On a high-traffic router with many processors and many IPv6 dst
entries, soft lockup in fib6_run_gc() can occur when number of
entries reaches gc_thresh.

This happens because fib6_run_gc() uses fib6_gc_lock to allow
only one thread to run the garbage collector but ip6_dst_gc()
doesn't update net->ipv6.ip6_rt_last_gc until fib6_run_gc()
returns. On a system with many entries, this can take some time
so that in the meantime, other threads pass the tests in
ip6_dst_gc() (ip6_rt_last_gc is still not updated) and wait for
the lock. They then have to run the garbage collector one after
another which blocks them for quite long.

Resolve this by replacing special value ~0UL of expire parameter
to fib6_run_gc() by explicit "force" parameter to choose between
spin_lock_bh() and spin_trylock_bh() and call fib6_run_gc() with
force=false if gc_thresh is reached but not max_size.
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>

John W. Linville says:

====================
This pull request is intended for the 3.11 stream.  It is a bit
larger than usual, as it includes pulls from most of my feeder trees
as well...

For the Bluetooth bits, Gustavo says:

"A few fixes and devices ID additions for 3.11:

 * There are 4 new ath3k device ids
 * Fixed stack memory usage in ath3k.
 * Fixed the init process of BlueFRITZ! devices, they were failing to init
   due to an unsupported command we sent.
 * Fixed wrong use of PTR_ERR in btusb code that was preventing intel devices
   to work properly.
 * Fixed race condition between hci_register_dev() and hci_dev_open() that
   could cause a NULL pointer dereference.
 * Fixed race condition that could call hci_req_cmd_complete() and make some
   devices to fail as showed in the log added to the commit message."

Regarding the NFC bits, Samuel says:

"We have:

1) A build failure fix for the NCI SPI transport layer due to a
   missing CRC_CCITT Kconfig dependency.

2) A netlink command rename: CMD_FW_UPLOAD was merged during the 3.11
   merge window but the typical terminology for loading a firmware to a
   target is firmware download rather than upload. In order to avoid any
   confusion in a file exported to userspace, we rename this command to
   CMD_FW_DOWNLOAD."

Samuel's item #2 isn't strictly a fix, but it seems safe and should
avoid confusion in the future.

As for the mac80211 bits, Johannes says:

"I only have three fixes this time, a fix for a suspend regression, a
patch correcting the initiator in regulatory code and one fix for mesh
station powersave."

With respect to the iwlwifi bits, Johannes says:

"We have a scan fix for passive channels, a new PCI device ID for an old
device, a NIC reset fix, an RF-Kill fix, a fix for powersave when GO
interfaces are present as well as an aggregation session fix (for a
corner case) and a workaround for a firmware design issue - it only
supports a single GTK in D3."

Bringing-up the rear with the Atheros trees, Kalle says:

"Geert Uytterhoeven fixed an ath10k build problem when NO_DMA=y. I added
a missing MAINTAINERS entry for ath10k and updated ath6kl git tree
location."

Along with the above...

Arend van Spriel fixes a brcmfmac WARNING when unplugging the device.

Avinash Patil proves a couple of minor mwifiex fixes relating to P2P mode.

Luciano Coelho updates the MAINTAINERS entry for the wilink drivers.

Stanislaw Gruszka brings an rt2x00 fix for a queue start/stop problem.

Stone Piao fixes another mwifiex problem, a command timeout related to P2P mode.

Tomasz Moń corrects an endian problem in mwifiex.

I'll remind my feeder maintainers to slowdown the patchflow.
Beyond that, please let me know if there are problems!
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless into for-davem

If there is no querier on a link then we won't get periodic reports and
therefore won't be able to learn about multicast listeners behind ports,
potentially leading to lost multicast packets, especially for multicast
listeners that joined before the creation of the bridge.

These lost multicast packets can appear since c5c23260
("bridge: Add multicast_querier toggle and disable queries by default")
in particular.

With this patch we are flooding multicast packets if our querier is
disabled and if we didn't detect any other querier.

A grace period of the Maximum Response Delay of the querier is added to
give multicast responses enough time to arrive and to be learned from
before disabling the flooding behaviour again.
Signed-off-by: Linus Lüssing <linus.luessing@web.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

Self explanitory dma_mapping_error addition to the 8139 driver, based on this:
https://bugzilla.redhat.com/show_bug.cgi?id=947250

It showed several backtraces arising for dma_map_* usage without checking the
return code on the mapping.  Add the check and abort the rx/tx operation if its
failed.  Untested as I have no hardware and the reporter has wandered off, but
seems pretty straightforward.
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
CC: "David S. Miller" <davem@davemloft.net>
CC: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The "pvc" struct has a hole after pvc.sap_family which is not cleared.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>

- fix the conversion between cpu and __le32
 - replace some pla_ocp and usb_ocp functions with generic_ocp function
Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Allocate the required memory before calling usb_control_msg. And
the additional memory copy is necessary.
Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Replace 0 with the result from usbnet_cdc_bind().
Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Don't replace the usb_control_msg() with usbnet_{read,write}_cmd()
which couldn't be called inside suspend/resume callback. Keep the
basic functions unlimited. Instead, using usb_autopm_get_interface()
and usb_autopm_put_interface() in r815x_mdio_{read,write}().
Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Some USB buffers use stack which may not be DMA-able.
Use the buffers from kmalloc to replace those one.
Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Pull networking fixes from David Miller:

 1) Fix association failures not triggering a connect-failure event in
    cfg80211, from Johannes Berg.

 2) Eliminate a potential NULL deref with older iptables tools when
    configuring xt_socket rules, from Eric Dumazet.

 3) Missing RTNL locking in wireless regulatory code, from Johannes
    Berg.

 4) Fix OOPS caused by firmware loading races in ath9k_htc, from Alexey
    Khoroshilov.

 5) Fix usb URB leak in usb_8dev CAN driver, also from Alexey
    Khoroshilov.

 6) VXLAN namespace teardown fails to unregister devices, from Stephen
    Hemminger.

 7) Fix multicast settings getting dropped by firmware in qlcnic driver,
    from Sucheta Chakraborty.

 8) Add sysctl range enforcement for tcp_syn_retries, from Michal Tesar.

 9) Fix a nasty bug in bridging where an active timer would get
    reinitialized with a setup_timer() call.  From Eric Dumazet.

10) Fix use after free in new mlx5 driver, from Dan Carpenter.

11) Fix freed pointer reference in ipv6 multicast routing on namespace
    cleanup, from Hannes Frederic Sowa.

12) Some usbnet drivers report TSO and SG in their feature set, but the
    usbnet layer doesn't really support them.  From Eric Dumazet.

13) Fix crash on EEH errors in tg3 driver, from Gavin Shan.

14) Drop cb_lock when requesting modules in genetlink, from Stanislaw
    Gruszka.

15) Kernel stack leaks in cbq scheduler and af_key pfkey messages, from
    Dan Carpenter.

16) FEC driver erroneously signals NETDEV_TX_BUSY on transmit leading to
    endless loops, from Uwe Kleine-König.

17) Fix hangs from loading mvneta driver, from Arnaud Patard.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (84 commits)
  mlx5: fix error return code in mlx5_alloc_uuars()
  mvneta: Try to fix mvneta when compiled as module
  mvneta: Fix hang when loading the mvneta driver
  atl1c: Fix misuse of netdev_alloc_skb in refilling rx ring
  genetlink: fix usage of NLM_F_EXCL or NLM_F_REPLACE
  af_key: more info leaks in pfkey messages
  net/fec: Don't let ndo_start_xmit return NETDEV_TX_BUSY without link
  net_sched: Fix stack info leak in cbq_dump_wrr().
  igb: fix vlan filtering in promisc mode when not in VT mode
  ixgbe: Fix Tx Hang issue with lldpad on 82598EB
  genetlink: release cb_lock before requesting additional module
  net: fec: workaround stop tx during errata ERR006358
  qlcnic: Fix diagnostic interrupt test for 83xx adapters.
  qlcnic: Fix setting Guest VLAN
  qlcnic: Fix operation type and command type.
  qlcnic: Fix initialization of work function.
  Revert "atl1c: Fix misuse of netdev_alloc_skb in refilling rx ring"
  atl1c: Fix misuse of netdev_alloc_skb in refilling rx ring
  net/tg3: Fix warning from pci_disable_device()
  net/tg3: Fix kernel crash
  ...

Samuel Ortiz <sameo@linux.intel.com> says:

'This is the second NFC fixes pull request for 3.11.

We have:

- A build failure fix for the NCI SPI transport layer due to a
  missing CRC_CCITT Kconfig dependency.

- A netlink command rename: CMD_FW_UPLOAD was merged during the 3.11
  merge window but the typical terminology for loading a firmware to a
  target is firmware download rather than upload. In order to avoid any
  confusion in a file exported to userspace, we rename this command into
  CMD_FW_DOWNLOAD."
Signed-off-by: John W. Linville <linville@tuxdriver.com>

Conflicts:
	net/bluetooth/hci_core.c

Soon the coelho@ti.com email will not be valid anymore, so change it
to my private one.

Cc: Luciano Coelho <luca@coelho.fi>
Signed-off-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

We missed bss_mode check for P2P client.
Signed-off-by: Stone Piao <piaoyun@marvell.com>
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

This patch fixes an issue wherein adhoc rates were being copied
into association request from P2P client.

Cc: <stable@vger.kernel.org> # 3.10.y
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Stone Piao <piaoyun@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

This patch fixes an issue wherein association would fail on P2P
interfaces. This happened because we are checking priv->mode
against NL80211_IFTYPE_STATION. While this check is correct for
infrastructure stations, it would fail P2P clients for which mode
is NL80211_IFTYPE_P2P_CLIENT.

Better check would be bss_role which has only 2 values: STA/AP.

Cc: <stable@vger.kernel.org> # 3.10.y
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Stone Piao <piaoyun@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

Pull MIPS fixes from Ralf Baechle:
 - BMIPS SMP fixes
 - a build fix necessary for older compilers
 - two more bugs found my Chandras' testing
 - and one more build fix

* 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus:
  MIPS: BMIPS: fix slave CPU booting when physical CPU is not 0
  MIPS: BMIPS: do not change interrupt routing depending on boot CPU
  MIPS: powertv: Fix arguments for free_reserved_area()
  MIPS: Set default CPU type for BCM47XX platforms
  MIPS: uapi/asm/siginfo.h: Fix GCC 4.1.2 compilation
  MIPS: Fix multiple definitions of UNCAC_BASE.

Pull Xen fixes from Konrad Rzeszutek Wilk:
 - Three fixes for ARM/ARM64 to either compile or not certain generic
   drivers
 - Fix for avoiding a potential deadlock when an user space event
   channel is destroyed.
 - Fix a workqueue resuming multiple times.

* tag 'stable/for-linus-3.11-rc3-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
  xen/tmem: do not allow XEN_TMEM on ARM64
  xen/evtchn: avoid a deadlock when unbinding an event channel
  xen/arm: enable PV control for ARM
  xen/arm64: Don't compile cpu hotplug
  xenbus: frontend resume cleanup

Merge tag 'xen-arm-3.11-rc2-warn-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/sstabellini/xen

Pull Xen ARM fix from Stefano Stabellini.

Update xen_restart to new calling convention.

* tag 'xen-arm-3.11-rc2-warn-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/sstabellini/xen:
  xen/arm,arm64: update xen_restart after ff701306 and 7b6d864b

Pull USB fixes from Greg KH:
 "Here are some tiny USB fixes for 3.11-rc4

  Nothing major, some gadget fixes, some new device ids, a new tiny
  driver for the ANT+ USB device, and a number of fixes for the mos7840
  driver that were much needed"

* tag 'usb-3.11-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  USB: serial: ftdi_sio: add more RT Systems ftdi devices
  usb: chipidea: fix the build error with randconfig
  usb: chipidea: cast PORTSC_PTS and DEVLC_PTS macros
  usb: gadget: udc-core: fix the typo of udc state attribute
  usb: gadget: f_phonet: remove unused preprocessor conditional
  usb: gadget: multi: fix error return code in cdc_do_config()
  USB: mos7840: fix pointer casts
  USB: mos7840: fix race in led handling
  USB: mos7840: fix device-type detection
  USB: mos7840: fix race in register handling
  USB: serial: add driver for Suunto ANT+ USB device
  usb: gadget: free opts struct on error recovery
  usb: gadget: ether: put_usb_function on unbind
  usb: musb: fix resource passed from glue layer to musb

Pull tty/serial driver fixes from Greg KH:
 "Here are 4 tiny tty and serial driver fixes for 3.11-rc4.

  Nothing big, a refcount leak, a module alias fix, and two fixes to the
  mxs-auart serial driver"

* tag 'tty-3.11-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  serial: arc_uart: Fix module alias
  tty_port: Fix refcounting leak in tty_port_tty_hangup()
  serial/mxs-auart: increase time to wait for transmitter to become idle
  serial/mxs-auart: fix race condition in interrupt handler

Fix to return -ENOMEM from the ioremap error handling
case instead of 0, as done elsewhere in this function.
Signed-off-by: Wei Yongjun <yongjun_wei@trendmicro.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>

When the mvneta driver is compiled as module, the clock is disabled before
it's loading. This will reset the registers values and all configuration
made by the bootloader.

This patch sets the "sgmii serdes configuration" register to a magical value
found in:
https://github.com/yellowback/ubuntu-precise-armadaxp/blob/master/arch/arm/mach-armadaxp/armada_xp_family/ctrlEnv/mvCtrlEnvLib.c

With this change, the interrupts are working/generated and ethernet is
working.
Signed-off-by: Arnaud Patard <arnaud.patard@rtp-net.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

When the mvneta driver is compiled, it'll be loaded with clocks disabled.
This implies that the clocks should be enabled again before any register
access or it'll hang.

To fix it:
- enable clock earlier
- move timer callback after setting timer.data
Signed-off-by: Arnaud Patard <arnaud.patard@rtp-net.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

On Mon, 2013-07-29 at 08:30 -0700, Eric Dumazet wrote:
> On Mon, 2013-07-29 at 13:09 +0100, Luis Henriques wrote:
>
> >
> > I confirm that I can't reproduce the issue using this patch.
> >
>
> Thanks, I'll send a polished patch, as this one had an error if
> build_skb() returns NULL (in case sk_buff allocation fails)

Please try the following patch : It should use 2K frags instead of 4K
for normal 1500 mtu

Thanks !

[PATCH] atl1c: use custom skb allocator

We had reports ( https://bugzilla.kernel.org/show_bug.cgi?id=54021 )
that using high order pages for skb allocations is problematic for atl1c

We do not know exactly what the problem is, but we suspect that crossing
4K pages is not well supported by this hardware.

Use a custom allocator, using page allocator and 2K fragments for
optimal stack behavior. We might make this allocator generic
in future kernels.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Luis Henriques <luis.henriques@canonical.com>
Cc: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Currently, it is not possible to use neither NLM_F_EXCL nor
NLM_F_REPLACE from genetlink. This is due to this checking in
genl_family_rcv_msg:

	if (nlh->nlmsg_flags & NLM_F_DUMP)

NLM_F_DUMP is NLM_F_MATCH|NLM_F_ROOT. Thus, if NLM_F_EXCL or
NLM_F_REPLACE flag is set, genetlink believes that you're
requesting a dump and it calls the .dumpit callback.

The solution that I propose is to refine this checking to
make it stricter:

	if ((nlh->nlmsg_flags & NLM_F_DUMP) == NLM_F_DUMP)

And given the combination NLM_F_REPLACE and NLM_F_EXCL does
not make sense to me, it removes the ambiguity.

There was a patch that tried to fix this some time ago (0ab03c2b
netlink: test for all flags of the NLM_F_DUMP composite) but it
tried to resolve this ambiguity in *all* existing netlink subsystems,
not only genetlink. That patch was reverted since it broke iproute2,
which is using NLM_F_ROOT to request the dump of the routing cache.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

This is inspired by a5cc68f3 "af_key: fix info leaks in notify
messages".  There are some struct members which don't get initialized
and could disclose small amounts of private information.
Acked-by: Mathias Krause <minipli@googlemail.com>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Loading a firmware into a target is typically called firmware
download, not firmware upload. So we rename the netlink API to
NFC_CMD_FW_DOWNLOAD in order to avoid any terminology confusion from
userspace.
Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>

Don't test for having link and let hardware deal with this situation.

Without this patch I see a machine running an -rt patched Linux being
stuck in sch_direct_xmit when it looses link while there is still a
packet to be sent. In this case the fec_enet_start_xmit routine returned
NETDEV_TX_BUSY which makes the network stack reschedule the packet and
so sch_direct_xmit calls fec_enet_start_xmit again.
I failed to reproduce a complete hang without -rt, but I think the
problem exists there, too.
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

tmem is not supported on arm or arm64 yet. Will revert this
once the Xen hypervisor supports it.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>

The current BMIPS SMP code assumes that the slave CPU is physical and
logical CPU 1, but on some systems such as BCM3368, the slave CPU is
physical CPU0. Fix the code to read the physical CPU (thread ID) we are
running this code on, and adjust the relocation vector address based on
it. This allows bringing up the second CPU on BCM3368 for instance.
Signed-off-by: Florian Fainelli <florian@openwrt.org>
Cc: linux-mips@linux-mips.org
Cc: cernekee@gmail.com
Cc: jogo@openwrt.org
Cc: blogic@openwrt.org
Patchwork: https://patchwork.linux-mips.org/patch/5621/Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Commit 4df715aa ("MIPS: BMIPS: support booting from physical CPU other
than 0") changed the interupt routing when we are booting from physical
CPU 0, but the settings are actually correct if we are booting from
physical CPU 0 or CPU 1. Revert that specific change.
Signed-off-by: Florian Fainelli <florian@openwrt.org>
Cc: linux-mips@linux-mips.org
Cc: cernekee@gmail.com
Cc: jogo@openwrt.org
Cc: blogic@openwrt.org
Patchwork: https://patchwork.linux-mips.org/patch/5622/Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Commit 6e7582bf
"MIPS: PowerTV: use free_reserved_area() to simplify code"

merged in 3.11-rc1, broke the build for the powertv defconfig with
the following build error:

arch/mips/powertv/asic/asic_devices.c: In function 'platform_release_memory':
arch/mips/powertv/asic/asic_devices.c:533:7: error: passing argument 1 of
'free_reserved_area' makes pointer from integer without a cast [-Werror]

The free_reserved_area() function expects a void * pointer for the start
address and a void * pointer for the end one.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/5624/Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

If neither BCM47XX_SSD nor BCM47XX_BCMA is selected, then no
CPU type is available leading to build problems. We fix
this problem by using MIPS32r1 as the default CPU type for
the BCM47XX platform.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Acked-by: Steven J. Hill <Steven.Hill@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/5618/Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

It wasn't until GCC 4.3 I believe that the __SIZEOF_*__ predefined macros
were added.  The change below switches <uapi/asm/siginfo.h> to the
_MIPS_SZLONG macro so that compilation with e.g. GCC 4.1.2 succeeds.
This is a user API header so I think this is even more important, for
older userland support.  The change adds an unsuccessful default too, to
catch any compiler configuration oddities.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/5630/Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Fix build error below:

arch/mips/include/asm/mach-generic/spaces.h:29:0: warning:
"UNCAC_BASE" redefined [enabled by default]
In file included from arch/mips/include/asm/addrspace.h:13:0,
                 from arch/mips/include/asm/barrier.h:11,
                 from arch/mips/include/asm/bitops.h:18,
                 from include/linux/bitops.h:22,
                 from include/linux/kernel.h:10,
                 from include/asm-generic/bug.h:13,
                 from arch/mips/include/asm/bug.h:41,
                 from include/linux/bug.h:4,
                 from include/linux/page-flags.h:9,
                 from kernel/bounds.c:9:
arch/mips/include/asm/mach-ar7/spaces.h:20:0: note: this is the
location of the previous definition
Signed-off-by: Steven J. Hill <Steven.Hill@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/5583/Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Unbinding an event channel (either with the ioctl or when the evtchn
device is closed) may deadlock because disable_irq() is called with
port_user_lock held which is also locked by the interrupt handler.

Think of the IOCTL_EVTCHN_UNBIND is being serviced, the routine has
just taken the lock, and an interrupt happens. The evtchn_interrupt
is invoked, tries to take the lock and spins forever.

A quick glance at the code shows that the spinlock is a local IRQ
variant. Unfortunately that does not help as "disable_irq() waits for
the interrupt handler on all CPUs to stop running.  If the irq occurs
on another VCPU, it tries to take port_user_lock and can't because
the unbind ioctl is holding it." (from David). Hence we cannot
depend on the said spinlock to protect us. We could make it a system
wide IRQ disable spinlock but there is a better way.

We can piggyback on the fact that the existence of the spinlock is
to make get_port_user() checks be up-to-date. And we can alter those
checks to not depend on the spin lock (as it's protected by u->bind_mutex
in the ioctl) and can remove the unnecessary locking (this is
IOCTL_EVTCHN_UNBIND) path.

In the interrupt handler we cannot use the mutex, but we do not
need it.

"The unbind disables the irq before making the port user stale, so when
you clear it you are guaranteed that the interrupt handler that might
use that port cannot be running." (from David).

Hence this patch removes the spinlock usage on the teardown path
and piggybacks on disable_irq happening before we muck with the
get_port_user() data. This ensures that the interrupt handler will
never run on stale data.
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
[v1: Expanded the commit description a bit]