]> git.openfabrics.org - ~shefty/rdma-dev.git/log
~shefty/rdma-dev.git
9 years agonet: fix a wrong assignment in skb_split()
Amerigo Wang [Tue, 19 Feb 2013 22:51:30 +0000 (22:51 +0000)]
net: fix a wrong assignment in skb_split()

commit c9af6db4c11ccc6c3e7f1 (net: Fix possible wrong checksum generation)
has a suspicous piece:

-       skb_shinfo(skb1)->gso_type = skb_shinfo(skb)->gso_type;
-
+       skb_shinfo(skb)->tx_flags = skb_shinfo(skb1)->tx_flags & SKBTX_SHARED_FRAG;

skb1 is the new skb, therefore should be on the left side of the assignment.
This patch fixes it.

Cc: Pravin B Shelar <pshelar@nicira.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <amwang@redhat.com>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'for-davem' of git://git.kernel.org/pub/scm/linux/kernel/git/linville...
David S. Miller [Wed, 20 Feb 2013 03:24:50 +0000 (22:24 -0500)]
Merge branch 'for-davem' of git://git./linux/kernel/git/linville/wireless-next

John W. Linville says:

====================
One last batch of stragglers intended for 3.9...

For the iwlwifi pull, Johannes says:

"I hadn't expected to ask you to pull iwlwifi-next again, but I have a
number of fixes most of which I'd also send in after rc1, so here it is.

The first commit is a merge error between mac80211-next and
iwlwifi-next; in addition I have fixes for P2P scanning and MVM driver
MAC (virtual interface) management from Ilan, a CT-kill (critical
temperature) fix from Eytan, and myself fixed three different little but
annoying bugs in the MVM driver.

The only ones I might not send for -rc1 are Emmanuel's debug patch, but
OTOH it should help greatly if there are any issues, and my own time
event debugging patch that I used to find the race condition but we
decided to keep it for the future."

For the mac80211 pull, Johannes says:

"Like iwlwifi-next, this would almost be suitable for rc1.

I have a fix for station management on non-TDLS drivers, a CAB queue
crash fix for mesh, a fix for an annoying (but harmless) warning, a
tracing fix and a documentation fix. Other than that, only a few mesh
cleanups."

Along with that is a fix for memory corruption in rtlwifi, an
orinoco_usb fix to avoid allocating a DMA buffer on the stack, an a
hostap fix to return -ENOMEM instead of -1 after a memory allocation
failure.  The remaining bits implement 802.11ac support for the mwifiex
driver -- I think that is still worth getting into 3.9.

Please let me know if there are problems!
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoip_gre: remove an extra dst_release()
Eric Dumazet [Tue, 19 Feb 2013 12:14:10 +0000 (12:14 +0000)]
ip_gre: remove an extra dst_release()

commit 68c331631143 (v4 GRE: Add TCP segmentation offload for GRE)
introduced a bug in error path.

dst is attached to skb, so will be released when skb is freed.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Pravin B Shelar <pshelar@nicira.com>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
John W. Linville [Tue, 19 Feb 2013 19:56:34 +0000 (14:56 -0500)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-next into for-davem

9 years agoppp: set qdisc_tx_busylock to avoid LOCKDEP splat
Eric Dumazet [Tue, 19 Feb 2013 18:42:03 +0000 (10:42 -0800)]
ppp: set qdisc_tx_busylock to avoid LOCKDEP splat

If a qdisc is installed on a ppp device, its possible to get
a lockdep splat under stress, because nested dev_queue_xmit() can
lock busylock a second time (on a different device, so its a false
positive)

Avoid this problem using a distinct lock_class_key for ppp
devices.

Reported-by: Yanko Kaneti <yaneti@declera.com>
Tested-by: Yanko Kaneti <yaneti@declera.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoatl1c: restore buffer state
Huang, Xiong [Tue, 19 Feb 2013 07:23:09 +0000 (07:23 +0000)]
atl1c: restore buffer state

in the previous commit : f1f220ea1dda078, the BUSY state of buffer is wrongly
deleted. this patch just restore it.

Signed-off-by: xiong <xiong@qca.qualcomm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fix a build failure when !CONFIG_PROC_FS
Cong Wang [Tue, 19 Feb 2013 02:47:05 +0000 (02:47 +0000)]
net: fix a build failure when !CONFIG_PROC_FS

When !CONFIG_PROC_FS dev_mcast_init() is not defined,
actually we can just merge dev_mcast_init() into
dev_proc_init().

Reported-by: Gao feng <gaofeng@cn.fujitsu.com>
Cc: Gao feng <gaofeng@cn.fujitsu.com>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: ipv4: fix waring -Wunused-variable
Gao feng [Tue, 19 Feb 2013 00:43:12 +0000 (00:43 +0000)]
net: ipv4: fix waring -Wunused-variable

the vars ip_rt_gc_timeout is used only when
CONFIG_SYSCTL is selected.

move these vars into CONFIG_SYSCTL.

Signed-off-by: Gao feng <gaofeng@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: proc: fix build failed when procfs is not configured
Gao feng [Tue, 19 Feb 2013 00:43:10 +0000 (00:43 +0000)]
net: proc: fix build failed when procfs is not configured

commit d4beaa66add8aebf83ab16d2fde4e4de8dac36df
"net: proc: change proc_net_fops_create to proc_create"
uses proc_create to replace proc_net_fops_create, when
CONFIG_PROC isn't configured, some build error will
occurs.

net/packet/af_packet.c: In function 'packet_net_init':
net/packet/af_packet.c:3831:48: error: 'packet_seq_fops' undeclared (first use in this function)
net/packet/af_packet.c:3831:48: note: each undeclared identifier is reported only once for each function it appears in

There may be other build fails like above,this patch
change proc_create from function to macros when CONFIG_PROC
is not configured,just like what proc_net_fops_create did
before this commit.

Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Gao feng <gaofeng@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoRevert "xen: netback: remove redundant xenvif_put"
David S. Miller [Tue, 19 Feb 2013 18:04:34 +0000 (13:04 -0500)]
Revert "xen: netback: remove redundant xenvif_put"

This reverts commit d37204566a61d5116d385ae909db8e14a734b30f.

This change is incorrect, as per Jan Beulich:

====================
But this is wrong from all we can tell, we discussed this before
(Wei pointed to the discussion in an earlier reply). The core of
it is that the put here parallels the one in netbk_tx_err(), and
the one in xenvif_carrier_off() matches the get from
xenvif_connect() (which normally would be done on the path
coming through xenvif_disconnect()).
====================

And a previous discussion of this issue is at:

http://marc.info/?l=xen-devel&m=136084174026977&w=2

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: move procfs code to net/core/net-procfs.c
Cong Wang [Mon, 18 Feb 2013 19:20:33 +0000 (19:20 +0000)]
net: move procfs code to net/core/net-procfs.c

Similar to net/core/net-sysfs.c, group procfs code to
a single unit.

Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoqmi_wwan, cdc-ether: add ADU960S
Dan Williams [Mon, 18 Feb 2013 17:25:09 +0000 (17:25 +0000)]
qmi_wwan, cdc-ether: add ADU960S

It advertises a standard CDC-ETHER interface, which actually should be
driven by qmi_wwan.

Signed-off-by: Dan Williams <dcbw@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobonding: set sysfs device_type to 'bond'
Doug Goldstein [Mon, 18 Feb 2013 14:59:23 +0000 (14:59 +0000)]
bonding: set sysfs device_type to 'bond'

Sets the sysfs device_type to 'bond' for udev. This allows udev rules to
be created for bond devices. This is similar to how other network
devices set their device_type.

Signed-off-by: Doug Goldstein <cardoe@cardoe.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobonding: fix bond_release_all inconsistencies
nikolay@redhat.com [Mon, 18 Feb 2013 14:09:42 +0000 (14:09 +0000)]
bonding: fix bond_release_all inconsistencies

This patch fixes the following inconsistencies in bond_release_all:
- IFF_BONDING flag is not stripped from slaves
- MTU is not restored
- no netdev notifiers are sent
Instead of trying to keep bond_release and bond_release_all in sync
I think we can re-use bond_release as the environment for calling it
is correct (RTNL is held). I have been running tests for the past
week and they came out successful. The only way for bond_release to fail
is for the slave to be attached in a different bond or to not be a slave
but that cannot happen as RTNL is held and no slave manipulations can be
achieved.

V2: As suggested bond_release is renamed to __bond_release_one with a
new parameter "all" introduced so to avoid calling unnecessary code while
destroying a bond, and a wrapper for it called bond_release is created
because of ndo_del_link. bond_release_all() is removed.

Signed-off-by: Nikolay Aleksandrov <nikolay@redhat.com>
Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agob44: use netdev_alloc_skb_ip_align()
Hauke Mehrtens [Mon, 18 Feb 2013 10:49:56 +0000 (10:49 +0000)]
b44: use netdev_alloc_skb_ip_align()

Without this patch b44 always allocates the 2 bytes needed for aligned
access on every platform, now it uses netdev_alloc_skb_ip_align().

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoxen: netback: remove redundant xenvif_put
Andrew Jones [Mon, 18 Feb 2013 10:29:20 +0000 (10:29 +0000)]
xen: netback: remove redundant xenvif_put

netbk_fatal_tx_err() calls xenvif_carrier_off(), which does
a xenvif_put(). As callers of netbk_fatal_tx_err should only
have one reference to the vif at this time, then the xenvif_put
in netbk_fatal_tx_err is one too many.

Signed-off-by: Andrew Jones <drjones@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: Do a sanity check on the gpio number
Fabio Estevam [Mon, 18 Feb 2013 10:20:31 +0000 (10:20 +0000)]
net: fec: Do a sanity check on the gpio number

Check whether the phy-reset GPIO is valid, prior to requesting it.

In the case a board does not provide a phy-reset GPIO, just returns immediately.

With such gpio validation in place, it is also safe to change from pr_debug to
dev_err in the case the gpio request fails.

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoip_gre: propogate target device GSO capability to the tunnel device
Dmitry Kravkov [Mon, 18 Feb 2013 09:50:53 +0000 (09:50 +0000)]
ip_gre: propogate target device GSO capability to the tunnel device

Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoip_gre: allow CSUM capable devices to handle packets
Dmitry Kravkov [Mon, 18 Feb 2013 09:50:52 +0000 (09:50 +0000)]
ip_gre: allow CSUM capable devices to handle packets

If device is not able to handle checksumming it will
be handled in dev_xmit

Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobonding: Fix initialize after use for 3ad machine state spinlock
nikolay@redhat.com [Mon, 18 Feb 2013 07:59:03 +0000 (07:59 +0000)]
bonding: Fix initialize after use for 3ad machine state spinlock

The 3ad machine state spinlock can be used before it is inititialized
while doing bond_enslave() (and the port is being initialized) since
port->slave is set before the lock is prepared, thus causing soft
lock-ups and a multitude of other nasty bugs.

[ Rename __initialize_port_locks() variable name to 'slave' -DaveM ]

Signed-off-by: Nikolay Aleksandrov <nikolay@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobonding: Fix race condition between bond_enslave() and bond_3ad_update_lacp_rate()
nikolay@redhat.com [Mon, 18 Feb 2013 07:59:02 +0000 (07:59 +0000)]
bonding: Fix race condition between bond_enslave() and bond_3ad_update_lacp_rate()

port->slave can be NULL since it's being initialized in bond_enslave
thus dereferencing a NULL pointer in bond_3ad_update_lacp_rate()
Also fix a minor bug, which could cause a port not to have
AD_STATE_LACP_TIMEOUT since there's no sync between
bond_3ad_update_lacp_rate() and bond_3ad_bind_slave(), by changing
the read_lock to a write_lock_bh in bond_3ad_update_lacp_rate().

Signed-off-by: Nikolay Aleksandrov <nikolay@redhat.com>
Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agob43: Increase number of RX DMA slots
Larry Finger [Sun, 17 Feb 2013 17:01:20 +0000 (17:01 +0000)]
b43: Increase number of RX DMA slots

Bastian Bittorf reported that some of the silent freezes on a Linksys WRT54G
were due to overflow of the RX DMA ring buffer, which was created with 64
slots. That finding reminded me that I was seeing similar crashed on a netbook,
which also has a relatively slow processor. After increasing the number of
slots to 128, runs on the netbook that previously failed now worked; however,
I found that 109 slots had been used in one test. For that reason, the number
of slots is being increased to 256.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Bastian Bittorf <bittorf@bluebottle.com>
Cc: Stable <stable@vger.kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoqlcnic: Bump up the version to 5.1.34
Jitendra Kalsaria [Mon, 18 Feb 2013 12:06:18 +0000 (12:06 +0000)]
qlcnic: Bump up the version to 5.1.34

Signed-off-by: Jitendra Kalsaria <jitendra.kalsaria@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoqlcnic: fix ping resumption to a VM after a live migration
Shahed Shaikh [Mon, 18 Feb 2013 13:22:37 +0000 (13:22 +0000)]
qlcnic: fix ping resumption to a VM after a live migration

Delete the MAC address of a VM, from the adapter's embedded switch,
after the VM had been migrated out of this adapter/server.

Signed-off-by: Shahed Shaikh <shahed.shaikh@qlogic.com>
Signed-off-by: Rajesh Borundia <rajesh.borundia@qlogic.com>
Signed-off-by: Jitendra Kalsaria <jitendra.kalsaria@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoqlcnic: fix estimation of receive MSS in case of LRO for 83xx adapter
Shahed Shaikh [Mon, 18 Feb 2013 12:06:16 +0000 (12:06 +0000)]
qlcnic: fix estimation of receive MSS in case of LRO for 83xx adapter

Set gso_size to MSS obtained from adapter to avoid incorrect estimation
of receive MSS, which would lead to delayed ACKs in some traffic patterns

Example:
Send two or three packets and wait for ack and only then send
remaining packets.

Signed-off-by: Shahed Shaikh <shahed.shaikh@qlogic.com>
Signed-off-by: Jitendra Kalsaria <jitendra.kalsaria@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoqlcnic: refactor Legacy interrupt handling for 83xx
Himanshu Madhani [Mon, 18 Feb 2013 12:06:15 +0000 (12:06 +0000)]
qlcnic: refactor Legacy interrupt handling for 83xx

o Cleanly seperate 83xx Legacy interrupt handling code from 82xx
o Update 83xx Legacy interrupt handling code to match with the spec

Signed-off-by: Himanshu Madhani <himanshu.madhani@qlogic.com>
Signed-off-by: Jitendra Kalsaria <jitendra.kalsaria@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoqlcnic: fix mailbox interrupt.
Sucheta Chakraborty [Mon, 18 Feb 2013 12:06:14 +0000 (12:06 +0000)]
qlcnic: fix mailbox interrupt.

o Config interrupt is not needed for mailbox interrupts.

Signed-off-by: Sucheta Chakraborty <sucheta.chakraborty@qlogic.com>
Signed-off-by: Jitendra Kalsaria <jitendra.kalsaria@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoqlcnic: fix unsupported CDRP command error message.
Jitendra Kalsaria [Mon, 18 Feb 2013 12:06:13 +0000 (12:06 +0000)]
qlcnic: fix unsupported CDRP command error message.

Added detailed error messages for FW CDRP command failure

Signed-off-by: Jitendra Kalsaria <jitendra.kalsaria@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'master' of git://1984.lsi.us.es/nf-next
David S. Miller [Tue, 19 Feb 2013 04:42:09 +0000 (23:42 -0500)]
Merge branch 'master' of git://1984.lsi.us.es/nf-next

Pablo Neira Ayuso says:

====================
The following patchset contain updates for your net-next tree, they are:

* Fix (for just added) connlabel dependencies, from Florian Westphal.

* Add aliasing support for conntrack, thus users can either use -m state
  or -m conntrack from iptables while using the same kernel module, from
  Jozsef Kadlecsik.

* Some code refactoring for the CT target to merge common code in
  revision 0 and 1, from myself.

* Add aliasing support for CT, based on patch from Jozsef Kadlecsik.

* Add one mutex per nfnetlink subsystem, from myself.

* Improved logging for packets that are dropped by helpers, from myself.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net into net
David S. Miller [Tue, 19 Feb 2013 04:32:49 +0000 (23:32 -0500)]
Merge git://git./linux/kernel/git/davem/net into net

Pull in 'net' to take in the bug fixes that didn't make it into
3.8-final.

Also, deal with the semantic conflict of the change made to
net/ipv6/xfrm6_policy.c   A missing rt6->n neighbour release
was added to 'net', but in 'net-next' we no longer cache the
neighbour entries in the ipv6 routes so that change is not
appropriate there.

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonetfilter: nf_ct_helper: better logging for dropped packets
Pablo Neira Ayuso [Sun, 10 Feb 2013 17:56:56 +0000 (18:56 +0100)]
netfilter: nf_ct_helper: better logging for dropped packets

Connection tracking helpers have to drop packets under exceptional
situations. Currently, the user gets the following logging message
in case that happens:

nf_ct_%s: dropping packet ...

However, depending on the helper, there are different reasons why a
packet can be dropped.

This patch modifies the existing code to provide more specific
error message in the scope of each helper to help users to debug
the reason why the packet has been dropped, ie:

nf_ct_%s: dropping packet: reason ...

Thanks to Joe Perches for many formatting suggestions.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
9 years agonet: wireless: hostap: hostap_ap.c: Return -ENOMEM instead of -1 for if kmalloc(...
Kumar Amit Mehta [Mon, 18 Feb 2013 12:13:44 +0000 (04:13 -0800)]
net: wireless: hostap: hostap_ap.c: Return -ENOMEM instead of -1 for if kmalloc() fails.

When memory allocation using, kmalloc() fails, report appropriate error value.

Signed-off-by: Kumar Amit Mehta <gmate.amit@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
9 years agonet: wireless: orinoco: orinoco_usb.c: fix DMA buffers on stack
Kumar Amit Mehta [Mon, 18 Feb 2013 11:37:44 +0000 (03:37 -0800)]
net: wireless: orinoco: orinoco_usb.c: fix DMA buffers on stack

This patch fixes an instance of DMA buffer on stack(being passed to
usb_control_msg) for the wireless USB version of the Agere Orinoco card driver.
It also fixes the missing audit for the return value of firmware download
routine. Found using smatch.

Signed-off-by: Kumar Amit Mehta <gmate.amit@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
9 years agortlwifi: usb: allocate URB control message setup_packet and data buffer separately
Jussi Kivilinna [Mon, 18 Feb 2013 08:29:30 +0000 (10:29 +0200)]
rtlwifi: usb: allocate URB control message setup_packet and data buffer separately

rtlwifi allocates both setup_packet and data buffer of control message urb,
using shared kmalloc in _usbctrl_vendorreq_async_write. Structure used for
allocating is:
struct {
u8 data[254];
struct usb_ctrlrequest dr;
};

Because 'struct usb_ctrlrequest' is __packed, setup packet is unaligned and
DMA mapping of both 'data' and 'dr' confuses ARM/sunxi, leading to memory
corruptions and freezes.

Patch changes setup packet to be allocated separately.

[v2]:
 - Use WARN_ON_ONCE instead of WARN_ON

Cc: <stable@vger.kernel.org>
Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
9 years agomwifiex: add 802.11AC support
Yogesh Ashok Powar [Sat, 16 Feb 2013 05:44:30 +0000 (21:44 -0800)]
mwifiex: add 802.11AC support

For STA mode, collect VHT realated IEs from the Beacons or Probe
Responses and append similar VHT related IEs to association
requests.

For AP mode, get VHT related capability information and share it
with cfg80211 at the time of wiphy register. This information is
further used by cfg80211 and hostapd to start an AP with 802.11AC
support.

Currently only 8897 supports 802.11AC.

Signed-off-by: Yogesh Ashok Powar <yogeshp@marvell.com>
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Nishant Sarmukadam <nishants@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: Frank Huang <frankh@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
9 years agomwifiex: coding style cleanup in bss parameter update
Yogesh Ashok Powar [Sat, 16 Feb 2013 05:44:29 +0000 (21:44 -0800)]
mwifiex: coding style cleanup in bss parameter update

They all can make one line.

Signed-off-by: Yogesh Ashok Powar <yogeshp@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
9 years agomwifiex: fix empty TX ring check for PCIe8897 while unloading driver
Avinash Patil [Sat, 16 Feb 2013 05:37:54 +0000 (21:37 -0800)]
mwifiex: fix empty TX ring check for PCIe8897 while unloading driver

While unloading driver, we free all pending TX packets by flushing
TX ring. There is unhandled case for PCIE8897 while checking for
ring empty condition.

This patch adds the handling by calling mwifiex_pcie_txbd_empty().

Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Yogesh Ashok Powar <yogeshp@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
9 years agoipv6: fix a sparse warning
Eric Dumazet [Mon, 18 Feb 2013 10:18:52 +0000 (10:18 +0000)]
ipv6: fix a sparse warning

net/ipv6/reassembly.c:82:72: warning: incorrect type in argument 3 (different base types)
net/ipv6/reassembly.c:82:72:    expected unsigned int [unsigned] [usertype] c
net/ipv6/reassembly.c:82:72:    got restricted __be32 [usertype] id

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'for-john' of git://git.kernel.org/pub/scm/linux/kernel/git/jberg/mac802...
John W. Linville [Mon, 18 Feb 2013 20:27:42 +0000 (15:27 -0500)]
Merge branch 'for-john' of git://git./linux/kernel/git/jberg/mac80211-next

9 years agoMerge branch 'for-davem' of git://git.kernel.org/pub/scm/linux/kernel/git/linville...
David S. Miller [Mon, 18 Feb 2013 20:12:07 +0000 (15:12 -0500)]
Merge branch 'for-davem' of git://git./linux/kernel/git/linville/wireless-next

John W. Linville says:

====================
This probably is the last big pull request for wireless bits
for 3.9.  Of course, I'm sure there will be a few stragglers here
and there...surely a few bug fixes as well... :-) (In fact, I see
that Johannes has already queued-up a few more for me while I was
preparing this...)

Included are a number of pulls...

For mac80211-next, Johannes says:

"The biggest change I have is undoubtedly Marco's mesh powersave
implementation. Beyond that, I have a patch from Emmanuel to modify the
DTIM period API in mac80211, scan improvements and a removal of some
previous workaround code from Stanislaw, dynamic short slot time from
Thomas and 64-bit station byte counters from Vladimir. I also made a
number of changes myself, some related to WoWLAN, some auth/deauth
improvements and most of them BSS list cleanups."

"This time, I have relatively large number of fixes in various areas of
the code (a memory leak in regulatory, an RX race in mac80211, the new
radar checking caused a P2P device problem, some mesh issues with
stations, an older bug in tracing and for kernel-doc) as well as a
number of small new features. The biggest (in the diffstat) is my work
on hidden SSID tracking."

"Please pull to get
 * radar detection work from Simon
 * mesh improvements from Thomas
 * a connection monitoring/powersave fix from Wojciech
 * TDLS-related station management work from Jouni
 * VLAN crypto fixes from Michael Braun
 * CCK support in minstrel_ht from Felix
 * an SMPS (not SMSP, oops) related improvement in mac80211 (Emmanuel)
 * some WoWLAN work from Amitkumar Karwar: pattern match offset and a
   documentation fix
 * some WoWLAN work from myself (TCP connection wakeup feature API)
 * and a lot of VHT (and some HT) work (also from myself)

And a number of more random cleanups/fixes. I merged mac80211/master to
avoid a merge problem there."

And regarding iwlwifi-next, Johannes says:

"We continue work on our new driver, but I also have a WoWLAN and AP mode
improvement for the previous driver and a change to use threaded
interrupts to prepare us for working with non-PCIe devices."

Regarding wl12xx, Luca says:

"A few more patches intended for 3.9.  Mostly some clean-ups I've been
doing to make it easier to support device-tree.  Also including one bug
fix for wl12xx where the rates we advertise were wrong and an update in
the wlconf structure to support newer firmwares."

For the nfc-next bits, Samuel says:

"This is the second NFC pull request for 3.9.

We have:

- A few pn533 fixes on top of Waldemar refactorization of the driver, one of
  them fixes target mode.

- A new driver for Inside Secure microread chipset. It supports two
  physical layers: i2c and MEI. The MEI one depends on a patchset that's
  been sent to Greg Kroah-Hartman for inclusion into the 3.9 kernel [1]. The
  dependency is a KConfig one which means this code is not buildable as long
  as the MEI API is not usptream."

"This 3rd NFC pull request for 3.9 contains a fix for the microread MEI
physical layer support, as the MEI bus API changed.

From the MEI code, we now pass the MEI id back to the driver probe routine,
and we also pass a name and a MEI id table through the mei_bus_driver
structure. A few renames as well like e.g. mei_bus_driver to mei_driver or
mei_bus_client to mei_device in order to be closer to the driver model
practices."

For the ath6kl bits, Kalle says:

"There's not anything special here, most of the patches are just code
cleanup. The only functional changes are using the beacon interval from user
space and fixing a crash which happens when inserting and removing the
module in a loop."

Also, I pulled the wireless tree in order to resolve some pending
merge issues.  On top of that, there is a bunch of work on brcmfmac
that leads up to P2P support.  Also, mwifiex, rtlwifi, and a variety
of other drivers see some basic cleanups and minor enhancements.

Please let me know if there are problems!
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoVSOCK: Don't reject PF_VSOCK protocol
Andy King [Mon, 18 Feb 2013 06:04:13 +0000 (06:04 +0000)]
VSOCK: Don't reject PF_VSOCK protocol

Allow our own family as the protocol value for socket creation.

Reported-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Andy King <acking@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoVSOCK: Remove hypervisor-only socket option
Andy King [Mon, 18 Feb 2013 06:04:12 +0000 (06:04 +0000)]
VSOCK: Remove hypervisor-only socket option

Remove hypervisor-only socket option.

Reported-by: Gerd Hoffmann <kraxel@redhat.com>
Acked-by: Dmitry Torokhov <dtor@vmware.com>
Signed-off-by: Andy King <acking@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoVSOCK: get rid of vsock_version.h
Dmitry Torokhov [Mon, 18 Feb 2013 06:04:11 +0000 (06:04 +0000)]
VSOCK: get rid of vsock_version.h

There isn't really a need to have a separate file for it.

Acked-by: Andy King <acking@vmware.com>
Signed-off-by: Dmitry Torokhov <dtor@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoVSOCK: get rid of EXPORT_SYMTAB
Dmitry Torokhov [Mon, 18 Feb 2013 06:04:10 +0000 (06:04 +0000)]
VSOCK: get rid of EXPORT_SYMTAB

This is the default behavior for a looooooong time.

Acked-by: Andy King <acking@vmware.com>
Signed-off-by: Dmitry Torokhov <dtor@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoxfrm: release neighbor upon dst destruction
Romain KUNTZ [Mon, 18 Feb 2013 02:36:24 +0000 (02:36 +0000)]
xfrm: release neighbor upon dst destruction

Neighbor is cloned in xfrm6_fill_dst but seems to never be released.
Neighbor entry should be released when XFRM6 dst entry is destroyed
in xfrm6_dst_destroy, otherwise references may be kept forever on
the device pointed by the neighbor entry.

I may not have understood all the subtleties of XFRM & dst so I would
be happy to receive comments on this patch.

Signed-off-by: Romain Kuntz <r.kuntz@ipflavors.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: proc: remove proc_net_remove
Gao feng [Mon, 18 Feb 2013 01:34:57 +0000 (01:34 +0000)]
net: proc: remove proc_net_remove

proc_net_remove has been replaced by remove_proc_entry.
we can remove it now.

Signed-off-by: Gao feng <gaofeng@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: proc: change proc_net_remove to remove_proc_entry
Gao feng [Mon, 18 Feb 2013 01:34:56 +0000 (01:34 +0000)]
net: proc: change proc_net_remove to remove_proc_entry

proc_net_remove is only used to remove proc entries
that under /proc/net,it's not a general function for
removing proc entries of netns. if we want to remove
some proc entries which under /proc/net/stat/, we still
need to call remove_proc_entry.

this patch use remove_proc_entry to replace proc_net_remove.
we can remove proc_net_remove after this patch.

Signed-off-by: Gao feng <gaofeng@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: proc: remove proc_net_fops_create
Gao feng [Mon, 18 Feb 2013 01:34:55 +0000 (01:34 +0000)]
net: proc: remove proc_net_fops_create

proc_net_fops_create has been replaced by proc_create,
we can remove it now.

Signed-off-by: Gao feng <gaofeng@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: proc: change proc_net_fops_create to proc_create
Gao feng [Mon, 18 Feb 2013 01:34:54 +0000 (01:34 +0000)]
net: proc: change proc_net_fops_create to proc_create

Right now, some modules such as bonding use proc_create
to create proc entries under /proc/net/, and other modules
such as ipv4 use proc_net_fops_create.

It looks a little chaos.this patch changes all of
proc_net_fops_create to proc_create. we can remove
proc_net_fops_create after this patch.

Signed-off-by: Gao feng <gaofeng@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agodrivers: net: davinci_cpdma: acknowledge interrupt properly
Mugunthan V N [Sun, 17 Feb 2013 22:19:20 +0000 (22:19 +0000)]
drivers: net: davinci_cpdma: acknowledge interrupt properly

CPDMA interrupts are not properly acknowledged which leads to interrupt
storm, only cpdma interrupt 0 is acknowledged in Davinci CPDMA driver.
Changed cpdma_ctlr_eoi api to acknowledge 1 and 2 interrupts which are
used for rx and tx respectively.

Reported-by: Pantelis Antoniou <panto@antoniou-consulting.com>
Signed-off-by: Mugunthan V N <mugunthanvnm@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoiwlwifi: dvm: improve (again) the prints in reclaim path
Emmanuel Grumbach [Thu, 14 Feb 2013 09:33:54 +0000 (11:33 +0200)]
iwlwifi: dvm: improve (again) the prints in reclaim path

I removed a bit too much info last time.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agoiwlwifi: dvm: fix delayed enter to CT-kill bug
Eytan Lifshitz [Fri, 15 Feb 2013 14:31:45 +0000 (16:31 +0200)]
iwlwifi: dvm: fix delayed enter to CT-kill bug

Theoretically, the card may not enter CTKILL:
In case the timer that iwl_prepare_ct_kill_task is setting,
will expire before tt->state revert to its previous state.

Signed-off-by: Eytan Lifshitz <eytan.lifshitz@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agoiwlwifi: mvm: Reserve MAC id 0 for managed interfaces
Ilan Peer [Wed, 13 Feb 2013 11:27:18 +0000 (13:27 +0200)]
iwlwifi: mvm: Reserve MAC id 0 for managed interfaces

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agoiwlwifi: mvm: Fix resource allocation for P2P Device
Ilan Peer [Wed, 13 Feb 2013 10:26:39 +0000 (12:26 +0200)]
iwlwifi: mvm: Fix resource allocation for P2P Device

The time event data structures are required also for P2P Device
interface.

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agoiwlwifi: mvm: Set the scan type according to the vif type
Ilan Peer [Mon, 11 Feb 2013 06:50:45 +0000 (08:50 +0200)]
iwlwifi: mvm: Set the scan type according to the vif type

The FW can differentiate between scans, according to the interface
type on which the scan was issues. Supply the interfaces type
information to the FW.

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agoiwlwifi: mvm: fix time event command handling race
Johannes Berg [Fri, 15 Feb 2013 23:11:34 +0000 (00:11 +0100)]
iwlwifi: mvm: fix time event command handling race

Occasionally, we would run into this warning:

  iwlwifi 0000:02:00.0: U iwl_mvm_protect_session extend 0x2601: only 200 ms left
  iwlwifi 0000:02:00.0: U iwl_mvm_remove_time_event Removing TE 0x2601
  iwlwifi 0000:02:00.0: I iwl_pcie_enqueue_hcmd Sending command TIME_EVENT_CMD (#29), seq: 0x0925, 60 bytes at 37[5]:9
  iwlwifi 0000:02:00.0: U iwl_pcie_send_hcmd_sync Attempting to send sync command TIME_EVENT_CMD
  iwlwifi 0000:02:00.0: U iwl_pcie_send_hcmd_sync Setting HCMD_ACTIVE for command TIME_EVENT_CMD
  iwlwifi 0000:02:00.0: I iwl_pcie_enqueue_hcmd Sending command TIME_EVENT_CMD (#29), seq: 0x0926, 60 bytes at 38[6]:9
  iwlwifi 0000:02:00.0: U iwl_mvm_time_event_response TIME_EVENT_CMD response - UID = 0x2601
  iwlwifi 0000:02:00.0: I iwl_pcie_hcmd_complete Clearing HCMD_ACTIVE for command TIME_EVENT_CMD
  iwlwifi 0000:02:00.0: U iwl_mvm_rx_time_event_notif Time event notification - UID = 0x2701 action 1
  wlan0: associate with 00:0a:b8:55:a8:30 (try 2/3)
  ------------[ cut here ]------------
  WARNING: at drivers/net/wireless/iwlwifi/mvm/time-event.c:269 iwl_mvm_time_event_send_add+0x163/0x1a0 [iwlmvm]()
  Modules linked in: [...]
  Call Trace:
   [<c1046e42>] warn_slowpath_common+0x72/0xa0
   [<c1046e92>] warn_slowpath_null+0x22/0x30
   [<f8cad913>] iwl_mvm_time_event_send_add+0x163/0x1a0 [iwlmvm]
   [<f8cadead>] iwl_mvm_protect_session+0xcd/0x1c0 [iwlmvm]
   [<f8ca2087>] iwl_mvm_mac_mgd_prepare_tx+0x67/0xa0 [iwlmvm]
   [<f882a130>] ieee80211_sta_work+0x8f0/0x1070 [mac80211]

The reason is a problem with asynchronous vs. synchronous
commands, what happens here is the following:
 * TE 0x2601 is removed, the TIME_EVENT_CMD for that is async
 * a new TE (will be 0x2701) is created, the TIME_EVENT_CMD
   for that is sync and also uses a notification wait for the
   response (to avoid another race condition)
 * the response for the TE 0x2601 removal comes from the
   firmware, and is handled by the notification wait handler
   that's really waiting for the second response, but can't
   tell the difference, we therefore see the message
   "TIME_EVENT_CMD response - UID = 0x2601" instead of
   "TIME_EVENT_CMD response - UID = 0x2701".

Fix this issue by making the TE removal synchronous as well,
this means that we wait for the response to that command
first, before there's any chance of sending a new one.

Also, to detect such issues more easily in the future, add
a warning to the notification handler that detects them.

Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agoiwlwifi: mvm: add debug statement to time event add
Johannes Berg [Fri, 15 Feb 2013 22:54:10 +0000 (23:54 +0100)]
iwlwifi: mvm: add debug statement to time event add

This is helpful for debugging the time event warning,
but also in general to see what's going on.

Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agoiwlwifi: mvm: fix GO powersave client manipulation
Johannes Berg [Fri, 15 Feb 2013 21:47:48 +0000 (22:47 +0100)]
iwlwifi: mvm: fix GO powersave client manipulation

All station commands must include a valid MAC ID,
the ID 0 is randomly valid in some cases, but we
must set the ID properly. Do that by passing the
right station and using its mac_id_n_color.

Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agoiwlwifi: mvm: program DTIM timings properly
Johannes Berg [Tue, 5 Feb 2013 17:10:04 +0000 (18:10 +0100)]
iwlwifi: mvm: program DTIM timings properly

For the firmware to know when DTIM beacons arrive
we have to program the DTIM time in TSF and system
time in the MAC context. Since mac80211 now tracks
the different times (on demand), this becomes easy.

Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agoiwlwifi: mvm: fix merge error
Johannes Berg [Fri, 15 Feb 2013 22:39:17 +0000 (23:39 +0100)]
iwlwifi: mvm: fix merge error

The iwlwifi-next tree removed IEEE80211_HW_NEED_DTIM_BEFORE_ASSOC
while the mac80211-next tree removed

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agoMerge remote-tracking branch 'wireless-next/master' into iwlwifi-next
Johannes Berg [Mon, 18 Feb 2013 19:16:12 +0000 (20:16 +0100)]
Merge remote-tracking branch 'wireless-next/master' into iwlwifi-next

9 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
John W. Linville [Mon, 18 Feb 2013 18:47:13 +0000 (13:47 -0500)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-next into for-davem

Conflicts:
drivers/net/wireless/iwlwifi/dvm/tx.c
drivers/net/wireless/ti/wlcore/sdio.c
drivers/net/wireless/ti/wlcore/spi.c

9 years agoMAINTAINERS: Jiri Pirko email change
Jiri Pirko [Fri, 15 Feb 2013 23:55:05 +0000 (23:55 +0000)]
MAINTAINERS: Jiri Pirko email change

Change email for team driver maintainership.

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agotg3: Use different macros for pci_chip_rev_id accesses
Joe Perches [Sat, 16 Feb 2013 11:20:04 +0000 (11:20 +0000)]
tg3: Use different macros for pci_chip_rev_id accesses

Upper case macros for various chip attributes are slightly
difficult to read and are a bit out of characterto the other
tg3_<foo> attribute functions.

Convert:

GET_ASIC_REV(tp->pci_chip_rev_id)       -> tg3_asic_rev(tp)
GET_CHIP_REV(tp->pci_chip_rev_id)       -> tg3_chip_rev(tp)

Remove:
GET_METAL_REV(tp->pci_chip_rev_id)      -> tg3_metal_rev(tp) (unused)

Add:
tg3_chip_rev_id(tp) for tp->pci_chip_rev_id so access styles
are similar to tg3_asic_rev and tg3_chip_rev.

These macros are not converted to static inline functions
because gcc (tested with 4.7.2) is currently unable to
optimize the object code it produces the same way and code
is otherwise larger.

Signed-off-by: Joe Perches <joe@perches.com>
Acked-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agotg3: Remove define and single use of GET_CHIP_REV_ID
Joe Perches [Sat, 16 Feb 2013 11:20:03 +0000 (11:20 +0000)]
tg3: Remove define and single use of GET_CHIP_REV_ID

It's the same value as tp->pci_chip_rev_id so use that
instead.  This makes all CHIPREV_ID_<foo> tests the same.

Signed-off-by: Joe Perches <joe@perches.com>
Acked-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobgmac: fix unaligned accesses to network headers
Hauke Mehrtens [Sat, 16 Feb 2013 11:10:38 +0000 (11:10 +0000)]
bgmac: fix unaligned accesses to network headers

Without this patch I get many unaligned access warnings per packet,
this patches fixes them all. This should improve performance on some
systems like mips.

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoip: fix warning in xfrm4_mode_tunnel_input
stephen hemminger [Sat, 16 Feb 2013 07:40:29 +0000 (07:40 +0000)]
ip: fix warning in xfrm4_mode_tunnel_input

Same problem as IPv6

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoipv6: fix warning in xfrm6_mode_tunnel_input
stephen hemminger [Sat, 16 Feb 2013 07:38:15 +0000 (07:38 +0000)]
ipv6: fix warning in xfrm6_mode_tunnel_input

Should not use assignment in conditional:
 warning: suggest parentheses around assignment used as truth value [-Wparentheses]

Problem introduced by:
commit 14bbd6a565e1bcdc240d44687edb93f721cfdf99
Author: Pravin B Shelar <pshelar@nicira.com>
Date:   Thu Feb 14 09:44:49 2013 +0000

    net: Add skb_unclone() helper function.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/net...
David S. Miller [Mon, 18 Feb 2013 17:35:42 +0000 (12:35 -0500)]
Merge branch 'master' of git://git./linux/kernel/git/jkirsher/net-next

Jeff Kirsher says:

====================
This series contains updates to e1000, e1000e, igb, igbvf and ixgbe.
The e1000, e1000e, igb and igbvf are single patch changes and the
remaining 11 patches are all against ixgbe.

The e1000 patch is a comment cleanup to align e1000 with the code
commenting style for /drivers/net.  It also contains a few other white
space cleanups (i.e. fix lines over 80 char, remove unnecessary blank
lines and fix the use of tabs/spaces).

The e1000e patch from Koki (Fujitsu) adds a warning when link speed is
downgraded due to SmartSpeed.

The igb patch from Stefan (Red Hat) increases the timeout in the ethtool
offline self-test because some i350 adapters would sometimes fail the
self-test because link auto negotiation may take longer than the current
4 second timeout.

The igbvf patch from Alex is meant to address several race issues that
become possible because next_to_watch could possibly be set to a value
that shows that the descriptor is done when it is not.  In order to correct
that we instead make next_to_watch a pointer that is set to NULL during
cleanup, and set to the eop_desc after the descriptor rings have been written.

The remaining patches for ixgbe are a mix of fixes and added support as well
as some cleanup.  Most notably is the added support for displaying the
number of Tx/Rx channels via ethtool by Alex.  Also Aurélien adds the
ability for reading data from SFP+ modules over i2c for diagnostic
monitoring.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fix a compile error when SOCK_REFCNT_DEBUG is enabled
Ying Xue [Fri, 15 Feb 2013 22:28:25 +0000 (22:28 +0000)]
net: fix a compile error when SOCK_REFCNT_DEBUG is enabled

When SOCK_REFCNT_DEBUG is enabled, below build error is met:

kernel/sysctl_binary.o: In function `sk_refcnt_debug_release':
include/net/sock.h:1025: multiple definition of `sk_refcnt_debug_release'
kernel/sysctl.o:include/net/sock.h:1025: first defined here
kernel/audit.o: In function `sk_refcnt_debug_release':
include/net/sock.h:1025: multiple definition of `sk_refcnt_debug_release'
kernel/sysctl.o:include/net/sock.h:1025: first defined here
make[1]: *** [kernel/built-in.o] Error 1
make: *** [kernel] Error 2

So we decide to make sk_refcnt_debug_release static to eliminate
the error.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: move ioctl functions into a separated file
Cong Wang [Fri, 15 Feb 2013 22:20:46 +0000 (22:20 +0000)]
net: move ioctl functions into a separated file

They well deserve a separated unit.

Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: ehea module param description fix
Dave Young [Fri, 15 Feb 2013 20:19:19 +0000 (20:19 +0000)]
net: ehea module param description fix

In ehea.h the minimal entries is 2^7 - 1:
#define EHEA_MIN_ENTRIES_QP  127

Thus change the module param description accordinglly

Signed-off-by: Dave Young <dyoung@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agocfg80211: fix station change if TDLS isn't supported
Johannes Berg [Mon, 18 Feb 2013 13:57:42 +0000 (14:57 +0100)]
cfg80211: fix station change if TDLS isn't supported

Larry noticed (and bisected) that commit df881293c6ba9a12868491a717b25
"cfg80211: Pass TDLS peer's QoS/HT/VHT information during set_station"
broke secure connections. This is is the case only for drivers that
don't support TDLS, where any kind of change, even just the change of
authorized flag that is required for normal operation, was rejected
now. To fix this, remove the checks. I have some patches that will add
proper verification for all the different cases later.

Cc: Jouni Malinen <j@w1.fi>
Bisected-by: Larry Finger <Larry.Finger@lwfinger.net>
Tested-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agoipv6: optimize inet6_hash_frag()
Eric Dumazet [Fri, 15 Feb 2013 18:14:48 +0000 (18:14 +0000)]
ipv6: optimize inet6_hash_frag()

Use ipv6_addr_hash() and a single jhash invocation.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'tipc_net-next' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg...
David S. Miller [Mon, 18 Feb 2013 17:22:17 +0000 (12:22 -0500)]
Merge branch 'tipc_net-next' of git://git./linux/kernel/git/paulg/linux

Paul Gortmaker says:

====================
Two relatively small cleanup patches here, plus a reimplementation
of the patch Neil had questions about[1] in the last development
cycle.

Tested on today's net-next, between 32 and 64 bit x86 machines using
the server/client in tipc-utils, as usual.

[1] http://patchwork.ozlabs.org/patch/204507/
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agomac80211: don't spam mesh probe response messages
Thomas Pedersen [Sat, 16 Feb 2013 18:41:06 +0000 (10:41 -0800)]
mac80211: don't spam mesh probe response messages

If mesh plink debugging is enabled, this gets annoying in
a crowded environment, fast.

Signed-off-by: Thomas Pedersen <thomas@cozybit.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: stringify mesh peering events
Thomas Pedersen [Sat, 16 Feb 2013 18:41:05 +0000 (10:41 -0800)]
mac80211: stringify mesh peering events

Convert mesh peering events into strings and make the
debug output a little easier to read. Also stop printing
the llid and plid since these don't change across peering
states and are random numbers anyway so they just amount
to noise.

Signed-off-by: Thomas Pedersen <thomas@cozybit.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: clean up mesh HT operation
Thomas Pedersen [Sat, 16 Feb 2013 18:41:04 +0000 (10:41 -0800)]
mac80211: clean up mesh HT operation

ieee80211_ht_cap_ie_to_sta_ht_cap() will clean up the
ht_supported flag and station bandwidth field for us
if the peer beacon doesn't have an HT capability element
(is operating as non-HT).

Also, we don't really need a special station ch_width
member to track the station operating mode any more so use
sta.bandwidth instead.

Signed-off-by: Thomas Pedersen <thomas@cozybit.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: fix harmless station flush warning
Johannes Berg [Fri, 15 Feb 2013 10:47:50 +0000 (11:47 +0100)]
mac80211: fix harmless station flush warning

If an interface is set down while authenticating or
associating, there's a station entry that will be
removed by the flushing in do_stop() and that will
cause a warning. It's otherwise harmless, but avoid
the warning by calling ieee80211_mgd_stop() first.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agocfg80211: add correct docbook entries
Johannes Berg [Fri, 15 Feb 2013 09:11:24 +0000 (10:11 +0100)]
cfg80211: add correct docbook entries

Update the 802.11 book to list the correct
data structures.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agoixgbe: Update DESC_NEEDED define to adjust for changes to MAX_SKB_FRAGS
Alexander Duyck [Sat, 26 Jan 2013 02:08:14 +0000 (02:08 +0000)]
ixgbe: Update DESC_NEEDED define to adjust for changes to MAX_SKB_FRAGS

Recent changes have made it so that MAX_SKB_FRAGS is now never less than 16.
As a result we were seeing issues on systems with 64K pages as it would
cause DESC_NEEDED to increase to 68, and we would need over 136 descriptors
free before clean_tx_irq would wake the queue.

This patch makes it so that DESC_NEEDED is always MAX_SKB_FRAGS + 4.  This
should prevent any possible deadlocks on the systems with 64K pages as we will
now only require 42 descriptors to wake.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoixgbe: fix Tx timeouts with BQL
Emil Tantilov [Fri, 25 Jan 2013 06:19:20 +0000 (06:19 +0000)]
ixgbe: fix Tx timeouts with BQL

This patch makes sure that TXDCTL.WTHRESH is set to 1 when BQL is enabled
and EITR is set to more than 100k interrupts per second to avoid Tx timeouts.

Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoixgbe: implement SFF diagnostic monitoring via ethtool
Aurélien Guillaume [Thu, 17 Jan 2013 06:55:24 +0000 (06:55 +0000)]
ixgbe: implement SFF diagnostic monitoring via ethtool

This patch adds support for reading data from SFP+ modules over i2c.

Signed-off-by: Aurélien Guillaume <footplus@gmail.com>
Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoixgbe: cleanup error checking in ixgbe_identify_sfp_module_generic()
Emil Tantilov [Fri, 18 Jan 2013 02:17:11 +0000 (02:17 +0000)]
ixgbe: cleanup error checking in ixgbe_identify_sfp_module_generic()

This patch replaces instances where a return code from i2c operations
were checked against a list of error codes with a much simpler
if ( status != 0 ) check.

Some whitespace cleanups included.

Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoixgbe: fix possible data corruption in read_i2c_byte
Emil Tantilov [Fri, 18 Jan 2013 02:16:41 +0000 (02:16 +0000)]
ixgbe: fix possible data corruption in read_i2c_byte

This patch makes sure that the SW lock is released after all i2c
operations complete in the retry code path.

Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoixgbe: Add support for set_channels ethtool operation
Alexander Duyck [Thu, 17 Jan 2013 08:39:33 +0000 (08:39 +0000)]
ixgbe: Add support for set_channels ethtool operation

This change adds support for the ethtool set_channels operation.

Since the ixgbe driver has to support DCB as well as the other modes the
assumption I made here is that the number of channels in DCB modes refers
to the number of queues per traffic class, not the number of queues total.

CC: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoixgbe: Add support for displaying the number of Tx/Rx channels
Alexander Duyck [Sat, 12 Jan 2013 06:33:52 +0000 (06:33 +0000)]
ixgbe: Add support for displaying the number of Tx/Rx channels

This patch adds support for the ethtool get_channels operation.

Since the ixgbe driver has to support DCB as well as the other modes the
assumption I made here is that the number of channels in DCB modes refers
to the number of queues per traffic class, not the number of queues total.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Reviewed-by: John Fastabend <john.r.fastabend@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoixgbe: Make ixgbe_setup_tc usable even when DCB is not enabled
Alexander Duyck [Sat, 12 Jan 2013 06:33:44 +0000 (06:33 +0000)]
ixgbe: Make ixgbe_setup_tc usable even when DCB is not enabled

The ixgbe_setup_tc code is essentially the same code we need any time we have
to update the number of queues.  As such I am making it available always and
just stripping the DCB specific bits out when DCB is disabled instead of
stripping the entire function.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Reviewed-by: John Fastabend <john.r.fastabend@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoixgbe: Update ixgbe driver to use __netdev_pick_tx in ixgbe_select_queue
Alexander Duyck [Sat, 12 Jan 2013 06:33:37 +0000 (06:33 +0000)]
ixgbe: Update ixgbe driver to use __netdev_pick_tx in ixgbe_select_queue

This change updates the ixgbe driver to use __netdev_pick_tx instead of
the current logic it is using to select a queue.  The main result of this
change is that ixgbe can now fully support XPS, and in the case of non-FCoE
enabled configs it means we don't need to have our own ndo_select_queue.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Reviewed-by: John Fastabend <john.r.fastabend@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoixgbe: Add function for setting XPS queue mapping
Alexander Duyck [Sat, 12 Jan 2013 06:33:31 +0000 (06:33 +0000)]
ixgbe: Add function for setting XPS queue mapping

This change adds support for ixgbe to configure the XPS queue mapping on
load.  The result of this change is that on open we will now be resetting
the number of Tx queues, and then setting the default configuration for XPS
based on if ATR is enabled or disabled.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Reviewed-by: John Fastabend <john.r.fastabend@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoixgbe: Define FCoE and Flow director limits much sooner to allow for changes
Alexander Duyck [Wed, 16 Jan 2013 01:35:35 +0000 (01:35 +0000)]
ixgbe: Define FCoE and Flow director limits much sooner to allow for changes

Instead of adjusting the FCoE and Flow director limits based on the number
of CPUs we can define them much sooner.  This allows the user to come
through later and adjust them once we have updated the code to support the
set_channels ethtool operation.

I am still allowing for FCoE and RSS queues to be separated if the number
queues is less than the number of CPUs.  This essentially treats the two
groupings like they are two separate traffic classes.

In addition I am changing the initialization to use the MAX_TX/RX_QUEUES
defines instead of trying to compute the value as it will be possible in
upcoming patches for the user to request the maximum number of queues.

I have also updated things so that the upper limit on queues is exactly 63
instead of allowing it to go up to 64.  The reason for this change is to
address the fact thqt the driver only supports up to 63 queue vectors since
the hardware supports 64 MSI-X vectors, but one must be reserved for "other"
causes.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoigb: increase timeout for ethtool offline self-test
Stefan Assmann [Sat, 2 Feb 2013 08:31:50 +0000 (08:31 +0000)]
igb: increase timeout for ethtool offline self-test

On several machines with i350 adapters the ethtool offline self-test sometimes
fails. This happens because link auto negotiation may take longer than the
timeout of 4 seconds. Increasing the timeout by 1 seconds resolves the issue.

Output from a failing i350 offline self-test:
while [ 1 ]; do ethtool -t eth2 offline; done
The test result is PASS
The test extra info:
Register test  (offline)         0
Eeprom test    (offline)         0
Interrupt test (offline)         0
Loopback test  (offline)         0
Link test   (on/offline)         0

The test result is FAIL
The test extra info:
Register test  (offline)         0
Eeprom test    (offline)         0
Interrupt test (offline)         0
Loopback test  (offline)         0
Link test   (on/offline)         1

The test result is PASS
The test extra info:
Register test  (offline)         0
Eeprom test    (offline)         0
Interrupt test (offline)         0
Loopback test  (offline)         0
Link test   (on/offline)         0

Signed-off-by: Stefan Assmann <sassmann@kpanic.de>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoigbvf: Make next_to_watch a pointer and adjust memory barriers to avoid races
Alexander Duyck [Thu, 31 Jan 2013 07:15:51 +0000 (07:15 +0000)]
igbvf: Make next_to_watch a pointer and adjust memory barriers to avoid races

This change is meant to address several race issues that become possible
because next_to_watch could possibly be set to a value that shows that the
descriptor is done when it is not.  In order to correct that we instead make
next_to_watch a pointer that is set to NULL during cleanup, and set to the
eop_desc after the descriptor rings have been written.

To enforce proper ordering the next_to_watch pointer is not set until after
a wmb writing the values to the last descriptor in a transmit.  In order to
guarantee that the descriptor is not read until after the eop_desc we use the
read_barrier_depends which is only really necessary on the alpha architecture.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Acked-by: Greg Rose <gregory.v.rose@intel.com>
Tested-by: Sibai Li <sibai.li@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoe1000e: display a warning message when SmartSpeed works
Koki Sanagi [Sun, 3 Feb 2013 14:03:55 +0000 (14:03 +0000)]
e1000e: display a warning message when SmartSpeed works

Current e1000e driver doesn't tell nothing when Link Speed is downgraded due to
SmartSpeed.  As a result, users suspect that there is something wrong with
NIC.  If the cause of it is SmartSpeed, there is no means to replace NIC. This
patch make e1000e notify users that SmartSpeed worked.

Signed-off-by: Koki Sanagi <sanagi.koki@jp.fujitsu.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoe1000: fix whitespace issues and multi-line comments
Jeff Kirsher [Sat, 9 Feb 2013 12:49:21 +0000 (12:49 +0000)]
e1000: fix whitespace issues and multi-line comments

Fixes whitespace issues, such as lines exceeding 80 chars, needless blank
lines and the use of spaces where tabs are needed.  In addition, fix
multi-line comments to align with the networking standard.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
9 years agotipc: remove redundant checking for the number of iovecs in a send request
Ying Xue [Thu, 31 Jan 2013 20:51:47 +0000 (21:51 +0100)]
tipc: remove redundant checking for the number of iovecs in a send request

As the number of iovecs in a send request is already limited within
UIO_MAXIOV(i.e. 1024) in __sys_sendmsg(), it's unnecessary to check it
again in TIPC stack.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
9 years agotipc: byte-based overload control on socket receive queue
Ying Xue [Sun, 20 Jan 2013 22:30:09 +0000 (23:30 +0100)]
tipc: byte-based overload control on socket receive queue

Change overload control to be purely byte-based, using
sk->sk_rmem_alloc as byte counter, and compare it to a calculated
upper limit for the socket receive queue.

For all connection messages, irrespective of message importance,
the overload limit is set to a constant value (i.e, 67MB). This
limit should normally never be reached because of the lower
limit used by the flow control algorithm, and is there only
as a last resort in case a faulty peer doesn't respect the send
window limit.

For datagram messages, message importance is taken into account
when calculating the overload limit. The calculation is based
on sk->sk_rcvbuf, and is hence configurable via the socket option
SO_RCVBUF.

Cc: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
9 years agotipc: eliminate duplicated discard_rx_queue routine
Ying Xue [Sun, 20 Jan 2013 22:30:08 +0000 (23:30 +0100)]
tipc: eliminate duplicated discard_rx_queue routine

The tipc function discard_rx_queue() is just a duplicated
implementation of __skb_queue_purge().  Remove the former
and directly invoke __skb_queue_purge().

In doing so, the underscores convey to the code reader, more
information about the current locking state that is assumed.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
9 years agotipc: fix missing spinlock init in broadcast code
Erik Hugne [Thu, 14 Feb 2013 14:43:33 +0000 (14:43 +0000)]
tipc: fix missing spinlock init in broadcast code

After commit 3c294cb3 "tipc: remove the bearer congestion mechanism",
we try to grab the broadcast bearer lock when sending multicast
messages over the broadcast link. This will cause an oops because
the lock is never initialized. This is an old bug, but the lock
was never actually used before commit 3c294cb3, so that why it was
not visible until now.  The oops will look something like:

BUG: spinlock bad magic on CPU#2, daemon/147
lock: bcast_bearer+0x48/0xffffffffffffd19a [tipc],
.magic: 00000000, .owner: <none>/-1, .owner_cpu: 0
Pid: 147, comm: daemon Not tainted 3.8.0-rc3+ #206
Call Trace:
spin_dump+0x8a/0x8f
spin_bug+0x21/0x26
do_raw_spin_lock+0x114/0x150
_raw_spin_lock_bh+0x19/0x20
tipc_bearer_blocked+0x1f/0x40 [tipc]
tipc_link_send_buf+0x82/0x280 [tipc]
? __alloc_skb+0x9f/0x2b0
tipc_bclink_send_msg+0x77/0xa0 [tipc]
tipc_multicast+0x11b/0x1b0 [tipc]
send_msg+0x225/0x530 [tipc]
sock_sendmsg+0xca/0xe0

The above can be triggered by running the multicast demo program.

Signed-off-by: Erik Hugne <erik.hugne@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: use skb_reset_mac_len() in dev_gro_receive()
Eric Dumazet [Thu, 14 Feb 2013 17:31:48 +0000 (17:31 +0000)]
net: use skb_reset_mac_len() in dev_gro_receive()

We no longer need to use mac_len, lets cleanup things.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>