firefly-linux-kernel-4.4.55.git
10 years agobnx2x: fix possible panic under memory stress
Eric Dumazet [Thu, 26 Jun 2014 07:44:02 +0000 (00:44 -0700)]
bnx2x: fix possible panic under memory stress

While it is legal to kfree(NULL), it is not wise to use :
put_page(virt_to_head_page(NULL))

 BUG: unable to handle kernel paging request at ffffeba400000000
 IP: [<ffffffffc01f5928>] virt_to_head_page+0x36/0x44 [bnx2x]

Reported-by: Michel Lespinasse <walken@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Ariel Elior <ariel.elior@qlogic.com>
Fixes: d46d132cc021 ("bnx2x: use netdev_alloc_frag()")
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipv4: irq safe sk_dst_[re]set() and ipv4_sk_update_pmtu() fix
Eric Dumazet [Mon, 30 Jun 2014 08:26:23 +0000 (01:26 -0700)]
ipv4: irq safe sk_dst_[re]set() and ipv4_sk_update_pmtu() fix

We have two different ways to handle changes to sk->sk_dst

First way (used by TCP) assumes socket lock is owned by caller, and use
no extra lock : __sk_dst_set() & __sk_dst_reset()

Another way (used by UDP) uses sk_dst_lock because socket lock is not
always taken. Note that sk_dst_lock is not softirq safe.

These ways are not inter changeable for a given socket type.

ipv4_sk_update_pmtu(), added in linux-3.8, added a race, as it used
the socket lock as synchronization, but users might be UDP sockets.

Instead of converting sk_dst_lock to a softirq safe version, use xchg()
as we did for sk_rx_dst in commit e47eb5dfb296b ("udp: ipv4: do not use
sk_dst_lock from softirq context")

In a follow up patch, we probably can remove sk_dst_lock, as it is
only used in IPv6.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Steffen Klassert <steffen.klassert@secunet.com>
Fixes: 9cb3a50c5f63e ("ipv4: Invalidate the socket cached route on pmtu events if possible")
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agopowerpc: bpf: Fix the broken LD_VLAN_TAG_PRESENT test
Denis Kirjanov [Wed, 25 Jun 2014 17:34:57 +0000 (21:34 +0400)]
powerpc: bpf: Fix the broken LD_VLAN_TAG_PRESENT test

We have to return the boolean here if the tag presents
or not, not just ANDing the TCI with the mask which results to:

[  709.412097] test_bpf: #18 LD_VLAN_TAG_PRESENT
[  709.412245] ret 4096 != 1
[  709.412332] ret 4096 != 1
[  709.412333] FAIL (2 times)

Signed-off-by: Denis Kirjanov <kda@linux-powerpc.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agopowerpc: bpf: Use correct mask while accessing the VLAN tag
Denis Kirjanov [Wed, 25 Jun 2014 17:34:56 +0000 (21:34 +0400)]
powerpc: bpf: Use correct mask while accessing the VLAN tag

To get a full tag (and not just a VID) we should access the TCI
except the VLAN_TAG_PRESENT field (which means that 802.1q header
is present). Also ensure that the VLAN_TAG_PRESENT stay on its place

Signed-off-by: Denis Kirjanov <kda@linux-powerpc.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipv6: Fix MLD Query message check
Hangbin Liu [Fri, 27 Jun 2014 01:57:53 +0000 (09:57 +0800)]
ipv6: Fix MLD Query message check

Based on RFC3810 6.2, we also need to check the hop limit and router alert
option besides source address.

Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Acked-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoudp: Add MIB counters for rcvbuferrors
James M Leddy [Wed, 25 Jun 2014 21:38:13 +0000 (17:38 -0400)]
udp: Add MIB counters for rcvbuferrors

Add MIB counters for rcvbuferrors in UDP to help diagnose problems.

Signed-off-by: James M Leddy <james.leddy@redhat.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobnx2x: Fix the MSI flags
Yijing Wang [Wed, 25 Jun 2014 04:22:56 +0000 (12:22 +0800)]
bnx2x: Fix the MSI flags

MSI-X should use PCI_MSIX_FLAGS not PCI_MSI_FLAGS.

Signed-off-by: Yijing Wang <wangyijing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'for-next' of git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Thu, 26 Jun 2014 04:47:28 +0000 (21:47 -0700)]
Merge branch 'for-next' of git://git.samba.org/sfrench/cifs-2.6

Pull CIFS fixes from Steve French:
 "Small set of misc cifs/smb3 fixes"

* 'for-next' of git://git.samba.org/sfrench/cifs-2.6:
  [CIFS] fix mount failure with broken pathnames when smb3 mount with mapchars option
  cifs: revalidate mapping prior to satisfying read_iter request with cache=loose
  fs/cifs: fix regression in cifs_create_mf_symlink()

10 years agoMerge tag 'hwmon-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck...
Linus Torvalds [Thu, 26 Jun 2014 04:38:45 +0000 (21:38 -0700)]
Merge tag 'hwmon-for-linus' of git://git./linux/kernel/git/groeck/linux-staging

Pull hwmon fixes from Guenter Roeck:
 "Various minor fixes"

* tag 'hwmon-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck/linux-staging:
  hwmon: (gpio-fan) Change name used in hwmon_device_register_with_groups
  hwmon: (emc1403) Fix missing 'select REGMAP_I2C' in Kconfig
  hwmon: (ntc_thermistor) Use the manufacturer name properly
  devicetree: bindings: Document murata vendor prefix
  hwmon: (w83l786ng) Report correct minimum fan speed

10 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Thu, 26 Jun 2014 04:08:24 +0000 (21:08 -0700)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) Fix crash in ipvs tot_stats estimator, from Julian Anastasov.

 2) Fix OOPS in nf_nat on netns removal, from Florian Westphal.

 3) Really really really fix locking issues in slip and slcan tty write
    wakeups, from Tyler Hall.

 4) Fix checksum offloading in fec driver, from Fugang Duan.

 5) Off by one in BPF instruction limit test, from Kees Cook.

 6) Need to clear all TSO capability flags when doing software TSO in
    tg3 driver, from Prashant Sreedharan.

 7) Fix memory leak in vlan_reorder_header() error path, from Li
    RongQing.

 8) Fix various bugs in xen-netfront and xen-netback multiqueue support,
    from David Vrabel and Wei Liu.

 9) Fix deadlock in cxgb4 driver, from Li RongQing.

10) Prevent double free of no-cache DST entries, from Eric Dumazet.

11) Bad csum_start handling in skb_segment() leads to crashes when
    forwarding, from Tom Herbert.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (76 commits)
  net: fix setting csum_start in skb_segment()
  ipv4: fix dst race in sk_dst_get()
  net: filter: Use kcalloc/kmalloc_array to allocate arrays
  trivial: net: filter: Change kerneldoc parameter order
  trivial: net: filter: Fix typo in comment
  net: allwinner: emac: Add missing free_irq
  cxgb4: use dev_port to identify ports
  xen-netback: bookkeep number of active queues in our own module
  tg3: Change nvram command timeout value to 50ms
  cxgb4: Not need to hold the adap_rcu_lock lock when read adap_rcu_list
  be2net: fix qnq mode detection on VFs
  of: mdio: fixup of_phy_register_fixed_link parsing of new bindings
  at86rf230: fix irq setup
  net: phy: at803x: fix coccinelle warnings
  net/mlx4_core: Fix the error flow when probing with invalid VF configuration
  tulip: Poll link status more frequently for Comet chips
  net: huawei_cdc_ncm: increase command buffer size
  drivers: net: cpsw: fix dual EMAC stall when connected to same switch
  xen-netfront: recreate queues correctly when reconnecting
  xen-netfront: fix oops when disconnected from backend
  ...

10 years agonet: fix setting csum_start in skb_segment()
Tom Herbert [Wed, 25 Jun 2014 19:51:01 +0000 (12:51 -0700)]
net: fix setting csum_start in skb_segment()

Dave Jones reported that a crash is occurring in

csum_partial
tcp_gso_segment
inet_gso_segment
? update_dl_migration
skb_mac_gso_segment
__skb_gso_segment
dev_hard_start_xmit
sch_direct_xmit
__dev_queue_xmit
? dev_hard_start_xmit
dev_queue_xmit
ip_finish_output
? ip_output
ip_output
ip_forward_finish
ip_forward
ip_rcv_finish
ip_rcv
__netif_receive_skb_core
? __netif_receive_skb_core
? trace_hardirqs_on
__netif_receive_skb
netif_receive_skb_internal
napi_gro_complete
? napi_gro_complete
dev_gro_receive
? dev_gro_receive
napi_gro_receive

It looks like a likely culprit is that SKB_GSO_CB()->csum_start is
not set correctly when doing non-scatter gather. We are using
offset as opposed to doffset.

Reported-by: Dave Jones <davej@redhat.com>
Tested-by: Dave Jones <davej@redhat.com>
Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Fixes: 7e2b10c1e52ca ("net: Support for multiple checksums with gso")
Acked-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge tag 'nfs-for-3.16-2' of git://git.linux-nfs.org/projects/trondmy/linux-nfs
Linus Torvalds [Thu, 26 Jun 2014 03:06:06 +0000 (20:06 -0700)]
Merge tag 'nfs-for-3.16-2' of git://git.linux-nfs.org/projects/trondmy/linux-nfs

Pull NFS client fixes from Trond Myklebust:
 "Highlights include:

   - Stable fix for a data corruption case due to incorrect cache
     validation
   - Fix a couple of false positive cache invalidations
   - Fix NFSv4 security negotiation issues"

* tag 'nfs-for-3.16-2' of git://git.linux-nfs.org/projects/trondmy/linux-nfs:
  NFSv4: test SECINFO RPC_AUTH_GSS pseudoflavors for support
  NFS Return -EPERM if no supported or matching SECINFO flavor
  NFS check the return of nfs4_negotiate_security in nfs4_submount
  NFS: Don't mark the data cache as invalid if it has been flushed
  NFS: Clear NFS_INO_REVAL_PAGECACHE when we update the file size
  nfs: Fix cache_validity check in nfs_write_pageuptodate()

10 years agoipv4: fix dst race in sk_dst_get()
Eric Dumazet [Tue, 24 Jun 2014 17:05:11 +0000 (10:05 -0700)]
ipv4: fix dst race in sk_dst_get()

When IP route cache had been removed in linux-3.6, we broke assumption
that dst entries were all freed after rcu grace period. DST_NOCACHE
dst were supposed to be freed from dst_release(). But it appears
we want to keep such dst around, either in UDP sockets or tunnels.

In sk_dst_get() we need to make sure dst refcount is not 0
before incrementing it, or else we might end up freeing a dst
twice.

DST_NOCACHE set on a dst does not mean this dst can not be attached
to a socket or a tunnel.

Then, before actual freeing, we need to observe a rcu grace period
to make sure all other cpus can catch the fact the dst is no longer
usable.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Dormando <dormando@rydia.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: filter: Use kcalloc/kmalloc_array to allocate arrays
Tobias Klauser [Tue, 24 Jun 2014 13:33:22 +0000 (15:33 +0200)]
net: filter: Use kcalloc/kmalloc_array to allocate arrays

Use kcalloc/kmalloc_array to make it clear we're allocating arrays. No
integer overflow can actually happen here, since len/flen is guaranteed
to be less than BPF_MAXINSNS (4096). However, this changed makes sure
we're not going to get one if BPF_MAXINSNS were ever increased.

Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Acked-by: Daniel Borkmann <dborkman@redhat.com>
Acked-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotrivial: net: filter: Change kerneldoc parameter order
Tobias Klauser [Tue, 24 Jun 2014 13:33:21 +0000 (15:33 +0200)]
trivial: net: filter: Change kerneldoc parameter order

Change the order of the parameters to sk_unattached_filter_create() in
the kerneldoc to reflect the order they appear in the actual function.

This fix is only cosmetic, in the generated doc they still appear in the
correct order without the fix.

Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Acked-by: Daniel Borkmann <dborkman@redhat.com>
Acked-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotrivial: net: filter: Fix typo in comment
Tobias Klauser [Tue, 24 Jun 2014 13:33:20 +0000 (15:33 +0200)]
trivial: net: filter: Fix typo in comment

Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Acked-by: Daniel Borkmann <dborkman@redhat.com>
Acked-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: allwinner: emac: Add missing free_irq
Maxime Ripard [Mon, 23 Jun 2014 20:49:40 +0000 (22:49 +0200)]
net: allwinner: emac: Add missing free_irq

If the mdio probe function fails in emac_open, the interrupt we just requested
isn't freed. If emac_open is called again, for example because we try to set up
the interface again, the kernel will oops because the interrupt wasn't properly
released.

Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Cc: <stable@vger.kernel.org> # 3.11+
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agocxgb4: use dev_port to identify ports
Thadeu Lima de Souza Cascardo [Sat, 21 Jun 2014 12:48:08 +0000 (09:48 -0300)]
cxgb4: use dev_port to identify ports

Commit 3f85944fe207d0225ef21a2c0951d4946fc9a95d ("net: Add sysfs file
for port number") introduce dev_port to network devices. cxgb4 adapters
have multiple ports on the same PCI function, and used dev_id to
identify those ports. That use was removed by commit
8c367fcbe6549195d2eb11e62bea233f811aad41 ("cxgb4: Do not set
net_device::dev_id to VI index"), since dev_id should be used only when
devices share the same MAC address.

Using dev_port for cxgb4 allows different ports on the same PCI function
to be identified.

Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoxen-netback: bookkeep number of active queues in our own module
Wei Liu [Mon, 23 Jun 2014 09:50:17 +0000 (10:50 +0100)]
xen-netback: bookkeep number of active queues in our own module

The original code uses netdev->real_num_tx_queues to bookkeep number of
queues and invokes netif_set_real_num_tx_queues to set the number of
queues. However, netif_set_real_num_tx_queues doesn't allow
real_num_tx_queues to be smaller than 1, which means setting the number
to 0 will not work and real_num_tx_queues is untouched.

This is bogus when xenvif_free is invoked before any number of queues is
allocated. That function needs to iterate through all queues to free
resources. Using the wrong number of queues results in NULL pointer
dereference.

So we bookkeep the number of queues in xen-netback to solve this
problem. This fixes a regression introduced by multiqueue patchset in
3.16-rc1.

There's another bug in original code that the real number of RX queues
is never set. In current Xen multiqueue design, the number of TX queues
and RX queues are in fact the same. We need to set the numbers of TX and
RX queues to the same value.

Also remove xenvif_select_queue and leave queue selection to core
driver, as suggested by David Miller.

Reported-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
CC: Ian Campbell <ian.campbell@citrix.com>
CC: Paul Durrant <paul.durrant@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotg3: Change nvram command timeout value to 50ms
Prashant Sreedharan [Sat, 21 Jun 2014 06:28:15 +0000 (23:28 -0700)]
tg3: Change nvram command timeout value to 50ms

Commit 506724c463fcd63477a5e404728a980b71f80bb7 "tg3: Override clock,
link aware and link idle mode during NVRAM dump" changed the timeout
value for nvram command execution from 100ms to 1ms. But the 1ms
timeout value was only sufficient for nvram read operations but not
write operations for most of the devices supported by tg3 driver.
This patch sets the MAX to 50ms. Also it uses usleep_range instead
of udelay.

Signed-off-by: Prashant Sreedharan <prashant@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Suggested-by: David Miller <davem@davemloft.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agohwmon: (gpio-fan) Change name used in hwmon_device_register_with_groups
Julien D'Ascenzio [Wed, 25 Jun 2014 20:00:31 +0000 (22:00 +0200)]
hwmon: (gpio-fan) Change name used in hwmon_device_register_with_groups

Since commit 648cd48c9e566f53c5df30d79857e0937ae13b09
The hwmon name attributes must not include '-' so the name must be
rename from gpio-fan to gpio_fan

Signed-off-by: Julien D'Ascenzio <jdascenzio@yahoo.fr>
Reviewed-by: Jean Delvare <jdelvare@suse.de>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
10 years agoMerge tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm...
Linus Torvalds [Wed, 25 Jun 2014 19:19:01 +0000 (12:19 -0700)]
Merge tag 'fixes-for-linus' of git://git./linux/kernel/git/arm/arm-soc

Pull ARM SoC fixes from Arnd Bergmann:
 "A new set of bug fixes for 3.16, containing patches for seven
  platforms:

  at91:
    - drivers/misc fix for Kconfig PWM symbol
    - correction of several values in DT after conversion to CCF
    - fix at91sam9261/at91sam9261ek mistake in slow crystal vs. slow RC osc

  imx:
    - Use GPIO for card CD/WP on imx51-babbage and eukrea-mbimxsd51,
      because controller base CD/WP is not working in esdhc driver due to
      runtime PM support
    - A couple of random ventana gw5xxx board fixes
    - Add IMX_IPUV3_CORE back to defconfig, which gets lost when moving
      IPUv3 driver out of staging tree
    - Fix enet/fec clock selection on imx6sl
    - Fix display node on imx53-m53evk board
    - A couple of Cubox-i updates from Russell, which were omitted from
      the merge window due to dependency

  integrator:
    - fix an OF-related regression against 3.15

  mvebu:
    - mvebu (v7)
       - Fix broken SoC ID detection
       - Select ARM_CPU_SUSPEND for v7
       - Remove armada38x compatible string (no users yet)
       - Enable Dove SoC in mvebu_v7_defconfig
    - kirkwood
       - Fix phy-connection-type on GuruPlug board

  qcom:
    - enable gsbi driver in defconfig
    - fix section mismatch warning in serial driver

  samsung:
    - use WFI macro in platform_do_lowpower because exynos cpuhotplug
      includes a hardcoded WFI instruction and it causes compile error
      in Thumb-2 mode.
    - fix GIC reg sizes for exynos4 SoCs
    - remove reset timer counter value during boot and resume for mct
      to fix a big jump in printk timestamps
    - fix pm code to check cortex-A9 for another exynos SoCs
    - don't rely on firmware's secondary_cpu_start for mcpm

  sti:
    - Ethernet clocks were wrongly defined for STiH415/416 platforms
    - STiH416 B2020 revision E DTS file name contained uppercase, change to
      lowercase"

* tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc: (33 commits)
  ARM: at91/dt: sam9261: remove slow RC osc
  ARM: at91/dt: define sam9261ek slow crystal frequency
  ARM: at91/dt: sam9261: correctly define mainck
  ARM: at91/dt: sam9n12: correct PLLA ICPLL and OUT values
  ARM: at91/dt: sam9x5: correct PLLA ICPLL and OUT values
  misc: atmel_pwm: fix Kconfig symbols
  ARM: integrator: fix OF-related regression
  ARM: mvebu: Fix the improper use of the compatible string armada38x using a wildcard
  ARM: dts: kirkwood: fix phy-connection-type for Guruplug
  ARM: EXYNOS: Don't rely on firmware's secondary_cpu_start for mcpm
  ARM: dts: imx51-eukrea-mbimxsd51-baseboard: unbreak esdhc.
  ARM: dts: imx51-babbage: Fix esdhc setup
  ARM: dts: mx5: Move the display out of soc {} node
  ARM: dts: mx5: Fix IPU port node placement
  ARM: mvebu: select ARM_CPU_SUSPEND for Marvell EBU v7 platforms
  ARM: mvebu: Fix broken SoC ID detection
  ARM: imx_v6_v7_defconfig: Enable CONFIG_IMX_IPUV3_CORE
  ARM: multi_v7_defconfig: Add QCOM GSBI driver
  ARM: stih41x: Rename stih416-b2020-revE.dts to stih416-b2020e.dts
  tty: serial: msm: Fix section mismatch warning
  ...

10 years agoMerge tag 'at91-fixes' of git://github.com/at91linux/linux-at91 into fixes
Arnd Bergmann [Wed, 25 Jun 2014 18:27:15 +0000 (20:27 +0200)]
Merge tag 'at91-fixes' of git://github.com/at91linux/linux-at91 into fixes

Merge "First AT91 fixes batch for 3.16" from Nicolas Ferre:

- drivers/misc fix for Kconfig PWM symbol
- correction of several values in DT after conversion to CCF
- fix at91sam9261/at91sam9261ek mistake in slow crystal vs. slow RC osc

* tag 'at91-fixes' of git://github.com/at91linux/linux-at91:
  ARM: at91/dt: sam9261: remove slow RC osc
  ARM: at91/dt: define sam9261ek slow crystal frequency
  ARM: at91/dt: sam9261: correctly define mainck
  ARM: at91/dt: sam9n12: correct PLLA ICPLL and OUT values
  ARM: at91/dt: sam9x5: correct PLLA ICPLL and OUT values
  misc: atmel_pwm: fix Kconfig symbols

10 years agoMerge tag 'mvebu-fixes-3.16' of git://git.infradead.org/linux-mvebu into fixes
Arnd Bergmann [Wed, 25 Jun 2014 13:34:00 +0000 (15:34 +0200)]
Merge tag 'mvebu-fixes-3.16' of git://git.infradead.org/linux-mvebu into fixes

Merge "mvebu fixes for v3.16" from Jason Cooper:

 - mvebu
    - Fix broken SoC ID detection
    - Select ARM_CPU_SUSPEND for v7
    - Remove armada38x compatible string (no users yet)

 - kirkwood
    - Fix phy-connection-type on GuruPlug board

* tag 'mvebu-fixes-3.16' of git://git.infradead.org/linux-mvebu:
  ARM: mvebu: Fix the improper use of the compatible string armada38x using a wildcard
  ARM: dts: kirkwood: fix phy-connection-type for Guruplug
  ARM: mvebu: select ARM_CPU_SUSPEND for Marvell EBU v7 platforms
  ARM: mvebu: Fix broken SoC ID detection

10 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph...
Linus Torvalds [Wed, 25 Jun 2014 17:34:17 +0000 (10:34 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/sage/ceph-client

Pull Ceph fix from Sage Weil:
 "This fixes a corner case for cloned RBD images"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client:
  rbd: handle parent_overlap on writes correctly

10 years agoARM: at91/dt: sam9261: remove slow RC osc
Alexandre Belloni [Mon, 23 Jun 2014 06:51:41 +0000 (08:51 +0200)]
ARM: at91/dt: sam9261: remove slow RC osc

The at91sam9261 doesn't actually have a slow RC oscillator, remove it from the
dtsi.

Signed-off-by: Alexandre Belloni <alexandre.belloni@free-electrons.com>
Acked-by: Boris BREZILLON <boris.brezillon@free-electrons.com>
Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com>
10 years agoARM: at91/dt: define sam9261ek slow crystal frequency
Alexandre Belloni [Sat, 14 Jun 2014 00:10:43 +0000 (02:10 +0200)]
ARM: at91/dt: define sam9261ek slow crystal frequency

Define at91sam9261ek's slow crystal frequencies.

Signed-off-by: Alexandre Belloni <alexandre.belloni@free-electrons.com>
Acked-by: Boris BREZILLON <boris.brezillon@free-electrons.com>
Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com>
10 years agoARM: at91/dt: sam9261: correctly define mainck
Alexandre Belloni [Fri, 13 Jun 2014 12:02:29 +0000 (14:02 +0200)]
ARM: at91/dt: sam9261: correctly define mainck

mainck (CKGR_MCFR register) is actually using main_osc (CKGR_MOR register).

Signed-off-by: Alexandre Belloni <alexandre.belloni@free-electrons.com>
Acked-by: Boris BREZILLON <boris.brezillon@free-electrons.com>
Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com>
10 years agoARM: at91/dt: sam9n12: correct PLLA ICPLL and OUT values
Alexandre Belloni [Fri, 13 Jun 2014 11:28:12 +0000 (13:28 +0200)]
ARM: at91/dt: sam9n12: correct PLLA ICPLL and OUT values

ICPLL can only take 0 or 1, it got mixed with OUT which can be in the [0-3]
range.

Signed-off-by: Alexandre Belloni <alexandre.belloni@free-electrons.com>
Acked-by: Boris BREZILLON <boris.brezillon@free-electrons.com>
Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com>
10 years agoARM: at91/dt: sam9x5: correct PLLA ICPLL and OUT values
Alexandre Belloni [Fri, 13 Jun 2014 11:25:34 +0000 (13:25 +0200)]
ARM: at91/dt: sam9x5: correct PLLA ICPLL and OUT values

ICPLL can only take 0 or 1, it got mixed with OUT which can be in the [0-3]
range.

Signed-off-by: Alexandre Belloni <alexandre.belloni@free-electrons.com>
Acked-by: Boris BREZILLON <boris.brezillon@free-electrons.com>
Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com>
10 years agohwmon: (emc1403) Fix missing 'select REGMAP_I2C' in Kconfig
Josef Gajdusek [Wed, 25 Jun 2014 14:21:21 +0000 (16:21 +0200)]
hwmon: (emc1403) Fix missing 'select REGMAP_I2C' in Kconfig

In commit 4cab259f, the emc1403 driver was converted to use regmap but the
necessary Kconfig option was not added.

Signed-off-by: Josef Gajdusek <atx@atx.name>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
10 years agohwmon: (ntc_thermistor) Use the manufacturer name properly
Naveen Krishna Chatradhi [Wed, 25 Jun 2014 06:29:31 +0000 (11:59 +0530)]
hwmon: (ntc_thermistor) Use the manufacturer name properly

Murata Manufacturing Co., Ltd is the vendor for
NTC (Negative Temperature coefficient) based Thermistors.
But, the driver extensively uses "NTC" as the vendor name.

This patch corrects the vendor name also updates the
compatibility strings according to the vendor-prefix.txt

Note: Drivers continue to support the previous compatible strings
but further addition of these compatible strings in device tree
is deprecated.

Signed-off-by: Naveen Krishna Chatradhi <ch.naveen@samsung.com>
Reviewed-by: Javier Martinez Canillas <javier.martinez@collabora.co.uk>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
10 years agodevicetree: bindings: Document murata vendor prefix
Naveen Krishna Chatradhi [Wed, 25 Jun 2014 06:29:30 +0000 (11:59 +0530)]
devicetree: bindings: Document murata vendor prefix

Add Murata Manufacturing Co., Ltd. to the list of device tree
vendor prefixes.

Murata manufactures NTC (Negative Temperature Coefficient) based
Thermistors for small scale applications like Mobiles and PDAs.

Signed-off-by: Naveen Krishna Chatradhi <ch.naveen@samsung.com>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
10 years agohwmon: (w83l786ng) Report correct minimum fan speed
Axel Lin [Wed, 25 Jun 2014 01:02:32 +0000 (09:02 +0800)]
hwmon: (w83l786ng) Report correct minimum fan speed

Current code is buggy, it shows the current fan speed as minimum fan speed.
Fix up show_fan_reg macro to correctly report fan and fan_min speed.

Signed-off-by: Axel Lin <axel.lin@ingics.com>
Acked-by: Kevin Lo <kevlo@kevlo.org>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
10 years agoMerge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc
Linus Torvalds [Wed, 25 Jun 2014 12:44:17 +0000 (05:44 -0700)]
Merge branch 'merge' of git://git./linux/kernel/git/benh/powerpc

Pull powerpc fixes and cleanups from Ben Herrenschmidt:
 "Here are a handful or two of powerpc fixes and simple/trivial
  cleanups.  A bunch of them fix ftrace with the new ABI v2 in Little
  Endian, the rest is a scattering of fairly simple things"

* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc:
  powerpc: Don't skip ePAPR spin-table CPUs
  powerpc/module: Fix TOC symbol CRC
  powerpc/powernv: Remove OPAL v1 takeover
  powerpc/kmemleak: Do not scan the DART table
  selftests/powerpc: Use the test harness for the TM DSCR test
  powerpc/cell: cbe_thermal.c: Cleaning up a variable is of the wrong type
  powerpc/kprobes: Fix jprobes on ABI v2 (LE)
  powerpc/ftrace: Use pr_fmt() to namespace error messages
  powerpc/ftrace: Fix nop of modules on 64bit LE (ABIv2)
  powerpc/ftrace: Fix inverted check of create_branch()
  powerpc/ftrace: Fix typo in mask of opcode
  powerpc: Add ppc_global_function_entry()
  powerpc/macintosh/smu.c: Fix closing brace followed by if
  powerpc: Remove __arch_swab*
  powerpc: Remove ancient DEBUG_SIG code
  powerpc/kerenl: Enable EEH for IO accessors

10 years agoMerge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost
Linus Torvalds [Wed, 25 Jun 2014 12:30:20 +0000 (05:30 -0700)]
Merge tag 'for_linus' of git://git./linux/kernel/git/mst/vhost

Pull vhost cleanups from Michael S Tsirkin:
 "Two cleanup patches removing code duplication that got introduced by
  changes in rc1.  Not fixing crashes, but I'd rather not carry the
  duplicate code until the next merge window"

* tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost:
  vhost-scsi: don't open-code kvfree
  vhost-net: don't open-code kvfree

10 years agoMerge tag 'trace-fixes-v3.16-rc1-v2' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Wed, 25 Jun 2014 12:08:09 +0000 (05:08 -0700)]
Merge tag 'trace-fixes-v3.16-rc1-v2' of git://git./linux/kernel/git/rostedt/linux-trace

Pull tracing cleanups and fixes from Steven Rostedt:
 "This includes three patches from Oleg Nesterov.  The first is a fix to
  a race condition that happens between enabling/disabling syscall
  tracepoints and new process creations (the check to go into the ptrace
  path for a process can be set when it shouldn't, or not set when it
  should).  Not a major bug but one that should be fixed and even
  applied to stable.

  The other two patches are cleanup/fixes that are not that critical,
  but for an -rc1 release would be nice to have.  They both deal with
  syscall tracepoints.

  It also includes a patch to introduce a new macro for the
  TRACE_EVENT() format called __field_struct().  Originally, __field()
  was used to record any variable into a trace event, but with the
  addition of setting the "is signed" attribute, the check causes
  anything but a primitive variable to fail to compile.  That is,
  structs and unions can't be used as they once were.  When the "is
  signed" check was introduce there were only primitive variables being
  recorded.  But that will change soon and it was reported that
  __field() causes build failures.

  To solve the __field() issue, __field_struct() is introduced to allow
  trace_events to be able to record complex types too"

* tag 'trace-fixes-v3.16-rc1-v2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  tracing: Add __field_struct macro for TRACE_EVENT()
  tracing: syscall_regfunc() should not skip kernel threads
  tracing: Change syscall_*regfunc() to check PF_KTHREAD and use for_each_process_thread()
  tracing: Fix syscall_*regfunc() vs copy_process() race

10 years agomisc: atmel_pwm: fix Kconfig symbols
Nicolas Ferre [Wed, 25 Jun 2014 09:33:44 +0000 (11:33 +0200)]
misc: atmel_pwm: fix Kconfig symbols

AT91 symbols AT91SAM9263, AT91SAM9RL, and AT91SAM9G45 do not exist and this
patch changes them to their correct ARCH_* version.
These symbols are chosen instead of the SOC_* ones because this driver is not
converted to DT.
Anyway, the ATMEL_PWM symbol and the associated driver will be removed soon,
during the move to the PWM sub-system.

Reported-by: Paul Bolle <pebolle@tiscali.nl>
Acked-by: Alexandre Belloni <alexandre.belloni@free-electrons.com>
Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com>
10 years agopowerpc: Don't skip ePAPR spin-table CPUs
Scott Wood [Wed, 25 Jun 2014 01:15:51 +0000 (20:15 -0500)]
powerpc: Don't skip ePAPR spin-table CPUs

Commit 59a53afe70fd530040bdc69581f03d880157f15a "powerpc: Don't setup
CPUs with bad status" broke ePAPR SMP booting.  ePAPR says that CPUs
that aren't presently running shall have status of disabled, with
enable-method being used to determine whether the CPU can be enabled.

Fix by checking for spin-table, which is currently the only supported
enable-method.

Signed-off-by: Scott Wood <scottwood@freescale.com>
Cc: Michael Neuling <mikey@neuling.org>
Cc: Emil Medve <Emilian.Medve@Freescale.com>
Cc: stable@vger.kernel.org
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
10 years agopowerpc/module: Fix TOC symbol CRC
Laurent Dufour [Tue, 24 Jun 2014 08:53:59 +0000 (10:53 +0200)]
powerpc/module: Fix TOC symbol CRC

The commit 71ec7c55ed91 introduced the magic symbol ".TOC." for ELFv2 ABI.
This symbol is built manually and has no CRC value computed. A zero value
is put in the CRC section to avoid modpost complaining about a missing CRC.
Unfortunately, this breaks the kernel module loading when the kernel is
relocated (kdump case for instance) because of the relocation applied to
the kcrctab values.

This patch compute a CRC value for the TOC symbol which will match the one
compute by the kernel when it is relocated - aka '0 - relocate_start' done in
maybe_relocated called by check_version (module.c).

Signed-off-by: Laurent Dufour <ldufour@linux.vnet.ibm.com>
Cc: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
10 years agopowerpc/powernv: Remove OPAL v1 takeover
Michael Ellerman [Tue, 24 Jun 2014 07:17:47 +0000 (17:17 +1000)]
powerpc/powernv: Remove OPAL v1 takeover

In commit 27f4488872d9 "Add OPAL takeover from PowerVM" we added support
for "takeover" on OPAL v1 machines.

This was a mode of operation where we would boot under pHyp, and query
for the presence of OPAL. If detected we would then do a special
sequence to take over the machine, and the kernel would end up running
in hypervisor mode.

OPAL v1 was never a supported product, and was never shipped outside
IBM. As far as we know no one is still using it.

Newer versions of OPAL do not use the takeover mechanism. Although the
query for OPAL should be harmless on machines with newer OPAL, we have
seen a machine where it causes a crash in Open Firmware.

The code in early_init_devtree() to copy boot_command_line into cmd_line
was added in commit 817c21ad9a1f "Get kernel command line accross OPAL
takeover", and AFAIK is only used by takeover, so should also be
removed.

Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
10 years agocxgb4: Not need to hold the adap_rcu_lock lock when read adap_rcu_list
Li RongQing [Fri, 20 Jun 2014 09:32:36 +0000 (17:32 +0800)]
cxgb4: Not need to hold the adap_rcu_lock lock when read adap_rcu_list

cxgb4_netdev maybe lead to dead lock, since it uses a spin lock, and be called
in both thread and softirq context, but not disable BH, the lockdep report is
below; In fact, cxgb4_netdev only reads adap_rcu_list with RCU protection, so
not need to hold spin lock again.
=================================
[ INFO: inconsistent lock state ]
3.14.7+ #24 Tainted: G         C O
---------------------------------
inconsistent {SOFTIRQ-ON-W} -> {IN-SOFTIRQ-W} usage.
radvd/3794 [HC0[0]:SC1[1]:HE1:SE0] takes:
 (adap_rcu_lock){+.?...}, at: [<ffffffffa09989ea>] clip_add+0x2c/0x116 [cxgb4]
{SOFTIRQ-ON-W} state was registered at:
  [<ffffffff810fca81>] __lock_acquire+0x34a/0xe48
  [<ffffffff810fd98b>] lock_acquire+0x82/0x9d
  [<ffffffff815d6ff8>] _raw_spin_lock+0x34/0x43
  [<ffffffffa09989ea>] clip_add+0x2c/0x116 [cxgb4]
  [<ffffffffa0998beb>] cxgb4_inet6addr_handler+0x117/0x12c [cxgb4]
  [<ffffffff815da98b>] notifier_call_chain+0x32/0x5c
  [<ffffffff815da9f9>] __atomic_notifier_call_chain+0x44/0x6e
  [<ffffffff815daa32>] atomic_notifier_call_chain+0xf/0x11
  [<ffffffff815b1356>] inet6addr_notifier_call_chain+0x16/0x18
  [<ffffffffa01f72e5>] ipv6_add_addr+0x404/0x46e [ipv6]
  [<ffffffffa01f8df0>] addrconf_add_linklocal+0x5f/0x95 [ipv6]
  [<ffffffffa01fc3e9>] addrconf_notify+0x632/0x841 [ipv6]
  [<ffffffff815da98b>] notifier_call_chain+0x32/0x5c
  [<ffffffff810e09a1>] __raw_notifier_call_chain+0x9/0xb
  [<ffffffff810e09b2>] raw_notifier_call_chain+0xf/0x11
  [<ffffffff8151b3b7>] call_netdevice_notifiers_info+0x4e/0x56
  [<ffffffff8151b3d0>] call_netdevice_notifiers+0x11/0x13
  [<ffffffff8151c0a6>] netdev_state_change+0x1f/0x38
  [<ffffffff8152f004>] linkwatch_do_dev+0x3b/0x49
  [<ffffffff8152f184>] __linkwatch_run_queue+0x10b/0x144
  [<ffffffff8152f1dd>] linkwatch_event+0x20/0x27
  [<ffffffff810d7bc0>] process_one_work+0x1cb/0x2ee
  [<ffffffff810d7e3b>] worker_thread+0x12e/0x1fc
  [<ffffffff810dd391>] kthread+0xc4/0xcc
  [<ffffffff815dc48c>] ret_from_fork+0x7c/0xb0
irq event stamp: 3388
hardirqs last  enabled at (3388): [<ffffffff810c6c85>]
__local_bh_enable_ip+0xaa/0xd9
hardirqs last disabled at (3387): [<ffffffff810c6c2d>]
__local_bh_enable_ip+0x52/0xd9
softirqs last  enabled at (3288): [<ffffffffa01f1d5b>]
rcu_read_unlock_bh+0x0/0x2f [ipv6]
softirqs last disabled at (3289): [<ffffffff815ddafc>]
do_softirq_own_stack+0x1c/0x30

other info that might help us debug this:
 Possible unsafe locking scenario:

       CPU0
       ----
  lock(adap_rcu_lock);
  <Interrupt>
    lock(adap_rcu_lock);

 *** DEADLOCK ***

5 locks held by radvd/3794:
 #0:  (sk_lock-AF_INET6){+.+.+.}, at: [<ffffffffa020b85a>]
rawv6_sendmsg+0x74b/0xa4d [ipv6]
 #1:  (rcu_read_lock){.+.+..}, at: [<ffffffff8151ac6b>]
rcu_lock_acquire+0x0/0x29
 #2:  (rcu_read_lock){.+.+..}, at: [<ffffffffa01f4cca>]
rcu_lock_acquire.constprop.16+0x0/0x30 [ipv6]
 #3:  (rcu_read_lock){.+.+..}, at: [<ffffffff810e09b4>]
rcu_lock_acquire+0x0/0x29
 #4:  (rcu_read_lock){.+.+..}, at: [<ffffffffa0998782>]
rcu_lock_acquire.constprop.40+0x0/0x30 [cxgb4]

stack backtrace:
CPU: 7 PID: 3794 Comm: radvd Tainted: G         C O 3.14.7+ #24
Hardware name: Supermicro X7DBU/X7DBU, BIOS 6.00 12/03/2007
 ffffffff81f15990 ffff88012fdc36a8 ffffffff815d0016 0000000000000006
 ffff8800c80dc2a0 ffff88012fdc3708 ffffffff815cc727 0000000000000001
 0000000000000001 ffff880100000000 ffffffff81015b02 ffff8800c80dcb58
Call Trace:
 <IRQ>  [<ffffffff815d0016>] dump_stack+0x4e/0x71
 [<ffffffff815cc727>] print_usage_bug+0x1ec/0x1fd
 [<ffffffff81015b02>] ? save_stack_trace+0x27/0x44
 [<ffffffff810fbfaa>] ? check_usage_backwards+0xa0/0xa0
 [<ffffffff810fc640>] mark_lock+0x11b/0x212
 [<ffffffff810fca0b>] __lock_acquire+0x2d4/0xe48
 [<ffffffff810fbfaa>] ? check_usage_backwards+0xa0/0xa0
 [<ffffffff810fbff6>] ? check_usage_forwards+0x4c/0xa6
 [<ffffffff810c6c8a>] ? __local_bh_enable_ip+0xaf/0xd9
 [<ffffffff810fd98b>] lock_acquire+0x82/0x9d
 [<ffffffffa09989ea>] ? clip_add+0x2c/0x116 [cxgb4]
 [<ffffffffa0998782>] ? rcu_read_unlock+0x23/0x23 [cxgb4]
 [<ffffffff815d6ff8>] _raw_spin_lock+0x34/0x43
 [<ffffffffa09989ea>] ? clip_add+0x2c/0x116 [cxgb4]
 [<ffffffffa09987b0>] ? rcu_lock_acquire.constprop.40+0x2e/0x30 [cxgb4]
 [<ffffffffa0998782>] ? rcu_read_unlock+0x23/0x23 [cxgb4]
 [<ffffffffa09989ea>] clip_add+0x2c/0x116 [cxgb4]
 [<ffffffffa0998beb>] cxgb4_inet6addr_handler+0x117/0x12c [cxgb4]
 [<ffffffff810fd99d>] ? lock_acquire+0x94/0x9d
 [<ffffffff810e09b4>] ? raw_notifier_call_chain+0x11/0x11
 [<ffffffff815da98b>] notifier_call_chain+0x32/0x5c
 [<ffffffff815da9f9>] __atomic_notifier_call_chain+0x44/0x6e
 [<ffffffff815daa32>] atomic_notifier_call_chain+0xf/0x11
 [<ffffffff815b1356>] inet6addr_notifier_call_chain+0x16/0x18
 [<ffffffffa01f72e5>] ipv6_add_addr+0x404/0x46e [ipv6]
 [<ffffffff810fde6a>] ? trace_hardirqs_on+0xd/0xf
 [<ffffffffa01fb634>] addrconf_prefix_rcv+0x385/0x6ea [ipv6]
 [<ffffffffa0207950>] ndisc_rcv+0x9d3/0xd76 [ipv6]
 [<ffffffffa020d536>] icmpv6_rcv+0x592/0x67b [ipv6]
 [<ffffffff810c6c85>] ? __local_bh_enable_ip+0xaa/0xd9
 [<ffffffff810c6c85>] ? __local_bh_enable_ip+0xaa/0xd9
 [<ffffffff810fd8dc>] ? lock_release+0x14e/0x17b
 [<ffffffffa020df97>] ? rcu_read_unlock+0x21/0x23 [ipv6]
 [<ffffffff8150df52>] ? rcu_read_unlock+0x23/0x23
 [<ffffffffa01f4ede>] ip6_input_finish+0x1e4/0x2fc [ipv6]
 [<ffffffffa01f540b>] ip6_input+0x33/0x38 [ipv6]
 [<ffffffffa01f5557>] ip6_mc_input+0x147/0x160 [ipv6]
 [<ffffffffa01f4ba3>] ip6_rcv_finish+0x7c/0x81 [ipv6]
 [<ffffffffa01f5397>] ipv6_rcv+0x3a1/0x3e2 [ipv6]
 [<ffffffff8151ef96>] __netif_receive_skb_core+0x4ab/0x511
 [<ffffffff810fdc94>] ? mark_held_locks+0x71/0x99
 [<ffffffff8151f0c0>] ? process_backlog+0x69/0x15e
 [<ffffffff8151f045>] __netif_receive_skb+0x49/0x5b
 [<ffffffff8151f0cf>] process_backlog+0x78/0x15e
 [<ffffffff8151f571>] ? net_rx_action+0x1a2/0x1cc
 [<ffffffff8151f47b>] net_rx_action+0xac/0x1cc
 [<ffffffff810c69b7>] ? __do_softirq+0xad/0x218
 [<ffffffff810c69ff>] __do_softirq+0xf5/0x218
 [<ffffffff815ddafc>] do_softirq_own_stack+0x1c/0x30
 <EOI>  [<ffffffff810c6bb6>] do_softirq+0x38/0x5d
 [<ffffffffa01f1d5b>] ? ip6_copy_metadata+0x156/0x156 [ipv6]
 [<ffffffff810c6c78>] __local_bh_enable_ip+0x9d/0xd9
 [<ffffffffa01f1d88>] rcu_read_unlock_bh+0x2d/0x2f [ipv6]
 [<ffffffffa01f28b4>] ip6_finish_output2+0x381/0x3d8 [ipv6]
 [<ffffffffa01f49ef>] ip6_finish_output+0x6e/0x73 [ipv6]
 [<ffffffffa01f4a70>] ip6_output+0x7c/0xa8 [ipv6]
 [<ffffffff815b1bfa>] dst_output+0x18/0x1c
 [<ffffffff815b1c9e>] ip6_local_out+0x1c/0x21
 [<ffffffffa01f2489>] ip6_push_pending_frames+0x37d/0x427 [ipv6]
 [<ffffffff81558af8>] ? skb_orphan+0x39/0x39
 [<ffffffffa020b85a>] ? rawv6_sendmsg+0x74b/0xa4d [ipv6]
 [<ffffffffa020ba51>] rawv6_sendmsg+0x942/0xa4d [ipv6]
 [<ffffffff81584cd2>] inet_sendmsg+0x3d/0x66
 [<ffffffff81508930>] __sock_sendmsg_nosec+0x25/0x27
 [<ffffffff8150b0d7>] sock_sendmsg+0x5a/0x7b
 [<ffffffff810fd8dc>] ? lock_release+0x14e/0x17b
 [<ffffffff8116d756>] ? might_fault+0x9e/0xa5
 [<ffffffff8116d70d>] ? might_fault+0x55/0xa5
 [<ffffffff81508cb1>] ? copy_from_user+0x2a/0x2c
 [<ffffffff8150b70c>] ___sys_sendmsg+0x226/0x2d9
 [<ffffffff810fcd25>] ? __lock_acquire+0x5ee/0xe48
 [<ffffffff810fde01>] ? trace_hardirqs_on_caller+0x145/0x1a1
 [<ffffffff8118efcb>] ? slab_free_hook.isra.71+0x50/0x59
 [<ffffffff8115c81f>] ? release_pages+0xbc/0x181
 [<ffffffff810fd99d>] ? lock_acquire+0x94/0x9d
 [<ffffffff81115e97>] ? read_seqcount_begin.constprop.25+0x73/0x90
 [<ffffffff8150c408>] __sys_sendmsg+0x3d/0x5b
 [<ffffffff8150c433>] SyS_sendmsg+0xd/0x19
 [<ffffffff815dc53d>] system_call_fastpath+0x1a/0x1f

Reported-by: Ben Greear <greearb@candelatech.com>
Cc: Casey Leedom <leedom@chelsio.com>
Cc: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: Li RongQing <roy.qing.li@gmail.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Casey Leedom <leedom@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoNFSv4: test SECINFO RPC_AUTH_GSS pseudoflavors for support
Andy Adamson [Thu, 12 Jun 2014 19:02:32 +0000 (15:02 -0400)]
NFSv4: test SECINFO RPC_AUTH_GSS pseudoflavors for support

Fix nfs4_negotiate_security to create an rpc_clnt used to test each SECINFO
returned pseudoflavor. Check credential creation  (and gss_context creation)
which is important for RPC_AUTH_GSS pseudoflavors which can fail for multiple
reasons including mis-configuration.

Don't call nfs4_negotiate in nfs4_submount as it was just called by
nfs4_proc_lookup_mountpoint (nfs4_proc_lookup_common)

Signed-off-by: Andy Adamson <andros@netapp.com>
[Trond: fix corrupt return value from nfs_find_best_sec()]
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
10 years agoNFS Return -EPERM if no supported or matching SECINFO flavor
Andy Adamson [Mon, 9 Jun 2014 19:33:20 +0000 (15:33 -0400)]
NFS Return -EPERM if no supported or matching SECINFO flavor

Do not return RPC_AUTH_UNIX if SEINFO reply tests fail. This
prevents an infinite loop of NFS4ERR_WRONGSEC for non RPC_AUTH_UNIX mounts.

Without this patch, a mount with no sec= option to a server
that does not include RPC_AUTH_UNIX in the
SECINFO return can be presented with an attemtp to use RPC_AUTH_UNIX
which will result in an NFS4ERR_WRONG_SEC which will prompt the SECINFO
call which will again try RPC_AUTH_UNIX....

Signed-off-by: Andy Adamson <andros@netapp.com>
Tested-By: Steve Dickson <steved@redhat.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
10 years agoNFS check the return of nfs4_negotiate_security in nfs4_submount
Andy Adamson [Mon, 9 Jun 2014 19:33:19 +0000 (15:33 -0400)]
NFS check the return of nfs4_negotiate_security in nfs4_submount

Signed-off-by: Andy Adamson <andros@netapp.com>
Tested-By: Steve Dickson <steved@redhat.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
10 years agoNFS: Don't mark the data cache as invalid if it has been flushed
Trond Myklebust [Fri, 20 Jun 2014 17:11:01 +0000 (13:11 -0400)]
NFS: Don't mark the data cache as invalid if it has been flushed

Now that we have functions such as nfs_write_pageuptodate() that use
the cache_validity flags to check if the data cache is valid or not,
it is a little more important to keep the flags in sync with the
state of the data cache.
In particular, we'd like to ensure that if the data cache is empty, we
don't start marking it as needing revalidation.

Reported-by: Scott Mayhew <smayhew@redhat.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
10 years agoNFS: Clear NFS_INO_REVAL_PAGECACHE when we update the file size
Trond Myklebust [Fri, 20 Jun 2014 17:16:38 +0000 (13:16 -0400)]
NFS: Clear NFS_INO_REVAL_PAGECACHE when we update the file size

In nfs_update_inode(), if the change attribute is seen to change on
the server, then we set NFS_INO_REVAL_PAGECACHE in order to make
sure that we check the file size.
However, if we also update the file size in the same function, we
don't need to check it again. So make sure that we clear the
NFS_INO_REVAL_PAGECACHE that was set earlier.

Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
10 years agonfs: Fix cache_validity check in nfs_write_pageuptodate()
Scott Mayhew [Fri, 20 Jun 2014 12:44:42 +0000 (08:44 -0400)]
nfs: Fix cache_validity check in nfs_write_pageuptodate()

NFS_INO_INVALID_DATA cannot be ignored, even if we have a delegation.

We're still having some problems with data corruption when multiple
clients are appending to a file and those clients are being granted
write delegations on open.

To reproduce:

Client A:
vi /mnt/`hostname -s`
while :; do echo "XXXXXXXXXXXXXXX" >>/mnt/file; sleep $(( $RANDOM % 5 )); done

Client B:
vi /mnt/`hostname -s`
while :; do echo "YYYYYYYYYYYYYYY" >>/mnt/file; sleep $(( $RANDOM % 5 )); done

What's happening is that in nfs_update_inode() we're recognizing that
the file size has changed and we're setting NFS_INO_INVALID_DATA
accordingly, but then we ignore the cache_validity flags in
nfs_write_pageuptodate() because we have a delegation.  As a result,
in nfs_updatepage() we're extending the write to cover the full page
even though we've not read in the data to begin with.

Signed-off-by: Scott Mayhew <smayhew@redhat.com>
Cc: <stable@vger.kernel.org> # v3.11+
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
10 years agoMerge git://git.kvack.org/~bcrl/aio-fixes
Linus Torvalds [Tue, 24 Jun 2014 21:00:13 +0000 (14:00 -0700)]
Merge git://git.kvack.org/~bcrl/aio-fixes

Pull aio fixes from Ben LaHaise:
 "These fix a kernel memory disclosure issue (arbitrary kmap() &
  copy_to_user()) revealed in CVE-2014-0206 by changes that were
  introduced in v3.10"

* git://git.kvack.org/~bcrl/aio-fixes:
  aio: fix kernel memory disclosure in io_getevents() introduced in v3.10
  aio: fix aio request leak when events are reaped by userspace

10 years agoMerge branch 'fixes' of git://ftp.arm.linux.org.uk/~rmk/linux-arm
Linus Torvalds [Tue, 24 Jun 2014 20:59:00 +0000 (13:59 -0700)]
Merge branch 'fixes' of git://ftp.arm.linux.org.uk/~rmk/linux-arm

Pull ARM fixes from Russell King:
 "A number of low impact fixes, the most noticable one is the thumb2
  frame pointer fix.  We also fix a regression caused during this merge
  window with ARM925 CPUs running with caches disabled, and fix a number
  of warnings"

* 'fixes' of git://ftp.arm.linux.org.uk/~rmk/linux-arm:
  ARM: arm925: ensure assembly sets up writethrough mapping
  ARM: perf: fix compiler warning with gcc 4.6.4 (and tidy code)
  ARM: l2c: fix dependencies on PL310 errata symbols
  ARM: 8069/1: Make thread_save_fp macro aware of THUMB2 mode
  ARM: 8068/1: scoop: Remove unused variable

10 years agoaio: fix kernel memory disclosure in io_getevents() introduced in v3.10
Benjamin LaHaise [Tue, 24 Jun 2014 17:32:51 +0000 (13:32 -0400)]
aio: fix kernel memory disclosure in io_getevents() introduced in v3.10

A kernel memory disclosure was introduced in aio_read_events_ring() in v3.10
by commit a31ad380bed817aa25f8830ad23e1a0480fef797.  The changes made to
aio_read_events_ring() failed to correctly limit the index into
ctx->ring_pages[], allowing an attacked to cause the subsequent kmap() of
an arbitrary page with a copy_to_user() to copy the contents into userspace.
This vulnerability has been assigned CVE-2014-0206.  Thanks to Mateusz and
Petr for disclosing this issue.

This patch applies to v3.12+.  A separate backport is needed for 3.10/3.11.

Signed-off-by: Benjamin LaHaise <bcrl@kvack.org>
Cc: Mateusz Guzik <mguzik@redhat.com>
Cc: Petr Matousek <pmatouse@redhat.com>
Cc: Kent Overstreet <kmo@daterainc.com>
Cc: Jeff Moyer <jmoyer@redhat.com>
Cc: stable@vger.kernel.org
10 years agoaio: fix aio request leak when events are reaped by userspace
Benjamin LaHaise [Tue, 24 Jun 2014 17:12:55 +0000 (13:12 -0400)]
aio: fix aio request leak when events are reaped by userspace

The aio cleanups and optimizations by kmo that were merged into the 3.10
tree added a regression for userspace event reaping.  Specifically, the
reference counts are not decremented if the event is reaped in userspace,
leading to the application being unable to submit further aio requests.
This patch applies to 3.12+.  A separate backport is required for 3.10/3.11.
This issue was uncovered as part of CVE-2014-0206.

Signed-off-by: Benjamin LaHaise <bcrl@kvack.org>
Cc: stable@vger.kernel.org
Cc: Kent Overstreet <kmo@daterainc.com>
Cc: Mateusz Guzik <mguzik@redhat.com>
Cc: Petr Matousek <pmatouse@redhat.com>
10 years ago[CIFS] fix mount failure with broken pathnames when smb3 mount with mapchars option
Steve French [Mon, 23 Jun 2014 01:38:49 +0000 (20:38 -0500)]
[CIFS] fix mount failure with broken pathnames when smb3 mount with mapchars option

When we SMB3 mounted with mapchars (to allow reserved characters : \ / > < * ?
via the Unicode Windows to POSIX remap range) empty paths
(eg when we open "" to query the root of the SMB3 directory on mount) were not
null terminated so we sent garbarge as a path name on empty paths which caused
SMB2/SMB2.1/SMB3 mounts to fail when mapchars was specified.  mapchars is
particularly important since Unix Extensions for SMB3 are not supported (yet)

Signed-off-by: Steve French <smfrench@gmail.com>
Cc: <stable@vger.kernel.org>
Reviewed-by: David Disseldorp <ddiss@suse.de>
10 years agoARM: integrator: fix OF-related regression
Linus Walleij [Tue, 24 Jun 2014 12:08:07 +0000 (14:08 +0200)]
ARM: integrator: fix OF-related regression

Commit 07e461cd7e73a84f0e3757932b93cc80976fd749
"of: Ensure unique names without sacrificing determinism"
caused a boot failure regression on the Integrator machines.

The problem is probably caused by fiddling too much with
the device tree population in the OF init function, such
as passing the SoC bus device as parent when populating
the device tree.

This patch fixes the problem by:

- Avoiding to explicitly look up the tree root
- Look up devices needed before device population from
  the match only, passing NULL as root
- Passing NULL as root and parent when calling
  of_platform_populate()

After this the Integrators boot again. Tested on
Integrator/AP and Integrator/CP.

Cc: Grant Likely <grant.likely@linaro.org>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
10 years agoARM: mvebu: Fix the improper use of the compatible string armada38x using a wildcard
Gregory CLEMENT [Mon, 23 Jun 2014 14:16:51 +0000 (16:16 +0200)]
ARM: mvebu: Fix the improper use of the compatible string armada38x using a wildcard

Wildcards in compatible strings should be avoid. "marvell,armada38x"
was recently introduced but was not yet used.

The armada 385 SoC is a superset of the armada 380 SoC (with more CPUs
and more PCIe slots). So this patch replaces the use of
"marvell,armada38x" by the "marvell,armada380" string.

Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Link: https://lkml.kernel.org/r/1403533011-21339-1-git-send-email-gregory.clement@free-electrons.com
Acked-by: Andrew Lunn <andrew@lunn.ch>
Cc: <stable@vger.kernel.org> # v3.15+
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
10 years agopowerpc/kmemleak: Do not scan the DART table
Catalin Marinas [Fri, 13 Jun 2014 08:44:21 +0000 (09:44 +0100)]
powerpc/kmemleak: Do not scan the DART table

The DART table allocation is registered to kmemleak via the
memblock_alloc_base() call. However, the DART table is later unmapped
and dart_tablebase VA no longer accessible. This patch tells kmemleak
not to scan this block and avoid an unhandled paging request.

Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
10 years agoselftests/powerpc: Use the test harness for the TM DSCR test
Michael Ellerman [Fri, 13 Jun 2014 11:16:04 +0000 (21:16 +1000)]
selftests/powerpc: Use the test harness for the TM DSCR test

This gives us standardised success/failure output and also handles
killing the test if it runs forever (2 minutes).

Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
10 years agopowerpc/cell: cbe_thermal.c: Cleaning up a variable is of the wrong type
Rickard Strandqvist [Sat, 14 Jun 2014 16:25:11 +0000 (18:25 +0200)]
powerpc/cell: cbe_thermal.c: Cleaning up a variable is of the wrong type

This variable is of the wrong type, everywhere it is used it
should be an unsigned int rather than a int.

Signed-off-by: Rickard Strandqvist <rickard_strandqvist@spectrumdigital.se>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
10 years agopowerpc/kprobes: Fix jprobes on ABI v2 (LE)
Michael Ellerman [Mon, 23 Jun 2014 03:23:31 +0000 (13:23 +1000)]
powerpc/kprobes: Fix jprobes on ABI v2 (LE)

In commit 721aeaa9 "Build little endian ppc64 kernel with ABIv2", we
missed some updates required in the kprobes code to make jprobes work
when the kernel is built with ABI v2.

Firstly update arch_deref_entry_point() to do the right thing. Now that
we have added ppc_global_function_entry() we can just always use that, it
will do the right thing for 32 & 64 bit and ABI v1 & v2.

Secondly we need to update the code that sets up the register state before
calling the jprobe handler. On ABI v1 we setup r2 to hold the TOC, on ABI
v2 we need to populate r12 with the function entry point address.

Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
10 years agopowerpc/ftrace: Use pr_fmt() to namespace error messages
Michael Ellerman [Tue, 17 Jun 2014 06:15:36 +0000 (16:15 +1000)]
powerpc/ftrace: Use pr_fmt() to namespace error messages

The printks() in our ftrace code have no prefix, so they appear on the
console with very little context, eg:

  Branch out of range

Use pr_fmt() & pr_err() to add a prefix. While we're at it, collapse a
few split lines that don't need to be, and add a missing newline to one
message.

Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
10 years agopowerpc/ftrace: Fix nop of modules on 64bit LE (ABIv2)
Michael Ellerman [Tue, 17 Jun 2014 06:15:35 +0000 (16:15 +1000)]
powerpc/ftrace: Fix nop of modules on 64bit LE (ABIv2)

There is a bug in the handling of the function entry when we are nopping
out a branch from a module in ftrace.

We compare the result of module_trampoline_target() with the value of
ppc_function_entry(), and expect them to be true. But they never will
be.

module_trampoline_target() will always return the global entry point of
the function, whereas ppc_function_entry() will always return the local.

Fix it by using the newly added ppc_global_function_entry().

Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
10 years agopowerpc/ftrace: Fix inverted check of create_branch()
Michael Ellerman [Tue, 17 Jun 2014 06:15:34 +0000 (16:15 +1000)]
powerpc/ftrace: Fix inverted check of create_branch()

In commit 24a1bdc35, "Fix ABIv2 issues with __ftrace_make_call", Anton
changed the logic that creates and patches the branch, and added a
thinko in the check of create_branch(). create_branch() returns the
instruction that was generated, so if we get zero then it succeeded.

The result is we can't ftrace modules:

  Branch out of range
  WARNING: at ../kernel/trace/ftrace.c:1638
  ftrace failed to modify [<d000000004ba001c>] fuse_req_init_context+0x1c/0x90 [fuse]

We should probably fix patch_instruction() to do that check and make the
API saner, but that's a separate patch. For now just invert the test.

Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
10 years agopowerpc/ftrace: Fix typo in mask of opcode
Michael Ellerman [Tue, 17 Jun 2014 06:15:33 +0000 (16:15 +1000)]
powerpc/ftrace: Fix typo in mask of opcode

In commit 24a1bdc35, "Fix ABIv2 issues with __ftrace_make_call", Anton
changed the logic that checks for the expected code sequence when
patching a module.

We missed the typo in the mask, 0xffff00000 should be 0xffff0000, which
has the effect of making the test always true.

That makes it impossible to ftrace against modules, eg:

  Unexpected call sequence: 48000008 e8410018
  WARNING: at ../kernel/trace/ftrace.c:1638
  ftrace failed to modify [<d000000007cf001c>] rng_dev_open+0x1c/0x70 [rng_core]

Reported-by: David Binderman <dcb314@hotmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
10 years agopowerpc: Add ppc_global_function_entry()
Michael Ellerman [Tue, 17 Jun 2014 06:15:32 +0000 (16:15 +1000)]
powerpc: Add ppc_global_function_entry()

ABIv2 has the concept of a global and local entry point to a function.
In most cases we are interested in the local entry point, and so that is
what ppc_function_entry() returns.

However we have a case in the ftrace code where we want the global entry
point, and there may be other places we need it too. Rather than special
casing each, add an accessor.

For ABIv1 and 32-bit there is only a single entry point, so we return
that. That means it's safe for the caller to use this without also
checking the ABI version.

Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
10 years agopowerpc/macintosh/smu.c: Fix closing brace followed by if
Rasmus Villemoes [Fri, 20 Jun 2014 19:44:27 +0000 (21:44 +0200)]
powerpc/macintosh/smu.c: Fix closing brace followed by if

A closing brace followed by "if" is almost certainly a mistake. Maybe
"else if" was meant, but in this case it doesn't really matter.

Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
10 years agopowerpc: Remove __arch_swab*
Benjamin Herrenschmidt [Tue, 24 Jun 2014 02:28:56 +0000 (12:28 +1000)]
powerpc: Remove __arch_swab*

The generic code uses gcc built-ins which work fine so there's no benefit
in implementing our own anymore.

We can't completely remove the ld/st_le* functions as some historical
cruft still uses them, but that's next on the radar

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
10 years agopowerpc: Remove ancient DEBUG_SIG code
Michael Ellerman [Mon, 23 Jun 2014 04:17:47 +0000 (14:17 +1000)]
powerpc: Remove ancient DEBUG_SIG code

We have some compile-time disabled debug code in signal_xx.c. It's from
some ancient time BG, almost certainly part of the original port, given
the very similar code on other arches.

The show_unhandled_signal logic, added in d0c3d534a438 (2.6.24) is
cleaner and prints more useful information, so drop the debug code.

Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
10 years agopowerpc/kerenl: Enable EEH for IO accessors
Gavin Shan [Mon, 23 Jun 2014 00:56:22 +0000 (10:56 +1000)]
powerpc/kerenl: Enable EEH for IO accessors

In arch/powerpc/kernel/iomap.c, lots of IO reading accessors missed
to check EEH error as Ben pointed. The patch fixes it.

For the writing accessors, we change the called functions only for
making them look similar to the reading counterparts.

Suggested-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Gavin Shan <gwshan@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
10 years agoMerge tag 'compress-3.16-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Tue, 24 Jun 2014 00:05:28 +0000 (17:05 -0700)]
Merge tag 'compress-3.16-rc3' of git://git./linux/kernel/git/gregkh/driver-core

Pull compress bugfixes from Greg KH:
 "Here are two bugfixes for some compression functions that resolve some
  errors when uncompressing some pathalogical data.  Both were found by
  Don A  Bailey"

* tag 'compress-3.16-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core:
  lz4: ensure length does not wrap
  lzo: properly check for overruns

10 years agoMerge branch 'akpm' (patches from Andrew Morton)
Linus Torvalds [Mon, 23 Jun 2014 23:48:14 +0000 (16:48 -0700)]
Merge branch 'akpm' (patches from Andrew Morton)

Merge fixes from Andrew Morton:
 "The nmi patch and watchdog patch aren't actually fixes - they're
  features which needed a few last-minutes touchups.

  Otherwise, a rather large batch of fixes - ocfs2 review takes a while
  and I got distracted and missed last week's batch"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>: (31 commits)
  ocfs2/dlm: do not purge lockres that is queued for assert master
  ocfs2: do not return DLM_MIGRATE_RESPONSE_MASTERY_REF to avoid endless,loop during umount
  ocfs2: manually do the iput once ocfs2_add_entry failed in ocfs2_symlink and ocfs2_mknod
  ocfs2: fix a tiny race when running dirop_fileop_racer
  ocfs2/dlm: fix misuse of list_move_tail() in dlm_run_purge_list()
  ocfs2: refcount: take rw_lock in ocfs2_reflink
  ocfs2: revert "ocfs2: fix NULL pointer dereference when dismount and ocfs2rec simultaneously"
  ocfs2: fix deadlock when two nodes are converting same lock from PR to EX and idletimeout closes conn
  ocfs2: should add inode into orphan dir after updating entry in ocfs2_rename()
  mm: fix crashes from mbind() merging vmas
  checkpatch: reduce false positives when checking void function return statements
  ia64: arch/ia64/include/uapi/asm/fcntl.h needs personality.h
  DMA, CMA: fix possible memory leak
  slab: fix oops when reading /proc/slab_allocators
  shmem: fix faulting into a hole while it's punched
  mm: let mm_find_pmd fix buggy race with THP fault
  mm: thp: fix DEBUG_PAGEALLOC oops in copy_page_rep()
  kernel/watchdog.c: print traces for all cpus on lockup detection
  nmi: provide the option to issue an NMI back trace to every cpu but current
  Documentation/accounting/getdelays.c: add missing null-terminate after strncpy call
  ...

10 years agoocfs2/dlm: do not purge lockres that is queued for assert master
Xue jiufei [Mon, 23 Jun 2014 20:22:09 +0000 (13:22 -0700)]
ocfs2/dlm: do not purge lockres that is queued for assert master

When workqueue is delayed, it may occur that a lockres is purged while it
is still queued for master assert.  it may trigger BUG() as follows.

N1                                         N2
dlm_get_lockres()
->dlm_do_master_requery
                                  is the master of lockres,
                                  so queue assert_master work

                                  dlm_thread() start running
                                  and purge the lockres

                                  dlm_assert_master_worker()
                                  send assert master message
                                  to other nodes
receiving the assert_master
message, set master to N2

dlmlock_remote() send create_lock message to N2, but receive DLM_IVLOCKID,
if it is RECOVERY lockres, it triggers the BUG().

Another BUG() is triggered when N3 become the new master and send
assert_master to N1, N1 will trigger the BUG() because owner doesn't
match.  So we should not purge lockres when it is queued for assert
master.

Signed-off-by: joyce.xue <xuejiufei@huawei.com>
Reviewed-by: Mark Fasheh <mfasheh@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoocfs2: do not return DLM_MIGRATE_RESPONSE_MASTERY_REF to avoid endless,loop during...
jiangyiwen [Mon, 23 Jun 2014 20:22:09 +0000 (13:22 -0700)]
ocfs2: do not return DLM_MIGRATE_RESPONSE_MASTERY_REF to avoid endless,loop during umount

The following case may lead to endless loop during umount.

node A         node B               node C       node D
umount volume,
migrate lockres1
to B
                                                 want to lock lockres1,
                                                 send
                                                 MASTER_REQUEST_MSG
                                                 to C
                                    init block mle
               send
               MIGRATE_REQUEST_MSG
               to C
                                    find a block
                                    mle, and then
                                    return
                                    DLM_MIGRATE_RESPONSE_MASTERY_REF
                                    to B
               set C in refmap
                                    umount successfully
               try to umount, endless
               loop occurs when migrate
               lockres1 since C is in
               refmap

So we can fix this endless loop case by only returning
DLM_MIGRATE_RESPONSE_MASTERY_REF if it has a mastery mle when receiving
MIGRATE_REQUEST_MSG.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: jiangyiwen <jiangyiwen@huawei.com>
Reviewed-by: Mark Fasheh <mfasheh@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Xue jiufei <xuejiufei@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoocfs2: manually do the iput once ocfs2_add_entry failed in ocfs2_symlink and ocfs2_mknod
jiangyiwen [Mon, 23 Jun 2014 20:22:09 +0000 (13:22 -0700)]
ocfs2: manually do the iput once ocfs2_add_entry failed in ocfs2_symlink and ocfs2_mknod

When the call to ocfs2_add_entry() failed in ocfs2_symlink() and
ocfs2_mknod(), iput() will not be called during dput(dentry) because no
d_instantiate(), and this will lead to umount hung.

Signed-off-by: jiangyiwen <jiangyiwen@huawei.com>
Cc: Joel Becker <jlbec@evilplan.org>
Reviewed-by: Mark Fasheh <mfasheh@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoocfs2: fix a tiny race when running dirop_fileop_racer
Yiwen Jiang [Mon, 23 Jun 2014 20:22:09 +0000 (13:22 -0700)]
ocfs2: fix a tiny race when running dirop_fileop_racer

When running dirop_fileop_racer we found a dead lock case.

2 nodes, say Node A and Node B, mount the same ocfs2 volume.  Create
/race/16/1 in the filesystem, and let the inode number of dir 16 is less
than the inode number of dir race.

Node A                            Node B
mv /race/16/1 /race/
                                  right after Node A has got the
                                  EX mode of /race/16/, and tries to
                                  get EX mode of /race
                                  ls /race/16/

In this case, Node A has got the EX mode of /race/16/, and wants to get EX
mode of /race/.  Node B has got the PR mode of /race/, and wants to get
the PR mode of /race/16/.  Since EX and PR are mutually exclusive, dead
lock happens.

This patch fixes this case by locking in ancestor order before trying
inode number order.

Signed-off-by: Yiwen Jiang <jiangyiwen@huawei.com>
Signed-off-by: Joseph Qi <joseph.qi@huawei.com>
Cc: Joel Becker <jlbec@evilplan.org>
Reviewed-by: Mark Fasheh <mfasheh@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoocfs2/dlm: fix misuse of list_move_tail() in dlm_run_purge_list()
Xue jiufei [Mon, 23 Jun 2014 20:22:08 +0000 (13:22 -0700)]
ocfs2/dlm: fix misuse of list_move_tail() in dlm_run_purge_list()

When a lockres in purge list but is still in use, it should be moved to
the tail of purge list.  dlm_thread will continue to check next lockres in
purge list.  However, code list_move_tail(&dlm->purge_list,
&lockres->purge) will do *no* movements, so dlm_thread will purge the same
lockres in this loop again and again.  If it is in use for a long time,
other lockres will not be processed.

Signed-off-by: Yiwen Jiang <jiangyiwen@huawei.com>
Signed-off-by: joyce.xue <xuejiufei@huawei.com>
Reviewed-by: Mark Fasheh <mfasheh@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoocfs2: refcount: take rw_lock in ocfs2_reflink
Wengang Wang [Mon, 23 Jun 2014 20:22:08 +0000 (13:22 -0700)]
ocfs2: refcount: take rw_lock in ocfs2_reflink

This patch tries to fix this crash:

 #5 [ffff88003c1cd690] do_invalid_op at ffffffff810166d5
 #6 [ffff88003c1cd730] invalid_op at ffffffff8159b2de
    [exception RIP: ocfs2_direct_IO_get_blocks+359]
    RIP: ffffffffa05dfa27  RSP: ffff88003c1cd7e8  RFLAGS: 00010202
    RAX: 0000000000000000  RBX: ffff88003c1cdaa8  RCX: 0000000000000000
    RDX: 000000000000000c  RSI: ffff880027a95000  RDI: ffff88003c79b540
    RBP: ffff88003c1cd858   R8: 0000000000000000   R9: ffffffff815f6ba0
    R10: 00000000000001c9  R11: 00000000000001c9  R12: ffff88002d271500
    R13: 0000000000000001  R14: 0000000000000000  R15: 0000000000001000
    ORIG_RAX: ffffffffffffffff  CS: 0010  SS: 0018
 #7 [ffff88003c1cd860] do_direct_IO at ffffffff811cd31b
 #8 [ffff88003c1cd950] direct_IO_iovec at ffffffff811cde9c
 #9 [ffff88003c1cd9b0] do_blockdev_direct_IO at ffffffff811ce764
#10 [ffff88003c1cdb80] __blockdev_direct_IO at ffffffff811ce7cc
#11 [ffff88003c1cdbb0] ocfs2_direct_IO at ffffffffa05df756 [ocfs2]
#12 [ffff88003c1cdbe0] generic_file_direct_write_iter at ffffffff8112f935
#13 [ffff88003c1cdc40] ocfs2_file_write_iter at ffffffffa0600ccc [ocfs2]
#14 [ffff88003c1cdd50] do_aio_write at ffffffff8119126c
#15 [ffff88003c1cddc0] aio_rw_vect_retry at ffffffff811d9bb4
#16 [ffff88003c1cddf0] aio_run_iocb at ffffffff811db880
#17 [ffff88003c1cde30] io_submit_one at ffffffff811dc238
#18 [ffff88003c1cde80] do_io_submit at ffffffff811dc437
#19 [ffff88003c1cdf70] sys_io_submit at ffffffff811dc530
#20 [ffff88003c1cdf80] system_call_fastpath at ffffffff8159a159

It crashes at
        BUG_ON(create && (ext_flags & OCFS2_EXT_REFCOUNTED));
in ocfs2_direct_IO_get_blocks.

ocfs2_direct_IO_get_blocks is expecting the OCFS2_EXT_REFCOUNTED be removed in
ocfs2_prepare_inode_for_write() if it was there. But no cluster lock is taken
during the time before (or inside) ocfs2_prepare_inode_for_write() and after
ocfs2_direct_IO_get_blocks().

It can happen in this case:

Node A(which crashes) Node B
------------------------                 ---------------------------
ocfs2_file_aio_write
  ocfs2_prepare_inode_for_write
    ocfs2_inode_lock
    ...
    ocfs2_inode_unlock
  #no refcount found
.... ocfs2_reflink
                                          ocfs2_inode_lock
                                          ...
                                          ocfs2_inode_unlock
                                          #now, refcount flag set on extent

                                        ...
                                        flush change to disk

ocfs2_direct_IO_get_blocks
  ocfs2_get_clusters
    #extent map miss
    #buffer_head miss
    read extents from disk
  found refcount flag on extent
  crash..

Fix:
Take rw_lock in ocfs2_reflink path

Signed-off-by: Wengang Wang <wen.gang.wang@oracle.com>
Reviewed-by: Mark Fasheh <mfasheh@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoocfs2: revert "ocfs2: fix NULL pointer dereference when dismount and ocfs2rec simulta...
Xue jiufei [Mon, 23 Jun 2014 20:22:08 +0000 (13:22 -0700)]
ocfs2: revert "ocfs2: fix NULL pointer dereference when dismount and ocfs2rec simultaneously"

75f82eaa502c ("ocfs2: fix NULL pointer dereference when dismount and
ocfs2rec simultaneously") may cause umount hang while shutting down
truncate log.

The situation is as followes:
ocfs2_dismout_volume
-> ocfs2_recovery_exit
  -> free osb->recovery_map
-> ocfs2_truncate_shutdown
  -> lock global bitmap inode
    -> ocfs2_wait_for_recovery
          -> check whether osb->recovery_map->rm_used is zero

Because osb->recovery_map is already freed, rm_used can be any other
values, so it may yield umount hang.

Signed-off-by: joyce.xue <xuejiufei@huawei.com>
Reviewed-by: Mark Fasheh <mfasheh@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoocfs2: fix deadlock when two nodes are converting same lock from PR to EX and idletim...
Tariq Saeed [Mon, 23 Jun 2014 20:22:08 +0000 (13:22 -0700)]
ocfs2: fix deadlock when two nodes are converting same lock from PR to EX and idletimeout closes conn

Orabug: 18639535

Two node cluster and both nodes hold a lock at PR level and both want to
convert to EX at the same time.  Master node 1 has sent BAST and then
closes the connection due to idletime out.  Node 0 receives BAST, sends
unlock req with cancel flag but gets error -ENOTCONN.  The problem is
this error is ignored in dlm_send_remote_unlock_request() on the
**incorrect** assumption that the master is dead.  See NOTE in comment
why it returns DLM_NORMAL.  Upon getting DLM_NORMAL, node 0 proceeds to
sends convert (without cancel flg) which fails with -ENOTCONN.  waits 5
sec and resends.

This time gets DLM_IVLOCKID from the master since lock not found in
grant, it had been moved to converting queue in response to conv PR->EX
req.  No way out.

Node 1 (master) Node 0
============== ======

  lock mode PR PR

  convert PR -> EX
  mv grant -> convert and que BAST
  ...
                     <-------- convert PR -> EX
  convert que looks like this: ((node 1, PR -> EX) (node 0, PR -> EX))
  ...
                        BAST (want PR -> NL)
                     ------------------>
  ...
  idle timout, conn closed
                                ...
                                In response to BAST,
                                sends unlock with cancel convert flag
                                gets -ENOTCONN. Ignores and
                                sends remote convert request
                                gets -ENOTCONN, waits 5 Sec, retries
  ...
  reconnects
                   <----------------- convert req goes through on next try
  does not find lock on grant que
                   status DLM_IVLOCKID
                   ------------------>
  ...

No way out.  Fix is to keep retrying unlock with cancel flag until it
succeeds or the master dies.

Signed-off-by: Tariq Saeed <tariq.x.saeed@oracle.com>
Reviewed-by: Mark Fasheh <mfasheh@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoocfs2: should add inode into orphan dir after updating entry in ocfs2_rename()
alex chen [Mon, 23 Jun 2014 20:22:07 +0000 (13:22 -0700)]
ocfs2: should add inode into orphan dir after updating entry in ocfs2_rename()

There are two files a and b in dir /mnt/ocfs2.

    node A                           node B

  mv a b
  In ocfs2_rename(), after calling
  ocfs2_orphan_add(), the inode of
  file b will be added into orphan
  dir.

  If ocfs2_update_entry() fails,
  ocfs2_rename return error and mv
  operation fails. But file b still
  exists in the parent dir.

  ocfs2_queue_orphan_scan
   -> ocfs2_queue_recovery_completion
   -> ocfs2_complete_recovery
   -> ocfs2_recover_orphans
  The inode of the file b will be
  put with iput().

  ocfs2_evict_inode
   -> ocfs2_delete_inode
   -> ocfs2_wipe_inode
   -> ocfs2_remove_inode
  OCFS2_VALID_FL in the inode
  i_flags will be cleared.

                                   The file b still can be accessed
                                   on node B.
                                   ls /mnt/ocfs2
                                   When first read the file b with
                                   ocfs2_read_inode_block(). It will
                                   validate the inode using
                                   ocfs2_validate_inode_block().
                                   Because OCFS2_VALID_FL not set in
                                   the inode i_flags, so the file
                                   system will be readonly.

So we should add inode into orphan dir after updating entry in
ocfs2_rename().

Signed-off-by: alex.chen <alex.chen@huawei.com>
Reviewed-by: Mark Fasheh <mfasheh@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agomm: fix crashes from mbind() merging vmas
Hugh Dickins [Mon, 23 Jun 2014 20:22:07 +0000 (13:22 -0700)]
mm: fix crashes from mbind() merging vmas

In v2.6.34 commit 9d8cebd4bcd7 ("mm: fix mbind vma merge problem")
introduced vma merging to mbind(), but it should have also changed the
convention of passing start vma from queue_pages_range() (formerly
check_range()) to new_vma_page(): vma merging may have already freed
that structure, resulting in BUG at mm/mempolicy.c:1738 and probably
worse crashes.

Fixes: 9d8cebd4bcd7 ("mm: fix mbind vma merge problem")
Reported-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Tested-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Christoph Lameter <cl@linux.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: <stable@vger.kernel.org> [2.6.34+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agocheckpatch: reduce false positives when checking void function return statements
Joe Perches [Mon, 23 Jun 2014 20:22:07 +0000 (13:22 -0700)]
checkpatch: reduce false positives when checking void function return statements

The previous patch had a few too many false positives on styles that
should be acceptable.

Signed-off-by: Joe Perches <joe@perches.com>
Tested-by: Anish Bhatt <anish@chelsio.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoia64: arch/ia64/include/uapi/asm/fcntl.h needs personality.h
Andrew Morton [Mon, 23 Jun 2014 20:22:07 +0000 (13:22 -0700)]
ia64: arch/ia64/include/uapi/asm/fcntl.h needs personality.h

fs/notify/fanotify/fanotify_user.c: In function 'SYSC_fanotify_init':
fs/notify/fanotify/fanotify_user.c:726: error: implicit declaration of function 'personality'
fs/notify/fanotify/fanotify_user.c:726: error: 'PER_LINUX32' undeclared (first use in this function)
fs/notify/fanotify/fanotify_user.c:726: error: (Each undeclared identifier is reported only once
fs/notify/fanotify/fanotify_user.c:726: error: for each function it appears in.)

Reported-by: Wu Fengguang <fengguang.wu@intel.com>
Cc: Will Woods <wwoods@redhat.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: <stable@vger.kernel.org> [3.15.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoDMA, CMA: fix possible memory leak
Joonsoo Kim [Mon, 23 Jun 2014 20:22:07 +0000 (13:22 -0700)]
DMA, CMA: fix possible memory leak

We should free memory for bitmap when we find zone mismatch, otherwise
this memory will leak.

Additionally, I copy code comment from PPC KVM's CMA code to inform why
we need to check zone mis-match.

Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Acked-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Reviewed-by: Michal Nazarewicz <mina86@mina86.com>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Acked-by: Minchan Kim <minchan@kernel.org>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: Marek Szyprowski <m.szyprowski@samsung.com>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: Alexander Graf <agraf@suse.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoslab: fix oops when reading /proc/slab_allocators
Joonsoo Kim [Mon, 23 Jun 2014 20:22:06 +0000 (13:22 -0700)]
slab: fix oops when reading /proc/slab_allocators

Commit b1cb0982bdd6 ("change the management method of free objects of
the slab") introduced a bug on slab leak detector
('/proc/slab_allocators').  This detector works like as following
decription.

 1. traverse all objects on all the slabs.
 2. determine whether it is active or not.
 3. if active, print who allocate this object.

but that commit changed the way how to manage free objects, so the logic
determining whether it is active or not is also changed.  In before, we
regard object in cpu caches as inactive one, but, with this commit, we
mistakenly regard object in cpu caches as active one.

This intoduces kernel oops if DEBUG_PAGEALLOC is enabled.  If
DEBUG_PAGEALLOC is enabled, kernel_map_pages() is used to detect who
corrupt free memory in the slab.  It unmaps page table mapping if object
is free and map it if object is active.  When slab leak detector check
object in cpu caches, it mistakenly think this object active so try to
access object memory to retrieve caller of allocation.  At this point,
page table mapping to this object doesn't exist, so oops occurs.

Following is oops message reported from Dave.

It blew up when something tried to read /proc/slab_allocators
(Just cat it, and you should see the oops below)

  Oops: 0000 [#1] PREEMPT SMP DEBUG_PAGEALLOC
  Modules linked in:
  [snip...]
  CPU: 1 PID: 9386 Comm: trinity-c33 Not tainted 3.14.0-rc5+ #131
  task: ffff8801aa46e890 ti: ffff880076924000 task.ti: ffff880076924000
  RIP: 0010:[<ffffffffaa1a8f4a>]  [<ffffffffaa1a8f4a>] handle_slab+0x8a/0x180
  RSP: 0018:ffff880076925de0  EFLAGS: 00010002
  RAX: 0000000000001000 RBX: 0000000000000000 RCX: 000000005ce85ce7
  RDX: ffffea00079be100 RSI: 0000000000001000 RDI: ffff880107458000
  RBP: ffff880076925e18 R08: 0000000000000001 R09: 0000000000000000
  R10: 0000000000000000 R11: 000000000000000f R12: ffff8801e6f84000
  R13: ffffea00079be100 R14: ffff880107458000 R15: ffff88022bb8d2c0
  FS:  00007fb769e45740(0000) GS:ffff88024d040000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: ffff8801e6f84ff8 CR3: 00000000a22db000 CR4: 00000000001407e0
  DR0: 0000000002695000 DR1: 0000000002695000 DR2: 0000000000000000
  DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000070602
  Call Trace:
    leaks_show+0xce/0x240
    seq_read+0x28e/0x490
    proc_reg_read+0x3d/0x80
    vfs_read+0x9b/0x160
    SyS_read+0x58/0xb0
    tracesys+0xd4/0xd9
  Code: f5 00 00 00 0f 1f 44 00 00 48 63 c8 44 3b 0c 8a 0f 84 e3 00 00 00 83 c0 01 44 39 c0 72 eb 41 f6 47 1a 01 0f 84 e9 00 00 00 89 f0 <4d> 8b 4c 04 f8 4d 85 c9 0f 84 88 00 00 00 49 8b 7e 08 4d 8d 46
  RIP   handle_slab+0x8a/0x180

To fix the problem, I introduce an object status buffer on each slab.
With this, we can track object status precisely, so slab leak detector
would not access active object and no kernel oops would occur.  Memory
overhead caused by this fix is only imposed to CONFIG_DEBUG_SLAB_LEAK
which is mainly used for debugging, so memory overhead isn't big
problem.

Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Reported-by: Dave Jones <davej@redhat.com>
Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Reviewed-by: Vladimir Davydov <vdavydov@parallels.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoshmem: fix faulting into a hole while it's punched
Hugh Dickins [Mon, 23 Jun 2014 20:22:06 +0000 (13:22 -0700)]
shmem: fix faulting into a hole while it's punched

Trinity finds that mmap access to a hole while it's punched from shmem
can prevent the madvise(MADV_REMOVE) or fallocate(FALLOC_FL_PUNCH_HOLE)
from completing, until the reader chooses to stop; with the puncher's
hold on i_mutex locking out all other writers until it can complete.

It appears that the tmpfs fault path is too light in comparison with its
hole-punching path, lacking an i_data_sem to obstruct it; but we don't
want to slow down the common case.

Extend shmem_fallocate()'s existing range notification mechanism, so
shmem_fault() can refrain from faulting pages into the hole while it's
punched, waiting instead on i_mutex (when safe to sleep; or repeatedly
faulting when not).

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Hugh Dickins <hughd@google.com>
Reported-by: Sasha Levin <sasha.levin@oracle.com>
Tested-by: Sasha Levin <sasha.levin@oracle.com>
Cc: Dave Jones <davej@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agomm: let mm_find_pmd fix buggy race with THP fault
Hugh Dickins [Mon, 23 Jun 2014 20:22:05 +0000 (13:22 -0700)]
mm: let mm_find_pmd fix buggy race with THP fault

Trinity has reported:

    BUG: unable to handle kernel NULL pointer dereference at 0000000000000018
    IP: __lock_acquire (kernel/locking/lockdep.c:3070 (discriminator 1))
    CPU: 6 PID: 16173 Comm: trinity-c364 Tainted: G        W
                            3.15.0-rc1-next-20140415-sasha-00020-gaa90d09 #398
    lock_acquire (arch/x86/include/asm/current.h:14
                  kernel/locking/lockdep.c:3602)
    _raw_spin_lock (include/linux/spinlock_api_smp.h:143
                    kernel/locking/spinlock.c:151)
    remove_migration_pte (mm/migrate.c:137)
    rmap_walk (mm/rmap.c:1628 mm/rmap.c:1699)
    remove_migration_ptes (mm/migrate.c:224)
    migrate_pages (mm/migrate.c:922 mm/migrate.c:960 mm/migrate.c:1126)
    migrate_misplaced_page (mm/migrate.c:1733)
    __handle_mm_fault (mm/memory.c:3762 mm/memory.c:3812 mm/memory.c:3925)
    handle_mm_fault (mm/memory.c:3948)
    __get_user_pages (mm/memory.c:1851)
    __mlock_vma_pages_range (mm/mlock.c:255)
    __mm_populate (mm/mlock.c:711)
    SyS_mlockall (include/linux/mm.h:1799 mm/mlock.c:817 mm/mlock.c:791)

I believe this comes about because, whereas collapsing and splitting THP
functions take anon_vma lock in write mode (which excludes concurrent
rmap walks), faulting THP functions (write protection and misplaced
NUMA) do not - and mostly they do not need to.

But they do use a pmdp_clear_flush(), set_pmd_at() sequence which, for
an instant (indeed, for a long instant, given the inter-CPU TLB flush in
there), leaves *pmd neither present not trans_huge.

Which can confuse a concurrent rmap walk, as when removing migration
ptes, seen in the dumped trace.  Although that rmap walk has a 4k page
to insert, anon_vmas containing THPs are in no way segregated from
4k-page anon_vmas, so the 4k-intent mm_find_pmd() does need to cope with
that instant when a trans_huge pmd is temporarily absent.

I don't think we need strengthen the locking at the THP end: it's easily
handled with an ACCESS_ONCE() before testing both conditions.

And since mm_find_pmd() had only one caller who wanted a THP rather than
a pmd, let's slightly repurpose it to fail when it hits a THP or
non-present pmd, and open code split_huge_page_address() again.

Signed-off-by: Hugh Dickins <hughd@google.com>
Reported-by: Sasha Levin <sasha.levin@oracle.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Konstantin Khlebnikov <koct9i@gmail.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Bob Liu <bob.liu@oracle.com>
Cc: Christoph Lameter <cl@gentwo.org>
Cc: Dave Jones <davej@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agomm: thp: fix DEBUG_PAGEALLOC oops in copy_page_rep()
Hugh Dickins [Mon, 23 Jun 2014 20:22:05 +0000 (13:22 -0700)]
mm: thp: fix DEBUG_PAGEALLOC oops in copy_page_rep()

Trinity has for over a year been reporting a CONFIG_DEBUG_PAGEALLOC oops
in copy_page_rep() called from copy_user_huge_page() called from
do_huge_pmd_wp_page().

I believe this is a DEBUG_PAGEALLOC false positive, due to the source
page being split, and a tail page freed, while copy is in progress; and
not a problem without DEBUG_PAGEALLOC, since the pmd_same() check will
prevent a miscopy from being made visible.

Fix by adding get_user_huge_page() and put_user_huge_page(): reducing to
the usual get_page() and put_page() on head page in the usual config;
but get and put references to all of the tail pages when
DEBUG_PAGEALLOC.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reported-by: Sasha Levin <sasha.levin@oracle.com>
Tested-by: Sasha Levin <sasha.levin@oracle.com>
Cc: Dave Jones <davej@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agokernel/watchdog.c: print traces for all cpus on lockup detection
Aaron Tomlin [Mon, 23 Jun 2014 20:22:05 +0000 (13:22 -0700)]
kernel/watchdog.c: print traces for all cpus on lockup detection

A 'softlockup' is defined as a bug that causes the kernel to loop in
kernel mode for more than a predefined period to time, without giving
other tasks a chance to run.

Currently, upon detection of this condition by the per-cpu watchdog
task, debug information (including a stack trace) is sent to the system
log.

On some occasions, we have observed that the "victim" rather than the
actual "culprit" (i.e.  the owner/holder of the contended resource) is
reported to the user.  Often this information has proven to be
insufficient to assist debugging efforts.

To avoid loss of useful debug information, for architectures which
support NMI, this patch makes it possible to improve soft lockup
reporting.  This is accomplished by issuing an NMI to each cpu to obtain
a stack trace.

If NMI is not supported we just revert back to the old method.  A sysctl
and boot-time parameter is available to toggle this feature.

[dzickus@redhat.com: add CONFIG_SMP in certain areas]
[akpm@linux-foundation.org: additional CONFIG_SMP=n optimisations]
[mq@suse.cz: fix warning]
Signed-off-by: Aaron Tomlin <atomlin@redhat.com>
Signed-off-by: Don Zickus <dzickus@redhat.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Mateusz Guzik <mguzik@redhat.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Jan Moskyto Matejka <mq@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agonmi: provide the option to issue an NMI back trace to every cpu but current
Aaron Tomlin [Mon, 23 Jun 2014 20:22:05 +0000 (13:22 -0700)]
nmi: provide the option to issue an NMI back trace to every cpu but current

Sometimes it is preferred not to use the trigger_all_cpu_backtrace()
routine when one wants to avoid capturing a back trace for current.  For
instance if one was previously captured recently.

This patch provides a new routine namely
trigger_allbutself_cpu_backtrace() which offers the flexibility to issue
an NMI to every cpu but current and capture a back trace accordingly.

Patch x86 and sparc to support new routine.

[dzickus@redhat.com: add stub in #else clause]
[dzickus@redhat.com: don't print message in single processor case, wrap with get/put_cpu based on Oleg's suggestion]
[sfr@canb.auug.org.au: undo C99ism]
Signed-off-by: Aaron Tomlin <atomlin@redhat.com>
Signed-off-by: Don Zickus <dzickus@redhat.com>
Acked-by: David S. Miller <davem@davemloft.net>
Cc: Mateusz Guzik <mguzik@redhat.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoDocumentation/accounting/getdelays.c: add missing null-terminate after strncpy call
Rickard Strandqvist [Mon, 23 Jun 2014 20:22:04 +0000 (13:22 -0700)]
Documentation/accounting/getdelays.c: add missing null-terminate after strncpy call

Added a guaranteed null-terminate after call to strncpy.

This was partly found using a static code analysis program called
cppcheck.

Signed-off-by: Rickard Strandqvist <rickard_strandqvist@spectrumdigital.se>
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agodrivers/memstick/host/rtsx_pci_ms.c: add cancel_work when remove driver
Micky Ching [Mon, 23 Jun 2014 20:22:04 +0000 (13:22 -0700)]
drivers/memstick/host/rtsx_pci_ms.c: add cancel_work when remove driver

Add cancel_work_sync() in rtsx_pci_ms_drv_remove() to cancel pending
request work when removing the driver.

Signed-off-by: Micky Ching <micky_ching@realsil.com.cn>
Cc: Samuel Ortiz <sameo@linux.intel.com> says:
Cc: Maxim Levitsky <maximlevitsky@gmail.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Alex Dubov <oakad@yahoo.com>
Cc: Roger Tseng <rogerable@realtek.com>
Cc: Wei WANG <wei_wang@realsil.com.cn>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agomm, pcp: allow restoring percpu_pagelist_fraction default
David Rientjes [Mon, 23 Jun 2014 20:22:04 +0000 (13:22 -0700)]
mm, pcp: allow restoring percpu_pagelist_fraction default

Oleg reports a division by zero error on zero-length write() to the
percpu_pagelist_fraction sysctl:

    divide error: 0000 [#1] SMP DEBUG_PAGEALLOC
    CPU: 1 PID: 9142 Comm: badarea_io Not tainted 3.15.0-rc2-vm-nfs+ #19
    Hardware name: Bochs Bochs, BIOS Bochs 01/01/2011
    task: ffff8800d5aeb6e0 ti: ffff8800d87a2000 task.ti: ffff8800d87a2000
    RIP: 0010: percpu_pagelist_fraction_sysctl_handler+0x84/0x120
    RSP: 0018:ffff8800d87a3e78  EFLAGS: 00010246
    RAX: 0000000000000f89 RBX: ffff88011f7fd000 RCX: 0000000000000000
    RDX: 0000000000000000 RSI: 0000000000000001 RDI: 0000000000000010
    RBP: ffff8800d87a3e98 R08: ffffffff81d002c8 R09: ffff8800d87a3f50
    R10: 000000000000000b R11: 0000000000000246 R12: 0000000000000060
    R13: ffffffff81c3c3e0 R14: ffffffff81cfddf8 R15: ffff8801193b0800
    FS:  00007f614f1e9740(0000) GS:ffff88011f440000(0000) knlGS:0000000000000000
    CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
    CR2: 00007f614f1fa000 CR3: 00000000d9291000 CR4: 00000000000006e0
    Call Trace:
      proc_sys_call_handler+0xb3/0xc0
      proc_sys_write+0x14/0x20
      vfs_write+0xba/0x1e0
      SyS_write+0x46/0xb0
      tracesys+0xe1/0xe6

However, if the percpu_pagelist_fraction sysctl is set by the user, it
is also impossible to restore it to the kernel default since the user
cannot write 0 to the sysctl.

This patch allows the user to write 0 to restore the default behavior.
It still requires a fraction equal to or larger than 8, however, as
stated by the documentation for sanity.  If a value in the range [1, 7]
is written, the sysctl will return EINVAL.

This successfully solves the divide by zero issue at the same time.

Signed-off-by: David Rientjes <rientjes@google.com>
Reported-by: Oleg Drokin <green@linuxhacker.ru>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agolib/Kconfig.debug: let FRAME_POINTER exclude SCORE, just like exclude most of other...
Chen Gang [Mon, 23 Jun 2014 20:22:04 +0000 (13:22 -0700)]
lib/Kconfig.debug: let FRAME_POINTER exclude SCORE, just like exclude most of other architectures

The related warning:

  scripts/kconfig/conf --allmodconfig Kconfig
  warning: (FAULT_INJECTION_STACKTRACE_FILTER && LATENCYTOP && KMEMCHECK && LOCKDEP) selects FRAME_POINTER which has unmet direct dependencies (DEBUG_KERNEL && (CRIS || M68K || FRV || UML || AVR32 || SUPERH || BLACKFIN || MN10300 || METAG) || ARCH_WANT_FRAME_POINTERS)

Signed-off-by: Chen Gang <gang.chen.5i5j@gmail.com>
Cc: Chen Liqin <liqin.linux@gmail.com>
Cc: Lennox Wu <lennox.wu@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agokernel/watchdog.c: remove preemption restrictions when restarting lockup detector
Don Zickus [Mon, 23 Jun 2014 20:22:03 +0000 (13:22 -0700)]
kernel/watchdog.c: remove preemption restrictions when restarting lockup detector

Peter Wu noticed the following splat on his machine when updating
/proc/sys/kernel/watchdog_thresh:

  BUG: sleeping function called from invalid context at mm/slub.c:965
  in_atomic(): 1, irqs_disabled(): 0, pid: 1, name: init
  3 locks held by init/1:
   #0:  (sb_writers#3){.+.+.+}, at: [<ffffffff8117b663>] vfs_write+0x143/0x180
   #1:  (watchdog_proc_mutex){+.+.+.}, at: [<ffffffff810e02d3>] proc_dowatchdog+0x33/0x110
   #2:  (cpu_hotplug.lock){.+.+.+}, at: [<ffffffff810589c2>] get_online_cpus+0x32/0x80
  Preemption disabled at:[<ffffffff810e0384>] proc_dowatchdog+0xe4/0x110

  CPU: 0 PID: 1 Comm: init Not tainted 3.16.0-rc1-testing #34
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
  Call Trace:
    dump_stack+0x4e/0x7a
    __might_sleep+0x11d/0x190
    kmem_cache_alloc_trace+0x4e/0x1e0
    perf_event_alloc+0x55/0x440
    perf_event_create_kernel_counter+0x26/0xe0
    watchdog_nmi_enable+0x75/0x140
    update_timers_all_cpus+0x53/0xa0
    proc_dowatchdog+0xe4/0x110
    proc_sys_call_handler+0xb3/0xc0
    proc_sys_write+0x14/0x20
    vfs_write+0xad/0x180
    SyS_write+0x49/0xb0
    system_call_fastpath+0x16/0x1b
  NMI watchdog: disabled (cpu0): hardware events not enabled

What happened is after updating the watchdog_thresh, the lockup detector
is restarted to utilize the new value.  Part of this process involved
disabling preemption.  Once preemption was disabled, perf tried to
allocate a new event (as part of the restart).  This caused the above
BUG_ON as you can't sleep with preemption disabled.

The preemption restriction seemed agressive as we are not doing anything
on that particular cpu, but with all the online cpus (which are
protected by the get_online_cpus lock).  Remove the restriction and the
BUG_ON goes away.

Signed-off-by: Don Zickus <dzickus@redhat.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Reported-by: Peter Wu <peter@lekensteyn.nl>
Tested-by: Peter Wu <peter@lekensteyn.nl>
Acked-by: David Rientjes <rientjes@google.com>
Cc: <stable@vger.kernel.org> [3.13+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoMAINTAINERS: SLAB maintainer update
Christoph Lameter [Mon, 23 Jun 2014 20:22:03 +0000 (13:22 -0700)]
MAINTAINERS: SLAB maintainer update

As discussed in various threads on the side:

Remove one inactive maintainer, add two new ones and update my email
address.  Plus add Andrew.  And fix the glob to include files like
mm/slab_common.c

Signed-off-by: Christoph Lameter <cl@linux.com>
Acked-by: David Rientjes <rientjes@google.com>
Acked-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Acked-by: Pekka Enberg <penberg@kernel.org>
Cc: Matt Mackall <mpm@selenic.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agohugetlb: fix copy_hugetlb_page_range() to handle migration/hwpoisoned entry
Naoya Horiguchi [Mon, 23 Jun 2014 20:22:03 +0000 (13:22 -0700)]
hugetlb: fix copy_hugetlb_page_range() to handle migration/hwpoisoned entry

There's a race between fork() and hugepage migration, as a result we try
to "dereference" a swap entry as a normal pte, causing kernel panic.
The cause of the problem is that copy_hugetlb_page_range() can't handle
"swap entry" family (migration entry and hwpoisoned entry) so let's fix
it.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Acked-by: Hugh Dickins <hughd@google.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: <stable@vger.kernel.org> [2.6.37+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agotmpfs: ZERO_RANGE and COLLAPSE_RANGE not currently supported
Hugh Dickins [Mon, 23 Jun 2014 20:22:03 +0000 (13:22 -0700)]
tmpfs: ZERO_RANGE and COLLAPSE_RANGE not currently supported

I was well aware of FALLOC_FL_ZERO_RANGE and FALLOC_FL_COLLAPSE_RANGE
support being added to fallocate(); but didn't realize until now that I
had been too stupid to future-proof shmem_fallocate() against new
additions.  -EOPNOTSUPP instead of going on to ordinary fallocation.

Signed-off-by: Hugh Dickins <hughd@google.com>
Reviewed-by: Lukas Czerner <lczerner@redhat.com>
Cc: <stable@vger.kernel.org> [3.15]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agomm, hotplug: probe interface is available on several platforms
David Rientjes [Mon, 23 Jun 2014 20:22:03 +0000 (13:22 -0700)]
mm, hotplug: probe interface is available on several platforms

Documentation/memory-hotplug.txt incorrectly states that the memory
driver "probe" interface is only supported on powerpc and is vague about
its application on x86.  Clarify the platforms that make this interface
available if memory hotplug is enabled.

Signed-off-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agokexec: save PG_head_mask in VMCOREINFO
Petr Tesarik [Mon, 23 Jun 2014 20:22:03 +0000 (13:22 -0700)]
kexec: save PG_head_mask in VMCOREINFO

To allow filtering of huge pages, makedumpfile must be able to identify
them in the dump.  This can be done by checking the appropriate page
flag, so communicate its value to makedumpfile through the VMCOREINFO
interface.

There's only one small catch.  Depending on how many page flags are
available on a given architecture, this bit can be called PG_head or
PG_compound.

I sent a similar patch back in 2012, but Eric Biederman did not like
using an #ifdef.  So, this time I'm adding a common symbol
(PG_head_mask) instead.

See https://lkml.org/lkml/2012/11/28/91 for the previous version.

Signed-off-by: Petr Tesarik <ptesarik@suse.cz>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
Cc: Eric Biederman <ebiederm@xmission.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Shaohua Li <shli@kernel.org>
Cc: Alexey Kardashevskiy <aik@ozlabs.ru>
Cc: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoCPU hotplug, smp: flush any pending IPI callbacks before CPU offline
Srivatsa S. Bhat [Mon, 23 Jun 2014 20:22:02 +0000 (13:22 -0700)]
CPU hotplug, smp: flush any pending IPI callbacks before CPU offline

There is a race between the CPU offline code (within stop-machine) and
the smp-call-function code, which can lead to getting IPIs on the
outgoing CPU, *after* it has gone offline.

Specifically, this can happen when using
smp_call_function_single_async() to send the IPI, since this API allows
sending asynchronous IPIs from IRQ disabled contexts.  The exact race
condition is described below.

During CPU offline, in stop-machine, we don't enforce any rule in the
_DISABLE_IRQ stage, regarding the order in which the outgoing CPU and
the other CPUs disable their local interrupts.  Due to this, we can
encounter a situation in which an IPI is sent by one of the other CPUs
to the outgoing CPU (while it is *still* online), but the outgoing CPU
ends up noticing it only *after* it has gone offline.

              CPU 1                                         CPU 2
          (Online CPU)                               (CPU going offline)

       Enter _PREPARE stage                          Enter _PREPARE stage

                                                     Enter _DISABLE_IRQ stage

                                                   =
       Got a device interrupt, and                 | Didn't notice the IPI
       the interrupt handler sent an               | since interrupts were
       IPI to CPU 2 using                          | disabled on this CPU.
       smp_call_function_single_async()            |
                                                   =

       Enter _DISABLE_IRQ stage

       Enter _RUN stage                              Enter _RUN stage

                                  =
       Busy loop with interrupts  |                  Invoke take_cpu_down()
       disabled.                  |                  and take CPU 2 offline
                                  =

       Enter _EXIT stage                             Enter _EXIT stage

       Re-enable interrupts                          Re-enable interrupts

                                                     The pending IPI is noted
                                                     immediately, but alas,
                                                     the CPU is offline at
                                                     this point.

This of course, makes the smp-call-function IPI handler code running on
CPU 2 unhappy and it complains about "receiving an IPI on an offline
CPU".

One real example of the scenario on CPU 1 is the block layer's
complete-request call-path:

__blk_complete_request() [interrupt-handler]
    raise_blk_irq()
        smp_call_function_single_async()

However, if we look closely, the block layer does check that the target
CPU is online before firing the IPI.  So in this case, it is actually
the unfortunate ordering/timing of events in the stop-machine phase that
leads to receiving IPIs after the target CPU has gone offline.

In reality, getting a late IPI on an offline CPU is not too bad by
itself (this can happen even due to hardware latencies in IPI
send-receive).  It is a bug only if the target CPU really went offline
without executing all the callbacks queued on its list.  (Note that a
CPU is free to execute its pending smp-call-function callbacks in a
batch, without waiting for the corresponding IPIs to arrive for each one
of those callbacks).

So, fixing this issue can be broken up into two parts:

1. Ensure that a CPU goes offline only after executing all the
   callbacks queued on it.

2. Modify the warning condition in the smp-call-function IPI handler
   code such that it warns only if an offline CPU got an IPI *and* that
   CPU had gone offline with callbacks still pending in its queue.

Achieving part 1 is straight-forward - just flush (execute) all the
queued callbacks on the outgoing CPU in the CPU_DYING stage[1],
including those callbacks for which the source CPU's IPIs might not have
been received on the outgoing CPU yet.  Once we do this, an IPI that
arrives late on the CPU going offline (either due to the race mentioned
above, or due to hardware latencies) will be completely harmless, since
the outgoing CPU would have executed all the queued callbacks before
going offline.

Overall, this fix (parts 1 and 2 put together) additionally guarantees
that we will see a warning only when the *IPI-sender code* is buggy -
that is, if it queues the callback _after_ the target CPU has gone
offline.

[1].  The CPU_DYING part needs a little more explanation: by the time we
execute the CPU_DYING notifier callbacks, the CPU would have already
been marked offline.  But we want to flush out the pending callbacks at
this stage, ignoring the fact that the CPU is offline.  So restructure
the IPI handler code so that we can by-pass the "is-cpu-offline?" check
in this particular case.  (Of course, the right solution here is to fix
CPU hotplug to mark the CPU offline _after_ invoking the CPU_DYING
notifiers, but this requires a lot of audit to ensure that this change
doesn't break any existing code; hence lets go with the solution
proposed above until that is done).

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Suggested-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Gautham R Shenoy <ego@linux.vnet.ibm.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Mike Galbraith <mgalbraith@suse.de>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rafael J. Wysocki <rjw@rjwysocki.net>
Cc: Rik van Riel <riel@redhat.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Sachin Kamat <sachin.kamat@samsung.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>