This patch stops the driver to invoke kthread (which remove the dead ioc)
for some time while EEH recovery has started.

[thenzl: add a 'non_operational_loop' reset.]
Signed-off-by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
Signed-off-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

The variable 'chain_flags' and 'phy_number' are initialized but never used
otherwise, So remove those unused variables.
Signed-off-by: Wei Yongjun <yongjun_wei@trendmicro.com.cn>
Signed-off-by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

linux/version.h is not necessary as detected by checkversion.pl script.
Signed-off-by: Sachin Kamat <sachin.kamat@linaro.org>
Acked-by: "Sreekanth Reddy" <Sreekanth.reddy@lsi.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

The intention in bnx2i_send_fw_iscsi_init_msg was to zero out
only the lower 32bits, but instead the whole mask64 is zeroed.
This patch fixes it.
Signed-off-by: Tomas Henzl <thenzl@redhat.com>
Acked-by: Eddie Wai <eddie.wai@broadcom.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Modified fnic driver to let hardware insert the COS value. Set bit
in descriptor to 0 telling hardware to use its lif COS configurations
to insert the COS value in the frames.
Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
Signed-off-by: Hiral Patel <hiralpat@cisco.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Updated MAINTAINERS list for fnic driver with appropriate names and email ids.
Signed-off-by: Brian Uchino <buchino@cisco.com>
Signed-off-by: Hiral Patel <hiralpat@cisco.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Signed-off-by: Vijaya Mohan Guvva <vmohan@brocade.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

ioc->diag_trigger_mpi is an SL_WH_MPI_TRIGGERS_T struct.

There is a cut and paste error here and SL_WH_EVENT_TRIGGERS_T is used
instead of SL_WH_MPI_TRIGGERS_T.  Since the SL_WH_EVENT_TRIGGERS_T is
smaller than SL_WH_MPI_TRIGGERS_T, it means we only clear part of the
buffer.

I've changed it to use sizeof(ioc->diag_trigger_mpi) which is a bit
simpler.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: "Sreekanth Reddy" <Sreekanth.reddy@lsi.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

This removes an open coded simple_open() function and
replaces file operations references to the function
with simple_open() instead.
Signed-off-by: Wei Yongjun <yongjun_wei@trendmicro.com.cn>
Acked-by: Naresh Kumar Inna <naresh@chelsio.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Signed-off-by: Manish Rangankar <manish.rangankar@qlogic.com>
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

In qla4xxx_get_ep_fwdb(), dst_addr is of type struct sockaddr.
We are copying sizeof(struct sockaddr_in6) bytes to dst_addr
which is 12 bytes larger. This will cause memory corruption.
So we change dst_addr to struct sockaddr_storage which is
of 128 byte, large enough to hold sizeof(struct sockaddr_in6).
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Manish Rangankar <manish.rangankar@qlogic.com>
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Issue:
Reset operation fails if port is in Link Down state

Fix:
Do not wait till HBA comes online after reset command is invoked.
Signed-off-by: Harish Zunjarrao <harish.zunjarrao@qlogic.com>
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Issue:
Mailbox command timed out after switching from polling mode to interrupt mode.

Events:-
 1. Mailbox interrupts are disabled
 2. FW generates AEN and at same time driver enables Mailbox Interrupt
 3. Driver issues new mailbox to Firmware

In above case driver will not get AEN interrupts generated by FW in step #2 as
FW generated this AEN when interrupts are disabled. During the same time
driver enabled the mailbox interrupt, so driver will not poll for interrupt.
Driver will never process AENs generated in step #2 and issues new mailbox to
FW, but now FW is not able to post mailbox completion as AENs generated before
are not processed by driver.

Fix:
Enable Mailbox / AEN interrupts before initializing FW in case of ISP83XX.
This will make sure we process all Mailbox and AENs in interrupt mode.
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

The ROM address handling in gdth_init_pci() is useless and possibly
dangerous.  This patch removes it.

"pci_resource_start(pdev, 8)" is not well-defined.  PCI resources 0-5 are
standard PCI BARs and 6 is the expansion ROM.  Resource 8 is either an
SR-IOV BAR (if CONFIG_PCI_IOV=y, resources 7-12 are SR-IOV BARs) or a
bridge window (resources 7-10).

The GDT device is neither an SR-IOV device nor a bridge, so in either case
resource 8 should be zero since struct pci_dev is allocated with kzalloc().

It is illegal for a driver to write an arbitrary address to the ROM BAR
because it has no way of knowing whether the ROM will conflict with another
device.

I think the only effect of the code being removed was to:

  1) Enable the ROM at 0xFEFF0000 (possibly causing a conflict with
     another device)
  2) Delay one millisecond
  3) Write zero to the ROM BAR, disabling it

I doubt the delay is needed, but I left it since it seems innocuous.
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Pull dmraid fix from NeilBrown:
 "Just one fix for md in 3.8

  dmraid assess redundancy and replacements slightly inaccurately which
  could lead to some degraded arrays failing to assemble."

* tag 'md-3.8-fixes' of git://neil.brown.name/md:
  DM-RAID: Fix RAID10's check for sufficient redundancy

Pull GFS2 fix from Steven Whitehouse.

* git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-3.0-fixes:
  GFS2: fix skip unlock condition

Pull IOMMU fix from Joerg Roedel:
 "One fix for the AMD IOMMU driver to work around broken BIOSes found in
  the field.  Some BIOSes forget to enable a workaround for a hardware
  problem which might cause the IOMMU to stop working under high load
  conditions.  The fix makes sure this workaround is enabled."

* tag 'iommu-fixes-v3.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu:
  IOMMU, AMD Family15h Model10-1Fh erratum 746 Workaround

Pull MFD fixes from Samuel Ortiz:
 "This is the first pull request for MFD fixes for 3.8

  We have some build failure fixes (twl4030, vexpress, abx500 and
  tps65910), some actual runtime oops and lockup fixes (rtsx, da9052),
  and some more hypothetical NULL pointers dereferences fixes for
  pcf50633 and max776xx.

  Then we also have additional rtsx fixes for a correct switch output
  voltage and clock divider correctness for rtl8411 (rtsx driver), and
  irqdomain fix for db8550-prcmu, and some more cosmetic fixes for
  arizona and wm5102."

* tag 'mfd-for-linus-3.8-1' of git://git.kernel.org/pub/scm/linux/kernel/git/sameo/mfd-2.6:
  mfd: rtsx: Fix oops when rtsx_pci_sdmmc is not probed
  mfd: wm5102: Fix definition of WM5102_MAX_REGISTER
  mfd: twl4030: Don't warn about uninitialized return code
  mfd: da9052/53 lockup fix
  mfd: rtsx: Add clock divider hook
  mmc: rtsx: Call MFD hook to switch output voltage
  mfd: rtsx: Add output voltage switch hook
  mfd: Fix compile errors and warnings when !CONFIG_AB8500_BM
  mfd: vexpress: Export global functions to fix build error
  mfd: arizona: Check errors from regcache_sync()
  mfd: tc3589x: Use simple irqdomain
  mfd: pcf50633: Init pcf->dev before using it
  mfd: max77693: Init max77693->dev before using it
  mfd: max77686: Init max77686->dev before using it
  mfd: db8500-prcmu: Fix irqdomain usage
  mfd: tps65910: Select REGMAP_IRQ in Kconfig to fix build error
  mfd: arizona: Disable control interface reporting for WM5102 and WM5110

Pull networking updates from David Miller:
 "Much more accumulated than I would have liked due to an unexpected
  bout with a nasty flu:

   1) AH and ESP input don't set ECN field correctly because the
      transport head of the SKB isn't set correctly, fix from Li
      RongQing.

   2) If netfilter conntrack zones are disabled, we can return an
      uninitialized variable instead of the proper error code.  Fix from
      Borislav Petkov.

   3) Fix double SKB free in ath9k driver beacon handling, from Felix
      Feitkau.

   4) Remove bogus assumption about netns cleanup ordering in
      nf_conntrack, from Pablo Neira Ayuso.

   5) Remove a bogus BUG_ON in the new TCP fastopen code, from Eric
      Dumazet.  It uses spin_is_locked() in it's test and is therefore
      unsuitable for UP.

   6) Fix SELINUX labelling regressions added by the tuntap multiqueue
      changes, from Paul Moore.

   7) Fix CRC errors with jumbo frame receive in tg3 driver, from Nithin
      Nayak Sujir.

   8) CXGB4 driver sets interrupt coalescing parameters only on first
      queue, rather than all of them.  Fix from Thadeu Lima de Souza
      Cascardo.

   9) Fix regression in the dispatch of read/write registers in dm9601
      driver, from Tushar Behera.

  10) ipv6_append_data miscalculates header length, from Romain KUNTZ.

  11) Fix PMTU handling regressions on ipv4 routes, from Steffen
      Klassert, Timo Teräs, and Julian Anastasov.

  12) In 3c574_cs driver, add necessary parenthesis to "x << y & z"
      expression.  From Nickolai Zeldovich.

  13) macvlan_get_size() causes underallocation netlink message space,
      fix from Eric Dumazet.

  14) Avoid division by zero in xfrm_replay_advance_bmp(), from Nickolai
      Zeldovich.  Amusingly the zero check was already there, we were
      just performing it after the modulus :-)

  15) Some more splice bug fixes from Eric Dumazet, which fix things
      mostly eminating from how we now more aggressively use high-order
      pages in SKBs.

  16) Fix size calculation bug when freeing hash tables in the IPSEC
      xfrm code, from Michal Kubecek.

  17) Fix PMTU event propagation into socket cached routes, from Steffen
      Klassert.

  18) Fix off by one in TX buffer release in netxen driver, from Eric
      Dumazet.

  19) Fix rediculous memory allocation requirements introduced by the
      tuntap multiqueue changes, from Jason Wang.

  20) Remove bogus AMD platform workaround in r8169 driver that causes
      major problems in normal operation, from Timo Teräs.

  21) virtio-net set affinity and select queue don't handle
      discontiguous cpu numbers properly, fix from Wanlong Gao.

  22) Fix a route refcounting issue in loopback driver, from Eric
      Dumazet.  There's a similar fix coming that we might add to the
      macvlan driver as well.

  23) Fix SKB leaks in batman-adv's distributed arp table code, from
      Matthias Schiffer.

  24) r8169 driver gives descriptor ownership back the hardware before
      we're done reading the VLAN tag out of it, fix from Francois
      Romieu.

  25) Checksums not calculated properly in GRE tunnel driver fix from
      Pravin B Shelar.

26) Fix SCTP memory leak on namespace exit."

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (101 commits)
  dm9601: support dm9620 variant
  SCTP: Free the per-net sysctl table on net exit. v2
  net: phy: icplus: fix broken INTR pin settings
  net: phy: icplus: Use the RGMII interface mode to configure clock delays
  IP_GRE: Fix kernel panic in IP_GRE with GRE csum.
  sctp: set association state to established in dupcook_a handler
  ip6mr: limit IPv6 MRT_TABLE identifiers
  r8169: fix vlan tag read ordering.
  net: cdc_ncm: use IAD provided by the USB core
  batman-adv: filter ARP packets with invalid MAC addresses in DAT
  batman-adv: check for more types of invalid IP addresses in DAT
  batman-adv: fix skb leak in batadv_dat_snoop_incoming_arp_reply()
  net: loopback: fix a dst refcounting issue
  virtio-net: reset virtqueue affinity when doing cpu hotplug
  virtio-net: split out clean affinity function
  virtio-net: fix the set affinity bug when CPU IDs are not consecutive
  can: pch_can: fix invalid error codes
  can: ti_hecc: fix invalid error codes
  can: c_can: fix invalid error codes
  r8169: remove the obsolete and incorrect AMD workaround
  ...

The IOMMU may stop processing page translations due to a perceived lack
of credits for writing upstream peripheral page service request (PPR)
or event logs. If the L2B miscellaneous clock gating feature is enabled
the IOMMU does not properly register credits after the log request has
completed, leading to a potential system hang.

BIOSes are supposed to disable L2B micellaneous clock gating by setting
L2_L2B_CK_GATE_CONTROL[CKGateL2BMiscDisable](D0F2xF4_x90[2]) = 1b. This
patch corrects that for those which do not enable this workaround.
Signed-off-by: Suravee Suthikulpanit <suravee.suthikulpanit@amd.com>
Acked-by: Borislav Petkov <bp@suse.de>
Cc: stable@vger.kernel.org
Signed-off-by: Joerg Roedel <joro@8bytes.org>

The recent commit fb6791d1
included the wrong logic.  The lvbptr check was incorrectly
added after the patch was tested.
Signed-off-by: David Teigland <teigland@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

dm9620 is a newer variant of dm9601 with more features (usb 2.0, checksum
offload, ..), but it can also be put in a dm9601 compatible mode, allowing
us to reuse the existing driver.

This does mean that the extended features like checksum offload cannot be
used, but that's hardly critical on a 100mbps interface.

Thanks to Sławek Wernikowski <slawek@wernikowski.net> for providing me
with a dm9620 based device to test.
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>

Per-net sysctl table needs to be explicitly freed at
net exit.  Otherwise we see the following with kmemleak:

unreferenced object 0xffff880402d08000 (size 2048):
  comm "chrome_sandbox", pid 18437, jiffies 4310887172 (age 9097.630s)
  hex dump (first 32 bytes):
    b2 68 89 81 ff ff ff ff 20 04 04 f8 01 88 ff ff  .h...... .......
    04 00 00 00 a4 01 00 00 00 00 00 00 00 00 00 00  ................
  backtrace:
    [<ffffffff815b4aad>] kmemleak_alloc+0x21/0x3e
    [<ffffffff81110352>] slab_post_alloc_hook+0x28/0x2a
    [<ffffffff81113fad>] __kmalloc_track_caller+0xf1/0x104
    [<ffffffff810f10c2>] kmemdup+0x1b/0x30
    [<ffffffff81571e9f>] sctp_sysctl_net_register+0x1f/0x72
    [<ffffffff8155d305>] sctp_net_init+0x100/0x39f
    [<ffffffff814ad53c>] ops_init+0xc6/0xf5
    [<ffffffff814ad5b7>] setup_net+0x4c/0xd0
    [<ffffffff814ada5e>] copy_net_ns+0x6d/0xd6
    [<ffffffff810938b1>] create_new_namespaces+0xd7/0x147
    [<ffffffff810939f4>] copy_namespaces+0x63/0x99
    [<ffffffff81076733>] copy_process+0xa65/0x1233
    [<ffffffff81077030>] do_fork+0x10b/0x271
    [<ffffffff8100a0e9>] sys_clone+0x23/0x25
    [<ffffffff815dda73>] stub_clone+0x13/0x20
    [<ffffffffffffffff>] 0xffffffffffffffff

I fixed the spelling of sysctl_header so the code actually
compiles. -- EWB.
Reported-by: Martin Mokrejs <mmokrejs@fold.natur.cuni.cz>
Signed-off-by: Vlad Yasevich <vyasevich@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch fixes the setting of the INTR pin that is
valid for IP101 A/G device and not for the IP1001.
Reported-by: Anunay Saxena <anunay.saxena@st.com>
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Like several other PHY devices which support RGMII, the IC+1001 allows
additional delays to by added to the RX_CLK and TX_CLK signals to
compensate for skew between the clock and data signals. Previously this
was always enabled, but this change makes use of the different RGMII
interface modes to allow the user to specify whether this should be
enabled.
Signed-off-by: Stuart Menefy <stuart.menefy@st.com>
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Due to IP_GRE GSO support, GRE can recieve non linear skb which
results in panic in case of GRE_CSUM.  Following patch fixes it by
using correct csum API.

Bug introduced in commit 6b78f16e (gre: add GSO support)
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

While sctp handling a duplicate COOKIE-ECHO and the action is
'Association restart', sctp_sf_do_dupcook_a() will processing
the unexpected COOKIE-ECHO for peer restart, but it does not set
the association state to SCTP_STATE_ESTABLISHED, so the association
could stuck in SCTP_STATE_SHUTDOWN_PENDING state forever.
This violates the sctp specification:
  RFC 4960 5.2.4. Handle a COOKIE ECHO when a TCB Exists
  Action
  A) In this case, the peer may have restarted. .....
     After this, the endpoint shall enter the ESTABLISHED state.

To resolve this problem, adding a SCTP_CMD_NEW_STATE cmd to the
command list before SCTP_CMD_REPLY cmd, this will set the restart
association to SCTP_STATE_ESTABLISHED state properly and also avoid
I-bit being set in the DATA chunk header when COOKIE_ACK is bundled
with DATA chunks.
Signed-off-by: Xufeng Zhang <xufeng.zhang@windriver.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Vlad Yasevich <vyasevich@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

We did this for IPv4 in b49d3c1e "net: ipmr: limit MRT_TABLE
identifiers" but we need to do it for IPv6 as well.  On IPv6 the name
is "pim6reg" instead of "pimreg" so there is one less digit allowed.

The strcpy() is in ip6mr_reg_vif().
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Control of receive descriptor must not be returned to ethernet chipset
before vlan tag processing is done.

VLAN tag receive word is now reset both in normal and error path.
Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Spotted-by: Timo Teras <timo.teras@iki.fi>
Cc: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

commit 9992c2e2 (net: cdc_ncm: workaround for missing CDC Union)
added code to lookup an IAD for the interface we are probing.
This is redundant.  The USB core has already done the lookup
and saved the result in the USB interface struct.  Use that
instead.

Cc: Greg Suarez <gsuarez@smithmicro.com>
Cc: Alexey Orishko <alexey.orishko@stericsson.com>
Signed-off-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: David S. Miller <davem@davemloft.net>

Included changes ares:
- fix an skb memleak in DAT
- fix the ARP filtering routine in DAT by preventing bogus entries to overwrite
  already existing ones in the local cache.
- fix the ARP filtering routine in DAT by preventing it to parse and add to the
  cache bogus entries
Signed-off-by: David S. Miller <davem@davemloft.net>

We never want multicast MAC addresses in the Distributed ARP Table, so it's
best to completely ignore ARP packets containing them where we expect unicast
addresses.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
Acked-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
Signed-off-by: Antonio Quartulli <ordex@autistici.org>

There are more types of IP addresses that may appear in ARP packets that we
don't want to process. While some of these should never appear in sane ARP
packets, a 0.0.0.0 source is used for duplicate address detection and thus seen
quite often.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
Acked-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
Signed-off-by: Antonio Quartulli <ordex@autistici.org>

The callers of batadv_dat_snoop_incoming_arp_reply() assume the skb has been
freed when it returns true; fix this by calling kfree_skb before returning as
it is done in batadv_dat_snoop_incoming_arp_request().
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
Acked-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Antonio Quartulli <ordex@autistici.org>

John W. Linville says:

====================
This is a batch of fixes intende for the 3.8 stream.

Regarding the iwlwifi bits, Johannes says this:

"Please pull to get a single fix from Emmanuel for a bug I introduced due
to misunderstanding the code."

Regarding the mac80211 bits, Johannes says this:

"I have a few small fixes for you:
 * some mesh frames would cause encryption warnings -- fixes from Bob
 * scanning would pretty much break an association if we transmitted
   anything to the AP while scanning -- fix from Stanislaw
 * mode injection was broken by channel contexts -- fix from Felix
 * FT roaming was broken: hardware crypto would get disabled by it"

Along with that, a handful of other fixes confined to specific drivers.

Avinash Patil fixes a typo in a NULL check in mwifiex.

Larry Finger fixes a build warning in rtlwifi.  Seems safe...

Stanislaw Gruszka fixes iwlegacy to prevent microcode errors when
switching from IBSS mode to STA mode.

Felix Fietkau provides a trio of ath9k fixes related to proper tuning.

Please let me know if there are problems!
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Ben Greear reported crashes in ip_rcv_finish() on a stress
test involving many macvlans.

We tracked the bug to a dst use after free. ip_rcv_finish()
was calling dst->input() and got garbage for dst->input value.

It appears the bug is in loopback driver, lacking
a skb_dst_force() before calling netif_rx().

As a result, a non refcounted dst, normally protected by a
RCU read_lock section, was escaping this section and could
be freed before the packet being processed.

  [<ffffffff813a3c4d>] loopback_xmit+0x64/0x83
  [<ffffffff81477364>] dev_hard_start_xmit+0x26c/0x35e
  [<ffffffff8147771a>] dev_queue_xmit+0x2c4/0x37c
  [<ffffffff81477456>] ? dev_hard_start_xmit+0x35e/0x35e
  [<ffffffff8148cfa6>] ? eth_header+0x28/0xb6
  [<ffffffff81480f09>] neigh_resolve_output+0x176/0x1a7
  [<ffffffff814ad835>] ip_finish_output2+0x297/0x30d
  [<ffffffff814ad6d5>] ? ip_finish_output2+0x137/0x30d
  [<ffffffff814ad90e>] ip_finish_output+0x63/0x68
  [<ffffffff814ae412>] ip_output+0x61/0x67
  [<ffffffff814ab904>] dst_output+0x17/0x1b
  [<ffffffff814adb6d>] ip_local_out+0x1e/0x23
  [<ffffffff814ae1c4>] ip_queue_xmit+0x315/0x353
  [<ffffffff814adeaf>] ? ip_send_unicast_reply+0x2cc/0x2cc
  [<ffffffff814c018f>] tcp_transmit_skb+0x7ca/0x80b
  [<ffffffff814c3571>] tcp_connect+0x53c/0x587
  [<ffffffff810c2f0c>] ? getnstimeofday+0x44/0x7d
  [<ffffffff810c2f56>] ? ktime_get_real+0x11/0x3e
  [<ffffffff814c6f9b>] tcp_v4_connect+0x3c2/0x431
  [<ffffffff814d6913>] __inet_stream_connect+0x84/0x287
  [<ffffffff814d6b38>] ? inet_stream_connect+0x22/0x49
  [<ffffffff8108d695>] ? _local_bh_enable_ip+0x84/0x9f
  [<ffffffff8108d6c8>] ? local_bh_enable+0xd/0x11
  [<ffffffff8146763c>] ? lock_sock_nested+0x6e/0x79
  [<ffffffff814d6b38>] ? inet_stream_connect+0x22/0x49
  [<ffffffff814d6b49>] inet_stream_connect+0x33/0x49
  [<ffffffff814632c6>] sys_connect+0x75/0x98

This bug was introduced in linux-2.6.35, in commit
7fee226a (net: add a noref bit on skb dst)

skb_dst_force() is enforced in dev_queue_xmit() for devices having a
qdisc.
Reported-by: Ben Greear <greearb@candelatech.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Tested-by: Ben Greear <greearb@candelatech.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add a cpu notifier to virtio-net, so that we can reset the
virtqueue affinity if the cpu hotplug happens. It improve
the performance through enabling or disabling the virtqueue
affinity after doing cpu hotplug.

Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: Eric Dumazet <erdnetdev@gmail.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: virtualization@lists.linux-foundation.org
Cc: netdev@vger.kernel.org
Signed-off-by: Wanlong Gao <gaowanlong@cn.fujitsu.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Split out the clean affinity function to virtnet_clean_affinity().

Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: Eric Dumazet <erdnetdev@gmail.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: virtualization@lists.linux-foundation.org
Cc: netdev@vger.kernel.org
Signed-off-by: Wanlong Gao <gaowanlong@cn.fujitsu.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

As Michael mentioned, set affinity and select queue will not work very
well when CPU IDs are not consecutive, this can happen with hot unplug.
Fix this bug by traversal the online CPUs, and create a per cpu variable
to find the mapping from CPU to the preferable virtual-queue.

Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: Eric Dumazet <erdnetdev@gmail.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: virtualization@lists.linux-foundation.org
Cc: netdev@vger.kernel.org
Signed-off-by: Wanlong Gao <gaowanlong@cn.fujitsu.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>