay Vosburgh <fubar@us.ibm.com>
CC: Andy Gospodarek <andy@greyhouse.net>
CC: Yuval Mintz <yuvalmin@broadcom.com>
Reported-by: Yuval Mintz <yuvalmin@broadcom.com>
Tested-by: Yuval Mintz <yuvalmin@broadcom.com>
Signed-off-by: Veaceslav Falico <vfalico@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoveth: Showing peer of veth type dev in ip link (kernel side)
Masatake YAMATO [Fri, 4 Oct 2013 02:34:21 +0000 (11:34 +0900)]
veth: Showing peer of veth type dev in ip link (kernel side)

ip link has ability to show extra information of net work device if
kernel provides sunh information. With this patch veth driver can
provide its peer ifindex information to ip command via netlink
interface.

Signed-off-by: Masatake YAMATO <yamato@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoRevert "xen-netback: improve ring effeciency for guest RX"
Wei Liu [Tue, 8 Oct 2013 09:54:21 +0000 (10:54 +0100)]
Revert "xen-netback: improve ring effeciency for guest RX"

This reverts commit 4f0581d25827d5e864bcf07b05d73d0d12a20a5c.

The named changeset is causing problem. Let's aim to make this part less
fragile before trying to improve things.

Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Cc: Ian Campbell <ian.campbell@citrix.com>
Cc: Annie Li <annie.li@oracle.com>
Cc: Matt Wilson <msw@amazon.com>
Cc: Xi Xiong <xixiong@amazon.com>
Cc: David Vrabel <david.vrabel@citrix.com>
Cc: Paul Durrant <paul.durrant@citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: wan: remove deprecated IRQF_DISABLED
Michael Opdenacker [Sat, 5 Oct 2013 04:45:30 +0000 (06:45 +0200)]
net: wan: remove deprecated IRQF_DISABLED

This patch proposes to remove the use of the IRQF_DISABLED flag

It's a NOOP since 2.6.35 and it will be removed one day.

Signed-off-by: Michael Opdenacker <michael.opdenacker@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoirda: remove deprecated IRQF_DISABLED
Michael Opdenacker [Sat, 5 Oct 2013 04:39:24 +0000 (06:39 +0200)]
irda: remove deprecated IRQF_DISABLED

This patch proposes to remove the use of the IRQF_DISABLED flag

It's a NOOP since 2.6.35 and it will be removed one day.

Signed-off-by: Michael Opdenacker <michael.opdenacker@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: hamradio/yam: remove deprecated IRQF_DISABLED
Michael Opdenacker [Sat, 5 Oct 2013 04:25:46 +0000 (06:25 +0200)]
net: hamradio/yam: remove deprecated IRQF_DISABLED

This patch proposes to remove the use of the IRQF_DISABLED flag

It's a NOOP since 2.6.35 and it will be removed one day.

Signed-off-by: Michael Opdenacker <michael.opdenacker@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: hamradio/scc: remove deprecated IRQF_DISABLED
Michael Opdenacker [Sat, 5 Oct 2013 04:22:30 +0000 (06:22 +0200)]
net: hamradio/scc: remove deprecated IRQF_DISABLED

This patch proposes to remove the use of the IRQF_DISABLED flag

It's a NOOP since 2.6.35 and it will be removed one day.

Signed-off-by: Michael Opdenacker <michael.opdenacker@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: fujitsu: Remove ISA depdendency from Kconfig
Matthew Whitehead [Fri, 4 Oct 2013 21:03:08 +0000 (17:03 -0400)]
net: fujitsu: Remove ISA depdendency from Kconfig

There no longer are ISA drivers in the fujitsu directory, so remove the
dependency from the Kconfig.

Signed-off-by: Matthew Whitehead <tedheadster@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-next
David S. Miller [Mon, 7 Oct 2013 19:40:44 +0000 (15:40 -0400)]
Merge git://git./linux/kernel/git/linville/wireless-next

Conflicts:
drivers/net/wireless/brcm80211/brcmfmac/dhd_bus.h
drivers/net/wireless/rtlwifi/rtl8188ee/phy.h
drivers/net/wireless/rtlwifi/rtl8192ce/phy.h
drivers/net/wireless/rtlwifi/rtl8192de/phy.h
drivers/net/wireless/rtlwifi/rtl8723ae/phy.h

Just some minor conflicts between the wireless-next changes
and Joe Perches's "extern" removal from function prototypes
in header files.

John W. Linville says:

====================
Regarding the Bluetooth bits, Gustavo says:

"The big work here is from Marcel and Johan. They did a lot of work
in the L2CAP, HCI and MGMT layers. The most important ones are the
addition of a new MGMT command to enable/disable LE advertisement
and the introduction of the HCI user channel to allow applications
to get directly and exclusive access to Bluetooth devices."

As to the ath10k bits, Kalle says:

"Bartosz dropped support for qca98xx hw1.0 hardware from ath10k, it's
just too much to support it. Michal added support for the new firmware
interface. Marek fixed WEP in AP and IBSS mode. Rest of the changes are
minor fixes or cleanups."

And also:

"Major changes are:

* throughput improvements including aligning the RX frames correctly and
  optimising HTT layer (Michal)

* remove qca98xx hw1.0 support (Bartosz)

* add support for firmware version 999.999.0.636 (Michal)

* firmware htt statistics support (Kalle)

* fix WEP in AP and IBSS mode (Marek)

* fix a mutex unlock balance in debugfs file (Shafi)

And of course there's a lot of smaller fixes and cleanup."

For the wl12xx bits, Luca says:

"Here are some patches intended for 3.13.  Eliad is upstreaming a bunch
of patches that have been pending in the internal tree.  Mostly bugfixes
and other small improvements."

Along with that...

Arend and friends bring us a batch of brcmfmac updates, Larry Finger
offers some rtlwifi refactoring, and Sujith sends the usual batch of
ath9k updates.  As usual, there are a number of other small updates
from a variety of players as well.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: Separate the close_list and the unreg_list v2
Eric W. Biederman [Sun, 6 Oct 2013 02:26:05 +0000 (19:26 -0700)]
net: Separate the close_list and the unreg_list v2

Separate the unreg_list and the close_list in dev_close_many preventing
dev_close_many from permuting the unreg_list.  The permutations of the
unreg_list have resulted in cases where the loopback device is accessed
it has been freed in code such as dst_ifdown.  Resulting in subtle memory
corruption.

This is the second bug from sharing the storage between the close_list
and the unreg_list.  The issues that crop up with sharing are
apparently too subtle to show up in normal testing or usage, so let's
forget about being clever and use two separate lists.

v2: Make all callers pass in a close_list to dev_close_many

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf-next
David S. Miller [Fri, 4 Oct 2013 17:26:38 +0000 (13:26 -0400)]
Merge branch 'master' of git://git./linux/kernel/git/pablo/nf-next

Pablo Neira Ayuso says:

====================
The following patchset contains Netfilter updates for your net-next tree,
mostly ipset improvements and enhancements features, they are:

* Don't call ip_nest_end needlessly in the error path from me, suggested
  by Pablo Neira Ayuso, from Jozsef Kadlecsik.

* Fixed sparse warnings about shadowed variable and missing rcu annotation
  and fix of "may be used uninitialized" warnings, also from Jozsef.

* Renamed simple macro names to avoid namespace issues, reported by David
  Laight, again from Jozsef.

* Use fix sized type for timeout in the extension part, and cosmetic
  ordering of matches and targets separatedly in xt_set.c, from Jozsef.

* Support package fragments for IPv4 protos without ports from Anders K.
  Pedersen. For example this allows a hash:ip,port ipset containing the
  entry 192.168.0.1,gre:0 to match all package fragments for PPTP VPN
  tunnels to/from the host. Without this patch only the first package
  fragment (with fragment offset 0) was matched.

* Introduced a new operation to get both setname and family, from Jozsef.
  ip[6]tables set match and SET target need to know the family of the set
  in order to reject adding rules which refer to a set with a non-mathcing
  family. Currently such rules are silently accepted and then ignored
  instead of generating an error message to the user.

* Reworked extensions support in ipset types from Jozsef. The approach of
  defining structures with all variations is not manageable as the
  number of extensions grows. Therefore a blob for the extensions is
  introduced, somewhat similar to conntrack. The support of extensions
  which need a per data destroy function is added as well.

* When an element timed out in a list:set type of set, the garbage
  collector skipped the checking of the next element. So the purging
  was delayed to the next run of the gc, fixed by Jozsef.

* A small Kconfig fix: NETFILTER_NETLINK cannot be selected and
  ipset requires it.

* hash:net,net type from Oliver Smith. The type provides the ability to
  store pairs of subnets in a set.

* Comment for ipset entries from Oliver Smith. This makes possible to
  annotate entries in a set with comments, for example:

  ipset n foo hash:net,net comment
  ipset a foo 10.0.0.0/21,192.168.1.0/24 comment "office nets A and B"

* Fix of hash types resizing with comment extension from Jozsef.

* Fix of new extensions for list:set type when an element is added
  into a slot from where another element was pushed away from Jozsef.

* Introduction of a common function for the listing of the element
  extensions from Jozsef.

* Net namespace support for ipset from Vitaly Lavrov.

* hash:net,port,net type from Oliver Smith, which makes possible
  to store the triples of two subnets and a protocol, port pair in
  a set.

* Get xt_TCPMSS working with net namespace, by Gao feng.

* Use the proper net netnamespace to allocate skbs, also by Gao feng.

* A couple of cleanups for the conntrack SIP helper, by Holger
  Eitzenberger.

* Extend cttimeout to allow setting default conntrack timeouts via
  nfnetlink, so we can get rid of all our sysctl/proc interfaces in
  the future for timeout tuning, from me.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotcp: shrink tcp6_timewait_sock by one cache line
Eric Dumazet [Thu, 3 Oct 2013 21:27:25 +0000 (14:27 -0700)]
tcp: shrink tcp6_timewait_sock by one cache line

While working on tcp listener refactoring, I found that it
would really make things easier if sock_common could include
the IPv6 addresses needed in the lookups, instead of doing
very complex games to get their values (depending on sock
being SYN_RECV, ESTABLISHED, TIME_WAIT)

For this to happen, I need to be sure that tcp6_timewait_sock
and tcp_timewait_sock consume same number of cache lines.

This is possible if we only use 32bits for tw_ttd, as we remove
one 32bit hole in inet_timewait_sock

inet_tw_time_stamp() is defined and used, even if its current
implementation looks like tcp_time_stamp : We might need finer
resolution for tcp_time_stamp in the future.

Before patch : sizeof(struct tcp6_timewait_sock) = 0xc8

After patch : sizeof(struct tcp6_timewait_sock) = 0xc0

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoatl1e: enable support for NETIF_F_RXALL and NETIF_F_RXCRC features
Andrea Merello [Thu, 3 Oct 2013 19:18:37 +0000 (21:18 +0200)]
atl1e: enable support for NETIF_F_RXALL and NETIF_F_RXCRC features

This patch allows (optionally, via ethtool) the atl1e NIC to:
- Receive bad frames (runt, bad-fcs, etc..)
- Receive full frames without stripping the FCS.

This has been tested on my board by injecting runt and bad-fcs
frames with a FPGA-based device.

The particular scenario of receiving very short frames (<4 bytes)
without passing FCS to the upper layer has been also tested:
This could be potentially dangerous because the driver performs a
4 byte subtraction on the frame length, but I finally have NOT
added anything to avoid this because it seems the NIC always
discards frames so much short..
If someone still have some reason to worry about this, please
tell me.. I will add an explicit SW check..

Signed-off-by: Andrea Merello <andrea.merello@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agowireless: rt2x00: rt2800usb: add new devices
Xose Vazquez Perez [Sat, 28 Sep 2013 11:07:10 +0000 (13:07 +0200)]
wireless: rt2x00: rt2800usb: add new devices

0411,0241  RT5572  BUFFALO WI-U2-300D Wireless LAN Adapter
0789,0170  RT3572  Logitec LAN-W300AN/U2
0846,9013  RT3573  NETGEAR Adaptador USB Inalambrico Movistar
0df6,006e  RT3573  Sitecom WiFi USB adapter N900
2001,3c1f  RT3573  D-Link DWA-162 Wireless N900 Dual Band Adapter
2001,3c20  RT5372  D-Link DWA-140 Wireless N USB Adapter(rev.D)
2001,3c21  RT5572  D-Link DWA-160 Xtreme N Dual Band USB Adapter(rev.C)
2001,3c22  RT5372  D-Link DWA-132 Wireless N USB Adapter(rev.B)
2001,3c23  RT5372  D-Link GO-USB-N300 Wireless N Easy USB Adapter
2019,ab29  ?       Planex GW-USMirco300
20f4,724a  RT5572  TRENDnet N600 Wireless Dual Band USB Adapter

Cc: Ivo van Doorn <IvDoorn@gmail.com>
Cc: Gertjan van Wingerde <gwingerde@gmail.com>
Cc: Helmut Schaa <helmut.schaa@googlemail.com>
Cc: John W. Linville <linville@tuxdriver.com>
Cc: users@rt2x00.serialmonkey.com
Cc: linux-wireless@vger.kernel.org
Signed-off-by: Xose Vazquez Perez <xose.vazquez@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agortlwifi: rtl8188ee: Convert driver to use new rtl_phy_scan_operation_backup() routine
Larry Finger [Thu, 26 Sep 2013 18:25:33 +0000 (13:25 -0500)]
rtlwifi: rtl8188ee: Convert driver to use new rtl_phy_scan_operation_backup() routine

Now that rtl_phy_scan_operation_backup() exists, convert rtl8188ee to use it.
Routine rtl88e_phy_scan_operation_backup() is no longer used and is removed.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agortlwifi: rtl8723ae: Convert driver to use new rtl_phy_scan_operation_backup() routine
Larry Finger [Thu, 26 Sep 2013 18:25:32 +0000 (13:25 -0500)]
rtlwifi: rtl8723ae: Convert driver to use new rtl_phy_scan_operation_backup() routine

Now that rtl_phy_scan_operation_backup() exists, convert rtl8723ae to use it.
Routine rtl8723ae_phy_scan_operation_backup() is no longer used and is removed.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agortlwifi: rtl8192ce: Convert driver to use new rtl_phy_scan_operation_backup() routine
Larry Finger [Thu, 26 Sep 2013 18:25:31 +0000 (13:25 -0500)]
rtlwifi: rtl8192ce: Convert driver to use new rtl_phy_scan_operation_backup() routine

Now that rtl_phy_scan_operation_backup() exists, convert rtl8192de to use it.
Routine rtl92d_phy_scan_operation_backup() is no longer used and is removed.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agortlwifi: rtl8192c: Remove rtl8192c_phy_scan_operation_backup()
Larry Finger [Thu, 26 Sep 2013 18:25:30 +0000 (13:25 -0500)]
rtlwifi: rtl8192c: Remove rtl8192c_phy_scan_operation_backup()

Drivers rtl8192ce and rtl8192cu no longer rely on this routine. It is removed.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agortlwifi: rtl8192ce: Convert driver to use new rtl_phy_scan_operation_backup() routine
Larry Finger [Thu, 26 Sep 2013 18:25:29 +0000 (13:25 -0500)]
rtlwifi: rtl8192ce: Convert driver to use new rtl_phy_scan_operation_backup() routine

Now that rtl_phy_scan_operation_backup() exists, convert rtl8192ce to use it.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agortlwifi: rtl8192cu: Convert to use new rtl_phy_scan_operation_backup() routine
Larry Finger [Thu, 26 Sep 2013 18:25:28 +0000 (13:25 -0500)]
rtlwifi: rtl8192cu: Convert to use new rtl_phy_scan_operation_backup() routine

Now that the routine exists in driver rtlwifi, convert rtl8192cu to use it.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agortlwifi: Implement a common rtl_phy_scan_operation_backup() routine
Larry Finger [Thu, 26 Sep 2013 18:25:27 +0000 (13:25 -0500)]
rtlwifi: Implement a common rtl_phy_scan_operation_backup() routine

Several of the drivers supported by rtlwifi individually implement the same
routine that supports scans. As a first step, create this routine in
driver rtlwifi.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agortlwifi: Remove all remaining references to variable 'noise' in rtl_stats struct
Larry Finger [Wed, 25 Sep 2013 17:57:49 +0000 (12:57 -0500)]
rtlwifi: Remove all remaining references to variable 'noise' in rtl_stats struct

This completes removal of all places that reference variable 'noise'
in the rtl_stats struct. The definition of the struct is unchanged.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agortlwifi: rtl8188ee: Fix smatch warning in rtl8188ee/hw.c
Larry Finger [Wed, 25 Sep 2013 17:57:48 +0000 (12:57 -0500)]
rtlwifi: rtl8188ee: Fix smatch warning in rtl8188ee/hw.c

Smatch lists the following:
  CHECK   drivers/net/wireless/rtlwifi/rtl8188ee/hw.c
drivers/net/wireless/rtlwifi/rtl8188ee/hw.c:149 _rtl88ee_set_fw_clock_on() info: ignoring unreachable code.
drivers/net/wireless/rtlwifi/rtl8188ee/hw.c:149 _rtl88ee_set_fw_clock_on() info: ignoring unreachable code.

This info message is the result of a real error due to a missing break statement
in a "while (1)" loop.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Stable <stable@vger.kernel.org> [3.10+]
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agortlwifi: Fix smatch warnings in usb.c
Larry Finger [Wed, 25 Sep 2013 17:57:47 +0000 (12:57 -0500)]
rtlwifi: Fix smatch warnings in usb.c

Smatch displays the following:
  CHECK   drivers/net/wireless/rtlwifi/usb.c
drivers/net/wireless/rtlwifi/usb.c:458 _rtl_usb_rx_process_agg() warn: assigning (-98) to unsigned variable 'stats.noise'
drivers/net/wireless/rtlwifi/usb.c:503 _rtl_usb_rx_process_noagg() warn: assigning (-98) to unsigned variable 'stats.noise'
drivers/net/wireless/rtlwifi/usb.c:596 _rtl_rx_get_padding() info: ignoring unreachable code.
drivers/net/wireless/rtlwifi/usb.c:596 _rtl_rx_get_padding() info: ignoring unreachable code.

The variable 'stats.noise' is not used, thus the initializers are removed.
The unreachable code info is fixed by including the appropriate section inside
#ifdef .. #endif constructions.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agortlwifi: Fix smatch warning in pci.c
Larry Finger [Wed, 25 Sep 2013 17:57:46 +0000 (12:57 -0500)]
rtlwifi: Fix smatch warning in pci.c

Smatch reports the following:
  CHECK   drivers/net/wireless/rtlwifi/pci.c
drivers/net/wireless/rtlwifi/pci.c:739 _rtl_pci_rx_interrupt() warn: assigning (-98) to unsigned variable 'stats.noise'

The variable 'stats.noise' is not used. That initializer is removed.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agortlwifi: rtl8192_common: Fix smatch errors and warnings in rtl8192c/dm_common.c
Larry Finger [Wed, 25 Sep 2013 17:57:45 +0000 (12:57 -0500)]
rtlwifi: rtl8192_common: Fix smatch errors and warnings in rtl8192c/dm_common.c

Smatch lists the following:
  CHECK   drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c
drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c:551 rtl92c_dm_pwdb_monitor() info: ignoring unreachable code.
drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c:551 rtl92c_dm_pwdb_monitor() info: ignoring unreachable code.
drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c:870 rtl92c_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'txpwr_level' 2 <= 2
drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c:870 rtl92c_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'txpwr_level' 2 <= 2
drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c:882 rtl92c_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'txpwr_level' 2 <= 2
drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c:883 rtl92c_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'txpwr_level' 2 <= 2
drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c:891 rtl92c_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'txpwr_level' 2 <= 2
drivers/net/wireless/rtlwifi/rtl8192c/dm_common.c:892 rtl92c_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'txpwr_level' 2 <= 2

The unreachable code message is fixed by deleting the code that follows a return.

The errors are fixed by increasing the size of txpwr_level.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agortlwifi: rtl8192cu: Fix smatch warning in rtl8192cu/trx.c
Larry Finger [Wed, 25 Sep 2013 17:57:44 +0000 (12:57 -0500)]
rtlwifi: rtl8192cu: Fix smatch warning in rtl8192cu/trx.c

Smatch lists the following:
  CHECK   drivers/net/wireless/rtlwifi/rtl8192cu/trx.c
drivers/net/wireless/rtlwifi/rtl8192cu/trx.c:367 _rtl_rx_process() warn: assigning (-98) to unsigned variable 'stats.noise'

This variable is unused, thus the initializer is removed, as well as a
comment referring to that variable.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agortlwifi: rtl8192de: Fix smatch warnings in rtl8192de/hw.c
Larry Finger [Wed, 25 Sep 2013 17:57:43 +0000 (12:57 -0500)]
rtlwifi: rtl8192de: Fix smatch warnings in rtl8192de/hw.c

Smatch lists the following:
  CHECK   drivers/net/wireless/rtlwifi/rtl8192de/hw.c
drivers/net/wireless/rtlwifi/rtl8192de/hw.c:1200 rtl92de_set_qos() info: ignoring unreachable code.
drivers/net/wireless/rtlwifi/rtl8192de/hw.c:1200 rtl92de_set_qos() info: ignoring unreachable code.

The dead code is deleted.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agortlwifi: rtl8192du: Fix smatch errors in /rtl8192de/dm.c
Larry Finger [Wed, 25 Sep 2013 17:57:42 +0000 (12:57 -0500)]
rtlwifi: rtl8192du: Fix smatch errors in /rtl8192de/dm.c

Smatch lists the following:
  CHECK   drivers/net/wireless/rtlwifi/rtl8192de/dm.c
drivers/net/wireless/rtlwifi/rtl8192de/dm.c:1054 rtl92d_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'ofdm_index' 2 <= 2
drivers/net/wireless/rtlwifi/rtl8192de/dm.c:1056 rtl92d_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'ofdm_index' 2 <= 2
drivers/net/wireless/rtlwifi/rtl8192de/dm.c:1126 rtl92d_dm_txpower_tracking_callback_thermalmeter() debug: remove_pools: nr_children over 4000 (4596). (rtlpriv->dbg.global_debuglevel merged)
drivers/net/wireless/rtlwifi/rtl8192de/dm.c:1126 rtl92d_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'cckswing_table_ch1ch13' 33 <= 255
drivers/net/wireless/rtlwifi/rtl8192de/dm.c:1129 rtl92d_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'cckswing_table_ch1ch13' 33 <= 255
drivers/net/wireless/rtlwifi/rtl8192de/dm.c:1132 rtl92d_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'cckswing_table_ch1ch13' 33 <= 255
drivers/net/wireless/rtlwifi/rtl8192de/dm.c:1135 rtl92d_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'cckswing_table_ch1ch13' 33 <= 255
drivers/net/wireless/rtlwifi/rtl8192de/dm.c:1138 rtl92d_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'cckswing_table_ch1ch13' 33 <= 255
drivers/net/wireless/rtlwifi/rtl8192de/dm.c:1141 rtl92d_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'cckswing_table_ch1ch13' 33 <= 255
drivers/net/wireless/rtlwifi/rtl8192de/dm.c:1144 rtl92d_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'cckswing_table_ch1ch13' 33 <= 255
drivers/net/wireless/rtlwifi/rtl8192de/dm.c:1147 rtl92d_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'cckswing_table_ch1ch13' 33 <= 255
drivers/net/wireless/rtlwifi/rtl8192de/dm.c:1151 rtl92d_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'cckswing_table_ch14' 33 <= 255
drivers/net/wireless/rtlwifi/rtl8192de/dm.c:1154 rtl92d_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'cckswing_table_ch14' 33 <= 255
drivers/net/wireless/rtlwifi/rtl8192de/dm.c:1157 rtl92d_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'cckswing_table_ch14' 33 <= 255
drivers/net/wireless/rtlwifi/rtl8192de/dm.c:1160 rtl92d_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'cckswing_table_ch14' 33 <= 255
drivers/net/wireless/rtlwifi/rtl8192de/dm.c:1163 rtl92d_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'cckswing_table_ch14' 33 <= 255
drivers/net/wireless/rtlwifi/rtl8192de/dm.c:1166 rtl92d_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'cckswing_table_ch14' 33 <= 255
drivers/net/wireless/rtlwifi/rtl8192de/dm.c:1169 rtl92d_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'cckswing_table_ch14' 33 <= 255
drivers/net/wireless/rtlwifi/rtl8192de/dm.c:1172 rtl92d_dm_txpower_tracking_callback_thermalmeter() error: buffer overflow 'cckswing_table_ch14' 33 <= 255

This patch fixes several off-by-one errors. It also removes a comment
referencing variable 'noise' in the rts_stats struct.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agort2800: add support for radio chip RF3070
Stanislaw Gruszka [Wed, 25 Sep 2013 13:34:55 +0000 (15:34 +0200)]
rt2800: add support for radio chip RF3070

Add support for new RF chip ID: 3070. It seems to be the same as 5370,
maybe vendor just put wrong value on the eeprom, but add this id anyway
since devices with it showed on the marked.

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobrcmfmac: rework rx path bus interface
Arend van Spriel [Wed, 25 Sep 2013 11:05:46 +0000 (13:05 +0200)]
brcmfmac: rework rx path bus interface

The brcmfmac has common and bus specific part. The rx function api
was using sk_buff_head. Changed to sk_buff instead.

Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Franky (Zhenhui) Lin <frankyl@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>
11 years agobrcmfmac: Use fw filename and nvram based of devid for sdio.
Hante Meuleman [Wed, 25 Sep 2013 11:05:45 +0000 (13:05 +0200)]
brcmfmac: Use fw filename and nvram based of devid for sdio.

SDIO firmware download routines uses one name for firmware file
and nvram file for all sdio devices. This is not user friendly.
Use fw filename and nvram filename based upon chip id and revision.

Reported-by: Stephen Warren <swarren@wwwdotorg.org>
Reviewed-by: Franky (Zhenhui) Lin <frankyl@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Hante Meuleman <meuleman@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobrcmfmac: fix sparse error 'bad constant expression'
Arend van Spriel [Wed, 25 Sep 2013 11:05:44 +0000 (13:05 +0200)]
brcmfmac: fix sparse error 'bad constant expression'

The definition of BRCMF_EVENTING_MASK_LEN results in a sparse
error message

.../fweh.c:331:22: error: bad constant expression
.../fweh.c:388:22: error: bad constant expression
.../dhd_common.c:256:22: error: bad constant expression

This is caused by the use of roundup() in BRCMF_EVENTING_MASK_LEN
and it is used to allocate an array variable on the stack. Better
use DIV_ROUND_UP() macro.

Reviewed-by: Franky (Zhenhui) Lin <frankyl@broadcom.com>
Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobrcmfmac: reserve memory for bus layer in sk_buff::cb
Franky Lin [Wed, 25 Sep 2013 11:05:43 +0000 (13:05 +0200)]
brcmfmac: reserve memory for bus layer in sk_buff::cb

Bus layer need to share sk_buff::cb with firmware signal feature. Reserve
necessary memory so they won't overwrite each other.

Reviewed-by: Arend Van Spriel <arend@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Franky Lin <frankyl@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobrcmfmac: add valid core index check in related functions
Franky Lin [Wed, 25 Sep 2013 11:05:42 +0000 (13:05 +0200)]
brcmfmac: add valid core index check in related functions

Perform a valid check for core index to avoid illegal address access.

Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Franky Lin <frankyl@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobrcmfmac: add BCM4339 SDIO interface support
Franky Lin [Wed, 25 Sep 2013 11:05:41 +0000 (13:05 +0200)]
brcmfmac: add BCM4339 SDIO interface support

BCM4339 is an a/b/g/n/ac 1x1 WiFi chip. This patch adds support for it through
SDIO interface.

Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Franky Lin <frankyl@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobrcmfmac: sync firmware event list
Franky Lin [Wed, 25 Sep 2013 11:05:40 +0000 (13:05 +0200)]
brcmfmac: sync firmware event list

Update event list to keep sync with firmware development. Use calculated event
mask length instead of hard-coded value.

Reviewed-by: Arend Van Spriel <arend@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Franky Lin <frankyl@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoMerge branch 'for-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/bluetoot...
John W. Linville [Thu, 3 Oct 2013 20:19:07 +0000 (16:19 -0400)]
Merge branch 'for-upstream' of git://git./linux/kernel/git/bluetooth/bluetooth-next

11 years agoMerge tag 'for-linville-20131001' of git://github.com/kvalo/ath
John W. Linville [Thu, 3 Oct 2013 20:16:34 +0000 (16:16 -0400)]
Merge tag 'for-linville-20131001' of git://github.com/kvalo/ath

11 years agoMerge branch 'for-linville' of git://git.kernel.org/pub/scm/linux/kernel/git/luca...
John W. Linville [Thu, 3 Oct 2013 20:15:35 +0000 (16:15 -0400)]
Merge branch 'for-linville' of git://git./linux/kernel/git/luca/wl12xx

11 years ago3com: Fix drivers/net/ethernet/3com/Kconfig references to PCMCIA and 3c515
Matthew Whitehead [Wed, 2 Oct 2013 15:08:03 +0000 (11:08 -0400)]
3com: Fix drivers/net/ethernet/3com/Kconfig references to PCMCIA and 3c515

The Vortex driver works with PCI and Cardbus devices, not PCMCIA.

There never was an EISA 3c515 card, only ISA, so remove that option.

Signed-off-by: Matthew Whitehead <tedheadster@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: ipv4: Change variable type to bool
Peter Senna Tschudin [Wed, 2 Oct 2013 12:19:51 +0000 (14:19 +0200)]
net: ipv4: Change variable type to bool

The variable fully_acked is only assigned the values true and false.
Change its type to bool.

The simplified semantic patch that find this problem is as
follows (http://coccinelle.lip6.fr/):

@exists@
type T;
identifier b;
@@
- T
+ bool
  b = ...;
  ... when any
  b = \(true\|false\)

Signed-off-by: Peter Senna Tschudin <peter.senna@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: myri10ge: Change variable type to bool
Peter Senna Tschudin [Wed, 2 Oct 2013 12:19:50 +0000 (14:19 +0200)]
net: myri10ge: Change variable type to bool

There is the rc variable on both myri10ge_ss_lock_napi and
myri10ge_ss_lock_poll functions. In both cases rc is only assigned the
values true and false. Both functions already return bool. Change rc
type to bool.

The simplified semantic patch that find this problem is as
follows (http://coccinelle.lip6.fr/):

@exists@
type T;
identifier b;
@@
- T
+ bool
  b = ...;
  ... when any
  b = \(true\|false\)

Signed-off-by: Peter Senna Tschudin <peter.senna@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: bnx2x: Change variable type to bool
Peter Senna Tschudin [Wed, 2 Oct 2013 12:19:49 +0000 (14:19 +0200)]
net: bnx2x: Change variable type to bool

The variable rc is only assigned the values true and false.
The function bnx2x_prev_is_path_marked already returns bool.
Change rc type to bool.

The simplified semantic patch that find this problem is as
follows (http://coccinelle.lip6.fr/):

@exists@
type T;
identifier b;
@@
- T
+ bool
  b = ...;
  ... when any
  b = \(true\|false\)

Signed-off-by: Peter Senna Tschudin <peter.senna@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: atl1c: Change variable type to bool
Peter Senna Tschudin [Wed, 2 Oct 2013 12:19:48 +0000 (14:19 +0200)]
net: atl1c: Change variable type to bool

The variable ret is only assigned the values true and false.
The function atl1c_read_eeprom already returns bool. Change
ret type to bool.

The simplified semantic patch that find this problem is as
follows (http://coccinelle.lip6.fr/):

@exists@
type T;
identifier b;
@@
- T
+ bool
  b = ...;
  ... when any
  b = \(true\|false\)

Signed-off-by: Peter Senna Tschudin <peter.senna@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'bond_hash'
David S. Miller [Thu, 3 Oct 2013 19:37:10 +0000 (15:37 -0400)]
Merge branch 'bond_hash'

Nikolay Aleksandrov says:

====================
This is a complete remake of my old patch that modified the bonding hash
functions to use skb_flow_dissect which was suggested by Eric Dumazet.
This time around I've left the old modes although using a new hash function
again suggested by Eric, which is the same for all modes. The only
difference is the way the headers are obtained. The old modes obtain them
as before in order to address concerns about speed, but the 2 new ones use
skb_flow_dissect. The unification of the hash function allows to remove a
pointer from struct bonding and also a few extra functions that dealt with
it. Two new functions are added which take care of the hashing based on
bond->params.xmit_policy only:
bond_xmit_hash() - global function, used by XOR and 3ad modes
bond_flow_dissect() - used by bond_xmit_hash() to obtain the necessary
headers and combine them according to bond->params.xmit_policy.
Also factor out the ports extraction from skb_flow_dissect and add a new
function - skb_flow_get_ports() which can be re-used.

v2: add the flow_dissector patch and use skb_flow_get_ports in patch 02
v3: fix a bug in the flow_dissector patch that caused a different thoff
    by modifying the thoff argument in skb_flow_get_ports directly, most
    of the users already do it anyway.
    Also add the necessary export symbol for skb_flow_get_ports.
v4: integrate the thoff bug fix in patch 01
v5: disintegrate the thoff bug fix and re-base on top of Eric's fix
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobonding: document the new xmit policy modes and update the changed ones
Nikolay Aleksandrov [Wed, 2 Oct 2013 11:39:26 +0000 (13:39 +0200)]
bonding: document the new xmit policy modes and update the changed ones

Add new documentation for encap2+3 and encap3+4, also update the formula
for the old modes due to the changes.

Signed-off-by: Nikolay Aleksandrov <nikolay@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobonding: modify the old and add new xmit hash policies
Nikolay Aleksandrov [Wed, 2 Oct 2013 11:39:25 +0000 (13:39 +0200)]
bonding: modify the old and add new xmit hash policies

This patch adds two new hash policy modes which use skb_flow_dissect:
3 - Encapsulated layer 2+3
4 - Encapsulated layer 3+4
There should be a good improvement for tunnel users in those modes.
It also changes the old hash functions to:
hash ^= (__force u32)flow.dst ^ (__force u32)flow.src;
hash ^= (hash >> 16);
hash ^= (hash >> 8);

Where hash will be initialized either to L2 hash, that is
SRCMAC[5] XOR DSTMAC[5], or to flow->ports which should be extracted
from the upper layer. Flow's dst and src are also extracted based on the
xmit policy either directly from the buffer or by using skb_flow_dissect,
but in both cases if the protocol is IPv6 then dst and src are obtained by
ipv6_addr_hash() on the real addresses. In case of a non-dissectable
packet, the algorithms fall back to L2 hashing.
The bond_set_mode_ops() function is now obsolete and thus deleted
because it was used only to set the proper hash policy. Also we trim a
pointer from struct bonding because we no longer need to keep the hash
function, now there's only a single hash function - bond_xmit_hash that
works based on bond->params.xmit_policy.

The hash function and skb_flow_dissect were suggested by Eric Dumazet.
The layer names were suggested by Andy Gospodarek, because I suck at
semantics.

Signed-off-by: Nikolay Aleksandrov <nikolay@redhat.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Acked-by: Veaceslav Falico <vfalico@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoflow_dissector: factor out the ports extraction in skb_flow_get_ports
Nikolay Aleksandrov [Wed, 2 Oct 2013 11:39:24 +0000 (13:39 +0200)]
flow_dissector: factor out the ports extraction in skb_flow_get_ports

Factor out the code that extracts the ports from skb_flow_dissect and
add a new function skb_flow_get_ports which can be re-used.

Suggested-by: Veaceslav Falico <vfalico@redhat.com>
Signed-off-by: Nikolay Aleksandrov <nikolay@redhat.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Veaceslav Falico <vfalico@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoinet: consolidate INET_TW_MATCH
Eric Dumazet [Wed, 2 Oct 2013 11:29:50 +0000 (04:29 -0700)]
inet: consolidate INET_TW_MATCH

TCP listener refactoring, part 2 :

We can use a generic lookup, sockets being in whatever state, if
we are sure all relevant fields are at the same place in all socket
types (ESTABLISH, TIME_WAIT, SYN_RECV)

This patch removes these macros :

 inet_addrpair, inet_addrpair, tw_addrpair, tw_portpair

And adds :

 sk_portpair, sk_addrpair, sk_daddr, sk_rcv_saddr

Then, INET_TW_MATCH() is really the same than INET_MATCH()

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoBluetooth: Only one command per L2CAP LE signalling is supported
Marcel Holtmann [Thu, 3 Oct 2013 08:26:37 +0000 (01:26 -0700)]
Bluetooth: Only one command per L2CAP LE signalling is supported

The Bluetooth specification makes it clear that only one command
should be present in the L2CAP LE signalling packet. So tighten
the checks here and restrict it to exactly one command.

This is different from L2CAP BR/EDR signalling where multiple
commands can be part of the same packet.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
11 years agoBluetooth: Check minimum length of SMP packets
Marcel Holtmann [Thu, 3 Oct 2013 08:23:08 +0000 (01:23 -0700)]
Bluetooth: Check minimum length of SMP packets

When SMP packets are received, make sure they contain at least 1 byte
header for the opcode. If not, drop the packet and disconnect the link.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
11 years agoBluetooth: Drop packets on ATT fixed channel on BR/EDR
Marcel Holtmann [Thu, 3 Oct 2013 09:54:11 +0000 (02:54 -0700)]
Bluetooth: Drop packets on ATT fixed channel on BR/EDR

The ATT fixed channel is only valid when using LE connections. On
BR/EDR it is required to go through L2CAP connection oriented
channel for ATT.

Drop ATT packets when they are received on a BR/EDR connection.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
11 years agoBluetooth: L2CAP connectionless channels are only valid for BR/EDR
Marcel Holtmann [Thu, 3 Oct 2013 07:03:39 +0000 (00:03 -0700)]
Bluetooth: L2CAP connectionless channels are only valid for BR/EDR

When receiving connectionless packets on a LE connection, just drop
the packet. There is no concept of connectionless channels for LE.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
11 years agoBluetooth: SMP packets are only valid on LE connections
Marcel Holtmann [Thu, 3 Oct 2013 07:00:57 +0000 (00:00 -0700)]
Bluetooth: SMP packets are only valid on LE connections

When receiving SMP packets on a BR/EDR connection, then just drop
the packet and do not try to process it.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
11 years agoBluetooth: Don't copy L2CAP LE signalling to raw sockets
Marcel Holtmann [Thu, 3 Oct 2013 06:51:49 +0000 (23:51 -0700)]
Bluetooth: Don't copy L2CAP LE signalling to raw sockets

The L2CAP raw sockets are only used for BR/EDR signalling. Packets
on LE links should not be forwarded there.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
11 years agoBluetooth: Fix switch statement order for L2CAP fixed channels
Marcel Holtmann [Thu, 3 Oct 2013 06:46:54 +0000 (23:46 -0700)]
Bluetooth: Fix switch statement order for L2CAP fixed channels

The switch statement for the various L2CAP fixed channel handlers
is not really ordered.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
11 years agoBluetooth: Allow changing device class when BR/EDR is disabled
Marcel Holtmann [Thu, 3 Oct 2013 06:37:29 +0000 (23:37 -0700)]
Bluetooth: Allow changing device class when BR/EDR is disabled

Changing the device class when BR/EDR is disabled has no visible
effect for remote devices. However to simplify the logic allow it
as long as the controller supports BR/EDR operations.

If it is not allowed, then the overall logic becomes rather
complicated since the class of device values would need clearing
or restoring when BR/EDR setting changes.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
11 years agoBluetooth: Restrict loading of long term keys to LE capable controllers
Marcel Holtmann [Thu, 3 Oct 2013 04:16:08 +0000 (21:16 -0700)]
Bluetooth: Restrict loading of long term keys to LE capable controllers

Loading long term keys into a BR/EDR only controller make no sense.
The kernel would never use any of these keys. So instead of allowing
userspace to waste memory, reject such operation with a not supported
error message.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
11 years agoBluetooth: Restrict loading of link keys to BR/EDR capable controllers
Marcel Holtmann [Thu, 3 Oct 2013 04:16:07 +0000 (21:16 -0700)]
Bluetooth: Restrict loading of link keys to BR/EDR capable controllers

Loading link keys into a LE only controller make no sense. The kernel
would never use any of these keys. So instead of allowing userspace
to waste memory, reject such operation with a not supported error
message.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
11 years agoBluetooth: Allow setting static address even if LE is disabled
Marcel Holtmann [Thu, 3 Oct 2013 05:10:32 +0000 (22:10 -0700)]
Bluetooth: Allow setting static address even if LE is disabled

Setting the static address does not depend on LE beeing enabled. It
only depends on a controller with LE support.

When depending on LE enabled this command becomes really complicated
since in case LE gets disabled, it would be required to clear the
static address and also its random address representation inside
the controller. With future support for private addresses such
complex setup should be avoided.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
11 years agoBluetooth: Restrict SSP setting changes to BR/EDR enabled controllers
Marcel Holtmann [Thu, 3 Oct 2013 04:31:52 +0000 (21:31 -0700)]
Bluetooth: Restrict SSP setting changes to BR/EDR enabled controllers

Only when BR/EDR is supported and enabled, allow changing of the SSP
setting. Just checking if the hardware supports SSP is not enough
since it might be the case that BR/EDR is disabled.

In the case that BR/EDR is disabled, but SSP supported by the
controller the not supported error message is now returned.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
11 years agoath10k: wmi: Convert use of 6 to ETH_ALEN
Joe Perches [Thu, 3 Oct 2013 03:39:11 +0000 (20:39 -0700)]
ath10k: wmi: Convert use of 6 to ETH_ALEN

Use the appropriate define instead of 6.

Signed-off-by: Joe Perches <joe@perches.com>
Noticed-by: Julia Lawall <julia.lawall@lip6.fr> via spatch script
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet:drivers/net: Miscellaneous conversions to ETH_ALEN
Joe Perches [Wed, 2 Oct 2013 02:04:40 +0000 (19:04 -0700)]
net:drivers/net: Miscellaneous conversions to ETH_ALEN

Convert the memset/memcpy uses of 6 to ETH_ALEN
where appropriate.

Also convert some struct definitions and u8 array
declarations of [6] to ETH_ALEN.

Signed-off-by: Joe Perches <joe@perches.com>
Acked-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoinclude/linux/skbuff.h: move CONFIG_XFRM check inside the skb_sec_path()
Denis Kirjanov [Wed, 2 Oct 2013 01:58:32 +0000 (05:58 +0400)]
include/linux/skbuff.h: move CONFIG_XFRM check inside the skb_sec_path()

And thus we have only one function definition

Signed-off-by: Denis Kirjanov <kda@linux-powerpc.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotcp: sndbuf autotuning improvements
Eric Dumazet [Tue, 1 Oct 2013 17:23:44 +0000 (10:23 -0700)]
tcp: sndbuf autotuning improvements

tcp_fixup_sndbuf() is underestimating initial send buffer requirements.

It was not noticed because big GSO packets were escaping the limitation,
but with smaller TSO packets (or TSO/GSO/SG off), application hits
sk_sndbuf before having a chance to fill enough packets in socket write
queue.

- initial cwnd can be bigger than 10 for specific routes

- SKB_TRUESIZE() is a bit under real needs in some cases,
  because of power-of-two rounding in kmalloc()

- Fast Recovery (RFC 5681 3.2) : Cubic needs 70% factor

- Extra cushion (application might react slowly to POLLOUT)

tcp_v4_conn_req_fastopen() needs to call tcp_init_metrics() before
calling tcp_init_buffer_space()

Then we realize tcp_new_space() should call tcp_fixup_sndbuf()
instead of duplicating this stuff.

Rename tcp_fixup_sndbuf() to tcp_sndbuf_expand() to be more
descriptive.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Maciej Żenczykowski <maze@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agofib_trie: avoid a redundant bit judgement in inflate
baker.zhang [Mon, 30 Sep 2013 23:45:09 +0000 (07:45 +0800)]
fib_trie: avoid a redundant bit judgement in inflate

Because 'node' is the i'st child of 'oldnode',
thus, here 'i' equals
tkey_extract_bits(node->key, oldtnode->pos, oldtnode->bits)

we just get 1 more bit,
and need not care the detail value of this bits.

I apologize for the mistake.

I generated the patch on a branch version,
and did not notice the put_child has been changed.

I have redone the test on HEAD version with my patch.

two cases are used.
case 1. inflate a node which has a leaf child node.
case 2: inflate a node which has a an child node with skipped bits

test env:
  ip link set eth0 up
  ip a add dev eth0 192.168.11.1/32
here, we just focus on route table(MAIN),
so I use a "192.168.11.1/32" address to simplify the test case.

call trace:
+ fib_insert_node
+ + trie_rebalance
+ + + resize
+ + + + inflate

Test case 1:  inflate a node which has a leaf child node.

===========================================================
step 1. prepare a fib trie
------------------------------------------
  ip r a 192.168.0.0/24 via 192.168.11.1
  ip r a 192.168.1.0/24 via 192.168.11.1

we get a fib trie.
root@baker:~# cat /proc/net/fib_trie
Main:
  +-- 192.168.0.0/23 1 0 0
   |-- 192.168.0.0
    /24 universe UNICAST
   |-- 192.168.1.0
    /24 universe UNICAST
Local:
.....

step 2. Add the third route
------------------------------------------
root@baker:~# ip r a 192.168.2.0/24 via 192.168.11.1

A fib_trie leaf will be inserted in fib_insert_node before trie_rebalance.

For function 'inflate':
'inflate' is called with following trie.
  +-- 192.168.0.0/22 1 1 0 <=== tn node
    +-- 192.168.0.0/23 1 0 0    <== node a
        |-- 192.168.0.0
          /24 universe UNICAST
        |-- 192.168.1.0
          /24 universe UNICAST
      |-- 192.168.2.0          <== leaf(node b)

When process node b, which is a leaf. here:
i is 1,
node key "192.168.2.0"
oldnode is (pos:22, bits:1)

unpatch source:
tkey_extract_bits(node->key, oldtnode->pos + oldtnode->bits, 1)
it equals:
tkey_extract_bits("192.168,2,0", 22 + 1, 1)

thus got 0, and call put_child(tn, 2*i, node); <== 2*i=2.

patched source:
tkey_extract_bits(node->key, oldtnode->pos, oldtnode->bits + 1),
tkey_extract_bits("192.168,2,0", 22, 1 + 1)  <== get 2.

Test case 2:  inflate a node which has a an child node with skipped bits
==========================================================================
step 1. prepare a fib trie.
  ip link set eth0 up
  ip a add dev eth0 192.168.11.1/32
  ip r a 192.168.128.0/24 via 192.168.11.1
  ip r a 192.168.0.0/24  via 192.168.11.1
  ip r a 192.168.16.0/24   via 192.168.11.1
  ip r a 192.168.32.0/24  via 192.168.11.1
  ip r a 192.168.48.0/24  via 192.168.11.1
  ip r a 192.168.144.0/24   via 192.168.11.1
  ip r a 192.168.160.0/24   via 192.168.11.1
  ip r a 192.168.176.0/24   via 192.168.11.1

check:
root@baker:~# cat /proc/net/fib_trie
Main:
  +-- 192.168.0.0/16 1 0 0
     +-- 192.168.0.0/18 2 0 0
        |-- 192.168.0.0
           /24 universe UNICAST
        |-- 192.168.16.0
           /24 universe UNICAST
        |-- 192.168.32.0
           /24 universe UNICAST
        |-- 192.168.48.0
           /24 universe UNICAST
     +-- 192.168.128.0/18 2 0 0
        |-- 192.168.128.0
           /24 universe UNICAST
        |-- 192.168.144.0
           /24 universe UNICAST
        |-- 192.168.160.0
           /24 universe UNICAST
        |-- 192.168.176.0
           /24 universe UNICAST
Local:
  ...

step 2. add a route to trigger inflate.
  ip r a 192.168.96.0/24   via 192.168.11.1

This command will call serveral times inflate.
In the first time, the fib_trie is:
________________________
+-- 192.168.128.0/(16, 1) <== tn node
 +-- 192.168.0.0/(17, 1)  <== node a
  +-- 192.168.0.0/(18, 2)
   |-- 192.168.0.0
   |-- 192.168.16.0
   |-- 192.168.32.0
   |-- 192.168.48.0
  |-- 192.168.96.0
 +-- 192.168.128.0/(18, 2) <== node b.
  |-- 192.168.128.0
  |-- 192.168.144.0
  |-- 192.168.160.0
  |-- 192.168.176.0

NOTE: node b is a interal node with skipped bits.
here,
i:1,
node->key "192.168.128.0",
oldnode:(pos:16, bits:1)
so
tkey_extract_bits(node->key, oldtnode->pos + oldtnode->bits, 1)
it equals:
tkey_extract_bits("192.168,128,0", 16 + 1, 1) <=== 0

tkey_extract_bits(node->key, oldtnode->pos, oldtnode->bits, 1)
it equals:
tkey_extract_bits("192.168,128,0", 16, 1+1) <=== 2

2*i + 0 == 2, so the result is same.

Signed-off-by: baker.zhang <baker.kernel@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoBluetooth: Fix memory leak with L2CAP signal channels
Marcel Holtmann [Wed, 2 Oct 2013 15:28:21 +0000 (08:28 -0700)]
Bluetooth: Fix memory leak with L2CAP signal channels

The wrong type of L2CAP signalling packets on the wrong type of
either BR/EDR or LE links need to be dropped. When that happens
the packet is dropped, but the memory not freed. So actually
free the memory as well.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
11 years agoisdn: eicon: free pointer after using it in log msg in divas_um_idi_delete_entity()
Jesper Juhl [Mon, 30 Sep 2013 19:25:27 +0000 (21:25 +0200)]
isdn: eicon: free pointer after using it in log msg in divas_um_idi_delete_entity()

Not really a problem, but nice IMHO; the Coverity static analyzer
complains that we use the pointer 'e' after it has been freed, so move
the freeing below the final use, even if that use is just using the
value of the pointer and not actually dereferencing it.

Signed-off-by: Jesper Juhl <jj@chaosbits.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoxen-netfront: convert to GRO API
Wei Liu [Mon, 30 Sep 2013 12:46:34 +0000 (13:46 +0100)]
xen-netfront: convert to GRO API

Anirban was seeing netfront received MTU size packets, which downgraded
throughput. The following patch makes netfront use GRO API which
improves throughput for that case.

Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Cc: Anirban Chakraborty <abchak@juniper.net>
Cc: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Konrad Wilk <konrad.wilk@oracle.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotg3: use phylib when robo switch is in use
Hauke Mehrtens [Sat, 28 Sep 2013 21:15:28 +0000 (23:15 +0200)]
tg3: use phylib when robo switch is in use

When a switch is connected as a PHY to the MAC driven by tg3, use
phylib and provide the phy address to tg3 from the sprom.

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Acked-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agossb: provide phy address for Gigabit Ethernet driver
Hauke Mehrtens [Sat, 28 Sep 2013 21:15:27 +0000 (23:15 +0200)]
ssb: provide phy address for Gigabit Ethernet driver

Add a function to provide the phy address which should be used to the
Gigabit Ethernet driver connected to ssb.

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Reviewed-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotg3: add support a phy at an address different than 01
Hauke Mehrtens [Sat, 28 Sep 2013 21:15:26 +0000 (23:15 +0200)]
tg3: add support a phy at an address different than 01

When phylib was in use tg3 only searched at address 01 on the mdio
bus and did not work with any other address. On the BCM4705 SoCs the
switch is connected as a PHY behind the MAC driven by tg3 and it is at
PHY address 30 in most cases. This is a preparation patch to allow
support for such switches.

phy_addr is set to TG3_PHY_MII_ADDR for all devices, which are using
phylib, so this should not change any behavior.

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Acked-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoBluetooth: Add the definition for Slave Page Response Timeout
DoHyun Pyun [Wed, 2 Oct 2013 12:54:53 +0000 (21:54 +0900)]
Bluetooth: Add the definition for Slave Page Response Timeout

The Slave Page Response Timeout event indicates to the Host that a
slave page response timeout has occurred in the BR/EDR Controller.

The Core Spec Addendum 4 adds this command in part B Connectionless
Slave Broadcast.

Bluetooth Core Specification Addendum 4 - Page 110

"7.7.72 Slave Page Response Timeout Event [New Section]
...
Note: this event will be generated if the slave BR/EDR Controller
responds to a page but does not receive the master FHS packet
(see Baseband, Section 8.3.3) within pagerespTO.

Event Parameters: NONE"

Signed-off-by: Dohyun Pyun <dh79.pyun@samsung.com>
Signed-off-by: C S Bhargava <cs.bhargava@samsung.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
11 years agoBluetooth: Add the definition and stcuture for Sync Train Complete
DoHyun Pyun [Wed, 2 Oct 2013 12:54:52 +0000 (21:54 +0900)]
Bluetooth: Add the definition and stcuture for Sync Train Complete

The Synchronization Train Complete event indicates that the Start
Synchronization Train command has completed.

The Core Spec Addendum 4 adds this command in part B Connectionless
Slave Broadcast.

Bluetooth Core Specification Addendum 4 - Page 103

"7.7.67 Synchronization Train Complete Event [New Section]
...

Event Parameters:

Status 0x00       Start Synchronization Train command completed
                  successfully.
       0x01-0xFF  Start Synchronization Train command failed.
                  See Part D, Error Codes, for error codes and
                  descriptions."

Signed-off-by: Dohyun Pyun <dh79.pyun@samsung.com>
Signed-off-by: C S Bhargava <cs.bhargava@samsung.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
11 years agoBluetooth: Add the definition for Start Synchronization Train
DoHyun Pyun [Wed, 2 Oct 2013 12:54:51 +0000 (21:54 +0900)]
Bluetooth: Add the definition for Start Synchronization Train

The Start_Synchronization_Train command controls the Synchronization
Train functionality in the BR/EDR Controller.

The Core Spec Addendum 4 adds this command in part B Connectionless
Slave Broadcast.

Bluetooth Core Specification Addendum 4 - Page 86

"7.1.51 Start Synchronization Train Command [New Section]
...
If connectionless slave broadcast mode is not enabled, the Command
Disallowed (0x0C) error code shall be returned. After receiving this
command and returning a Command Status event, the Baseband starts
attempting to send synchronization train packets containing information
related to the enabled Connectionless Slave Broadcast packet timing.

Note: The AFH_Channel_Map used in the synchronization train packets is
configured by the Set_AFH_Channel_Classification command and the local
channel classification in the BR/EDR Controller.

The synchronization train packets will be sent using the parameters
specified by the latest Write_Synchronization_Train_Parameters command.
The Synchronization Train will continue until synchronization_trainTO
slots (as specified in the last Write_Synchronization_Train command)
have passed or until the Host disables the Connectionless Slave Broadcast
logical transport."

Signed-off-by: Dohyun Pyun <dh79.pyun@samsung.com>
Signed-off-by: C S Bhargava <cs.bhargava@samsung.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
11 years agoBluetooth: Add the definition and structure for Set CSB
DoHyun Pyun [Wed, 2 Oct 2013 12:54:50 +0000 (21:54 +0900)]
Bluetooth: Add the definition and structure for Set CSB

he Set_Connectionless_Slave_Broadcast command controls the
Connectionless Slave Broadcast functionality in the BR/EDR
Controller.

The Core Spec Addendum 4 adds this command in part B Connectionless
Slave Broadcast.

Bluetooth Core Specification Addendum 4 - Page 78

"7.1.49 Set Connectionless Slave Broadcast Command [New Section]
...
The LT_ADDR indicated in the Set_Connectionless_Slave_Broadcast shall be
pre-allocated using the HCI_Set_Reserved_LT_ADDR command. If the
LT_ADDR has not been reserved, the Unknown Connection Identifier (0x02)
error code shall be returned. If the controller is unable to reserve
sufficient bandwidth for the requested activity, the Connection Rejected
Due to Limited Resources (0x0D) error code shall be returned.

The LPO_Allowed parameter informs the BR/EDR Controller whether it is
allowed to sleep.

The Packet_Type parameter specifies which packet types are allowed. The
Host shall either enable BR packet types only, or shall enable EDR and DM1
packet types only.

The Interval_Min and Interval_Max parameters specify the range from which
the BR/EDR Controller must select the Connectionless Slave Broadcast
Interval. The selected Interval is returned."

Signed-off-by: Dohyun Pyun <dh79.pyun@samsung.com>
Signed-off-by: C S Bhargava <cs.bhargava@samsung.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
11 years agoBluetooth: Add the structure for Write Sync Train Parameters
DoHyun Pyun [Wed, 2 Oct 2013 12:54:49 +0000 (21:54 +0900)]
Bluetooth: Add the structure for Write Sync Train Parameters

The Write_Synchronization_Train_Parameters command configures
the Synchronization Train functionality in the BR/EDR Controller.

The Core Spec Addendum 4 adds this command in part B Connectionless
Slave Broadcast.

Bluetooth Core Specification Addendum 4 - Page 97

"7.3.90 Write Synchronization Train Parameters Command [New Section]
...
Note: The AFH_Channel_Map used in the Synchronization Train packets is
configured by the Set_AFH_Channel_Classification command and the local
channel classification in the BR/EDR Controller.

Interval_Min and Interval_Max specify the allowed range of
Sync_Train_Interval. Refer to [Vol. 2], Part B, section 2.7.2 for
a detailed description of Sync_Train_Interval. The BR/EDR Controller shall
select an interval from this range and return it in Sync_Train_Interval.
If the Controller is unable to select a value from this range, it shall
return the Invalid HCI Command Parameters (0x12) error code.

Once started (via the Start_Synchronization_Train Command) the
Synchronization Train will continue until synchronization_trainTO slots have
passed or Connectionless Slave Broadcast has been disabled."

Signed-off-by: Dohyun Pyun <dh79.pyun@samsung.com>
Signed-off-by: C S Bhargava <cs.bhargava@samsung.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
11 years agoBluetooth: Add the definition and structure for Set CSB Data
DoHyun Pyun [Wed, 2 Oct 2013 12:54:48 +0000 (21:54 +0900)]
Bluetooth: Add the definition and structure for Set CSB Data

The Set_Connectionless_Slave_Broadcast_Data command provides the
ability for the Host to set Connectionless Slave Broadcast data in
the BR/EDR Controller.

The Core Spec Addendum 4 adds this command in part B Connectionless
Slave Broadcast.

Bluetooth Core Specification Addendum 4 - Page 93

"7.3.88 Set Connectionless Slave Broadcast Data Command [New Section]
...
If connectionless slave broadcast mode is disabled, this data shall be
kept by the BR/EDR Controller and used once connectionless slave broadcast
mode is enabled. If connectionless slave broadcast mode is enabled,
and this command is successful, this data will be sent starting with
the next Connectionless Slave Broadcast instant.

The Data_Length field may be zero, in which case no data needs to be
provided.

The Host may fragment the data using the Fragment field in the command. If
the combined length of the fragments exceeds the capacity of the largest
allowed packet size specified in the Set Connectionless Slave Broadcast
command, all fragments associated with the data being assembled shall be
discarded and the Invalid HCI Command Parameters error (0x12) shall be
returned."

Signed-off-by: Dohyun Pyun <dh79.pyun@samsung.com>
Signed-off-by: C S Bhargava <cs.bhargava@samsung.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
11 years agoBluetooth: Add the definition and structure for Delete Reserved LT_ADDR
DoHyun Pyun [Wed, 2 Oct 2013 12:54:47 +0000 (21:54 +0900)]
Bluetooth: Add the definition and structure for Delete Reserved LT_ADDR

The Delete_Reserved_LT_ADDR command requests that the BR/EDR
Controller cancel the reservation for a specific LT_ADDR reserved for the
purposes of Connectionless Slave Broadcast.

The Core Spec Addendum 4 adds this command in part B Connectionless
Slave Broadcast.

Bluetooth Core Specification Addendum 4 - Page 92

"7.3.87 Delete Reserved LT_ADDR Command [New Section]
...
If the LT_ADDR indicated in the LT_ADDR parameter is not reserved by the
BR/EDR Controller, it shall return the Unknown Connection Identifier (0x02)
error code.
If connectionless slave broadcast mode is still active, then the Controller
shall return the Command Disallowed (0x0C) error code."

Signed-off-by: Dohyun Pyun <dh79.pyun@samsung.com>
Signed-off-by: C S Bhargava <cs.bhargava@samsung.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
11 years agoBluetooth: Add the definition and structure for Set Reserved LT_ADDR
DoHyun Pyun [Wed, 2 Oct 2013 12:54:46 +0000 (21:54 +0900)]
Bluetooth: Add the definition and structure for Set Reserved LT_ADDR

The Set_Reserved_LT_ADDR command allows the host to request that the
BR/EDR Controller reserve a specific LT_ADDR for Connectionless Slave
Broadcast.

The Core Spec Addendum 4 adds this command in part B Connectionless
Slave Broadcast.

Bluetooth Core Specification Addendum 4 - Page 90

"7.3.86 Set Reserved LT_ADDR Command [New Section]
...
If the LT_ADDR indicated in the LT_ADDR parameter is already in use by the
BR/EDR Controller, it shall return the ACL Connection Already Exists (0x0B)
error code. If the LT_ADDR indicated in the LT_ADDR parameter is out of
range, the controller shall return the Invalid HCI Command Parameters (0x12)
error code. If the command succeeds, then the reserved LT_ADDR shall be
used when issuing subsequent Set Connectionless Slave Broadcast Data and
Set Connectionless Slave Broadcast commands.
To ensure that the reserved LT_ADDR is not already allocated, it is
recommended that this command be issued at some point after HCI_Reset is
issued but before page scanning is enabled or paging is initiated."

Signed-off-by: Dohyun Pyun <dh79.pyun@samsung.com>
Signed-off-by: C S Bhargava <cs.bhargava@samsung.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
11 years agoiwlwifi: pcie: fix merge damage
Johannes Berg [Mon, 30 Sep 2013 09:02:46 +0000 (11:02 +0200)]
iwlwifi: pcie: fix merge damage

The merge b35c8097 seems to have lost commit eabc4ac5d,
put the code back.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoBluetooth: Increment management interface revision
Marcel Holtmann [Wed, 2 Oct 2013 12:18:31 +0000 (05:18 -0700)]
Bluetooth: Increment management interface revision

This patch increments the management interface revision due to the
various fixes, improvements and other changes that have gone in
lately.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
11 years agoBluetooth: Fix advertising data flags with disabled BR/EDR
Johan Hedberg [Wed, 2 Oct 2013 13:02:24 +0000 (16:02 +0300)]
Bluetooth: Fix advertising data flags with disabled BR/EDR

We shouldn't include the simultaneous LE & BR/EDR flags in the LE
advertising data if BR/EDR is disabled on a dual-mode controller. This
patch fixes this issue and ensures that the create_ad function generates
the correct flags when BR/EDR is disabled.

Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
11 years agoBluetooth: Fix REJECTED vs NOT_SUPPORTED mgmt responses
Johan Hedberg [Wed, 2 Oct 2013 12:45:22 +0000 (15:45 +0300)]
Bluetooth: Fix REJECTED vs NOT_SUPPORTED mgmt responses

The REJECTED management response should mainly be used when the adapter
is in a state where we cannot accept some command or a specific
parameter value. The NOT_SUPPORTED response in turn means that the
adapter really cannot support the command or parameter value.

This patch fixes this distinction and adds two helper functions to
easily get the appropriate LE or BR/EDR related status response.

Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
11 years agoBluetooth: Add management command for setting static address
Marcel Holtmann [Wed, 2 Oct 2013 11:41:30 +0000 (04:41 -0700)]
Bluetooth: Add management command for setting static address

On dual-mode BR/EDR/LE and LE only controllers it is possible
to configure a random address. There are two types or random
addresses, one is static and the other private. Since the
random private addresses require special privacy feature to
be supported, the configuration of these two are kept separate.

This command allows for setting the static random address. It is
only supported on controllers with LE support. The static random
address is suppose to be valid for the lifetime of the controller
or at least until the next power cycle. To ensure such behavior,
setting of the address is limited to when the controller is
powered off.

The special BDADDR_ANY address (00:00:00:00:00:00) can be used to
disable the static address. This is also the default value.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
11 years agoBluetooth: Restrict disabling of HS when controller is powered off
Marcel Holtmann [Wed, 2 Oct 2013 07:27:02 +0000 (00:27 -0700)]
Bluetooth: Restrict disabling of HS when controller is powered off

Disabling the high speed setting when the controller is powered on has
too many side effects that are not taken care of. And in general it
is not an useful operation anyway. So just make such a command fail
with a rejection error message.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
11 years agoBluetooth: Add a new mgmt_set_bredr command
Johan Hedberg [Wed, 2 Oct 2013 10:43:14 +0000 (13:43 +0300)]
Bluetooth: Add a new mgmt_set_bredr command

This patch introduces a new mgmt command for enabling/disabling BR/EDR
functionality. This can be convenient when one wants to make a dual-mode
controller behave like a single-mode one. The command is only available
for dual-mode controllers and requires that LE is enabled before using
it. The BR/EDR setting can be enabled at any point, however disabling it
requires the controller to be powered off (otherwise a "rejected"
response will be sent).

Disabling the BR/EDR setting will automatically disable all other BR/EDR
related settings.

Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
11 years agoBluetooth: Introduce a new HCI_BREDR_ENABLED flag
Johan Hedberg [Wed, 2 Oct 2013 10:43:13 +0000 (13:43 +0300)]
Bluetooth: Introduce a new HCI_BREDR_ENABLED flag

To allow treating dual-mode (BR/EDR/LE) controllers as single-mode ones
(LE-only) we want to introduce a new HCI_BREDR_ENABLED flag to track
whether BR/EDR is enabled or not (previously we simply looked at the
feature bit with lmp_bredr_enabled).

This patch add the new flag and updates the relevant places to test
against it instead of using lmp_bredr_enabled. The flag is by default
enabled when registering an adapter and only cleared if necessary once
the local features have been read during the HCI init procedure.

We cannot completely block BR/EDR usage in case user space uses raw HCI
sockets but the patch tries to block this in places where possible, such
as the various BR/EDR specific ioctls.

Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
11 years agoBluetooth: btmrvl: add calibration data download support
Amitkumar Karwar [Tue, 1 Oct 2013 19:19:15 +0000 (12:19 -0700)]
Bluetooth: btmrvl: add calibration data download support

A text file containing calibration data in hex format can
be provided at following path:

/lib/firmware/mrvl/sd8797_caldata.conf

The data will be downloaded to firmware during initialization.

Reviewed-by: Mike Frysinger <vapier@chromium.org>
Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: Hyuckjoo Lee <hyuckjoo.lee@samsung.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
11 years agoBluetooth: btmrvl: add setup handler
Amitkumar Karwar [Tue, 1 Oct 2013 19:19:14 +0000 (12:19 -0700)]
Bluetooth: btmrvl: add setup handler

Move initialization code to hdev's setup handler.

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
11 years agoBluetooth: btmrvl: get rid of struct btmrvl_cmd
Amitkumar Karwar [Tue, 1 Oct 2013 19:19:13 +0000 (12:19 -0700)]
Bluetooth: btmrvl: get rid of struct btmrvl_cmd

Replace this proprietary structure with the standard one
(struct hci_command_hdr).

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
11 years agoBluetooth: btmrvl: add btmrvl_send_sync_cmd() function
Amitkumar Karwar [Tue, 1 Oct 2013 19:19:12 +0000 (12:19 -0700)]
Bluetooth: btmrvl: add btmrvl_send_sync_cmd() function

Command preparation code is used multiple times. This patch
separate out this common code and create btmrvl_send_sync_cmd()
function.

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
11 years agoBluetooth: Fix workqueue synchronization in hci_dev_open
Johan Hedberg [Tue, 1 Oct 2013 19:44:50 +0000 (22:44 +0300)]
Bluetooth: Fix workqueue synchronization in hci_dev_open

When hci_sock.c calls hci_dev_open it needs to ensure that there isn't
pending work in progress, such as that which is scheduled for the
initial setup procedure or the one for automatically powering off after
the setup procedure. This adds the necessary calls to ensure that any
previously scheduled work is completed before attempting to call
hci_dev_do_open.

This patch fixes a race with old user space versions where we might
receive a HCIDEVUP ioctl before the setup procedure has been completed.
When that happens the setup procedures callback may fail early and leave
the device in an inconsistent state, causing e.g. the setup callback to
be (incorrectly) called more than once.

Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>