GitHub/LineageOS/android_kernel_motorola_exynos9610.git
9 years agomac80211: OCB: remove pointless check for broadcast BSSID
Johannes Berg [Wed, 22 Apr 2015 13:10:45 +0000 (15:10 +0200)]
mac80211: OCB: remove pointless check for broadcast BSSID

The OCB input path already checked that the BSSID is the broadcast
address, so the later check can never fail.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: clean up/rename prepare_for_handlers()
Johannes Berg [Wed, 22 Apr 2015 13:08:39 +0000 (15:08 +0200)]
mac80211: clean up/rename prepare_for_handlers()

The function really shouldn't be called prepare_for_handlers(),
all it does is check if the frame should be dropped. Rename it
to ieee80211_accept_frame() and clean it up a bit.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: remove IEEE80211_RX_RA_MATCH
Johannes Berg [Wed, 22 Apr 2015 12:48:34 +0000 (14:48 +0200)]
mac80211: remove IEEE80211_RX_RA_MATCH

With promisc support gone, only AP and P2P-Device type interfaces
still clear IEEE80211_RX_RA_MATCH. In both cases this isn't really
necessary though, so we can remove that flag and the code.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: remove support for IFF_PROMISC
Johannes Berg [Wed, 22 Apr 2015 12:40:58 +0000 (14:40 +0200)]
mac80211: remove support for IFF_PROMISC

This support is essentially useless as typically networks are encrypted,
frames will be filtered by hardware, and rate scaling will be done with
the intended recipient in mind. For real monitoring of the network, the
monitor mode support should be used instead.

Removing it removes a lot of corner cases.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: make station hash table max_size configurable
Johannes Berg [Thu, 23 Apr 2015 15:26:06 +0000 (17:26 +0200)]
mac80211: make station hash table max_size configurable

Allow debug builds to configure the station hash table maximum
size in order to run with hash collisions in limited scenarios
such as hwsim testing. The default remains 0 which effectively
means no limit.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: allow segmentation offloads
Johannes Berg [Tue, 14 Apr 2015 12:50:41 +0000 (14:50 +0200)]
mac80211: allow segmentation offloads

Implement the necessary software segmentation on the normal
TX path so that fast-xmit can use segmentation offload if
the hardware (or driver) supports it.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: allow drivers to support S/G
Johannes Berg [Mon, 13 Apr 2015 14:58:25 +0000 (16:58 +0200)]
mac80211: allow drivers to support S/G

If drivers want to support S/G (really just gather DMA on TX) then
we can now easily support this on the fast-xmit path since it just
needs to write to the ethernet header (and already has a check for
that being possible.)

However, disallow this on the regular TX path (which has to handle
fragmentation, software crypto, etc.) by calling skb_linearize().

Also allow the related HIGHDMA since that's not interesting to the
code in mac80211 at all anyway.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: allow checksum offload only in fast-xmit
Johannes Berg [Fri, 10 Apr 2015 12:10:10 +0000 (14:10 +0200)]
mac80211: allow checksum offload only in fast-xmit

When we go through the complete TX processing, there are a number
of things like fragmentation and software crypto that require the
checksum to be calculated already.

In favour of maintainability, instead of adding the necessary call
to skb_checksum_help() in all the places that need it, just do it
once before the regular TX processing.

Right now this only affects the TI wlcore and QCA ath10k drivers
since they're the only ones using checksum offload. The previous
commits enabled fast-xmit for them in almost all cases.

For wlcore this even fixes a corner case: when a key fails to be
programmed to hardware software encryption gets used, encrypting
frames with a bad checksum.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agoath10k: enable IEEE80211_HW_SUPPORT_FAST_XMIT
Johannes Berg [Fri, 10 Apr 2015 12:18:40 +0000 (14:18 +0200)]
ath10k: enable IEEE80211_HW_SUPPORT_FAST_XMIT

The driver can clearly enable fast-xmit since it does rate
control in the device and thus must do duration calculation
there as well.

Acked-by: Kalle Valo <kvalo@qca.qualcomm.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agowlcore: enable IEEE80211_HW_SUPPORT_FAST_XMIT
Johannes Berg [Fri, 10 Apr 2015 12:13:14 +0000 (14:13 +0200)]
wlcore: enable IEEE80211_HW_SUPPORT_FAST_XMIT

The driver can clearly enable fast-xmit since it does rate
control in the device and thus must do duration calculation
there as well.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: extend fast-xmit to cover IBSS
Johannes Berg [Tue, 14 Apr 2015 08:28:37 +0000 (10:28 +0200)]
mac80211: extend fast-xmit to cover IBSS

IBSS can be supported very easily since it uses the standard station
authorization state etc. so it just needs to be covered by the header
building switch statement.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: extend fast-xmit for more ciphers
Johannes Berg [Fri, 10 Apr 2015 12:03:17 +0000 (14:03 +0200)]
mac80211: extend fast-xmit for more ciphers

When crypto is offloaded then in some cases it's all handled
by the device, and in others only some space for the IV must
be reserved in the frame. Handle both of these cases in the
fast-xmit path, up to a limit of 18 bytes of space for IVs.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: extend fast-xmit to driver fragmentation
Johannes Berg [Fri, 10 Apr 2015 12:02:08 +0000 (14:02 +0200)]
mac80211: extend fast-xmit to driver fragmentation

If the driver handles fragmentation then it wouldn't
be done in software so we can still use the fast-xmit
path in that case.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211_hwsim: enable IEEE80211_HW_SUPPORT_FAST_XMIT
Johannes Berg [Mon, 30 Mar 2015 13:18:58 +0000 (15:18 +0200)]
mac80211_hwsim: enable IEEE80211_HW_SUPPORT_FAST_XMIT

For hwsim, the duration field in frames is already not valid for
the common case of HT/VHT MCSes, so there's little point in trying
to keep it accurate for the legacy rates. Enable the fast-xmit code
to allow testing that, although given the dependency on hardware
crypto it will only be enabled in open network configurations.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: add TX fastpath
Johannes Berg [Sat, 21 Mar 2015 14:25:43 +0000 (15:25 +0100)]
mac80211: add TX fastpath

In order to speed up mac80211's TX path, add the "fast-xmit" cache
that will cache the data frame 802.11 header and other data to be
able to build the frame more quickly. This cache is rebuilt when
external triggers imply changes, but a lot of the checks done per
packet today are simplified away to the check for the cache.

There's also a more detailed description in the code.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: Fix mac80211.h docbook comments
Jonathan Corbet [Mon, 13 Apr 2015 16:27:35 +0000 (18:27 +0200)]
mac80211: Fix mac80211.h docbook comments

A couple of enums in mac80211.h became structures recently, but the
comments didn't follow suit, leading to errors like:

  Error(.//include/net/mac80211.h:367): Cannot parse enum!
  Documentation/DocBook/Makefile:93: recipe for target 'Documentation/DocBook/80211.xml' failed
  make[1]: *** [Documentation/DocBook/80211.xml] Error 1
  Makefile:1361: recipe for target 'mandocs' failed
  make: *** [mandocs] Error 2

Fix the comments comments accordingly.  Added a couple of other small
comment fixes while I was there to silence other recently-added docbook
warnings.

Reported-by: Jim Davis <jim.epost@gmail.com>
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: lock rate control
Johannes Berg [Thu, 5 Mar 2015 15:10:08 +0000 (16:10 +0100)]
mac80211: lock rate control

Both minstrel (reported by Sven Eckelmann) and the iwlwifi rate
control aren't properly taking concurrency into account. It's
likely that the same is true for other rate control algorithms.

In the case of minstrel this manifests itself in crashes when an
update and other data access are run concurrently, for example
when the stations change bandwidth or similar. In iwlwifi, this
can cause firmware crashes.

Since fixing all rate control algorithms will be very difficult,
just provide locking for invocations. This protects the internal
data structures the algorithms maintain.

I've manipulated hostapd to test this, by having it change its
advertised bandwidth roughly ever 150ms. At the same time, I'm
running a flood ping between the client and the AP, which causes
this race of update vs. get_rate/status to easily happen on the
client. With this change, the system survives this test.

Reported-by: Sven Eckelmann <sven@open-mesh.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: introduce plink lock for plink fields
Bob Copeland [Mon, 13 Apr 2015 21:26:28 +0000 (17:26 -0400)]
mac80211: introduce plink lock for plink fields

The mesh plink code uses sta->lock to serialize access to the
plink state fields between the peer link state machine and the
peer link timer.  Some paths (e.g. those involving
mps_qos_null_tx()) unfortunately hold this spinlock across
frame tx, which is soon to be disallowed.  Add a new spinlock
just for plink access.

Signed-off-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Fri, 17 Apr 2015 20:31:08 +0000 (16:31 -0400)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) Fix verifier memory corruption and other bugs in BPF layer, from
    Alexei Starovoitov.

 2) Add a conservative fix for doing BPF properly in the BPF classifier
    of the packet scheduler on ingress.  Also from Alexei.

 3) The SKB scrubber should not clear out the packet MARK and security
    label, from Herbert Xu.

 4) Fix oops on rmmod in stmmac driver, from Bryan O'Donoghue.

 5) Pause handling is not correct in the stmmac driver because it
    doesn't take into consideration the RX and TX fifo sizes.  From
    Vince Bridgers.

 6) Failure path missing unlock in FOU driver, from Wang Cong.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (44 commits)
  net: dsa: use DEVICE_ATTR_RW to declare temp1_max
  netns: remove BUG_ONs from net_generic()
  IB/ipoib: Fix ndo_get_iflink
  sfc: Fix memcpy() with const destination compiler warning.
  altera tse: Fix network-delays and -retransmissions after high throughput.
  net: remove unused 'dev' argument from netif_needs_gso()
  act_mirred: Fix bogus header when redirecting from VLAN
  inet_diag: fix access to tcp cc information
  tcp: tcp_get_info() should fetch socket fields once
  net: dsa: mv88e6xxx: Add missing initialization in mv88e6xxx_set_port_state()
  skbuff: Do not scrub skb mark within the same name space
  Revert "net: Reset secmark when scrubbing packet"
  bpf: fix two bugs in verification logic when accessing 'ctx' pointer
  bpf: fix bpf helpers to use skb->mac_header relative offsets
  stmmac: Configure Flow Control to work correctly based on rxfifo size
  stmmac: Enable unicast pause frame detect in GMAC Register 6
  stmmac: Read tx-fifo-depth and rx-fifo-depth from the devicetree
  stmmac: Add defines and documentation for enabling flow control
  stmmac: Add properties for transmit and receive fifo sizes
  stmmac: fix oops on rmmod after assigning ip addr
  ...

9 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc
Linus Torvalds [Fri, 17 Apr 2015 20:19:26 +0000 (16:19 -0400)]
Merge git://git./linux/kernel/git/davem/sparc

Pull sparc updates from David Miller:
 "The PowerPC folks have a really nice scalable IOMMU pool allocator
  that we wanted to make use of for sparc.  So here we have a series
  that abstracts out their code into a common layer that anyone can make
  use of.

  Sparc is converted, and the PowerPC folks have reviewed and ACK'd this
  series and plan to convert PowerPC over as well"

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc:
  iommu-common: Fix PARISC compile-time warnings
  sparc: Make LDC use common iommu poll management functions
  sparc: Make sparc64 use scalable lib/iommu-common.c functions
  sparc: Break up monolithic iommu table/lock into finer graularity pools and lock

9 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux-tile
Linus Torvalds [Fri, 17 Apr 2015 20:03:05 +0000 (16:03 -0400)]
Merge git://git./linux/kernel/git/cmetcalf/linux-tile

Pull arch/tile updates from Chris Metcalf:
 "These are mostly nohz_full changes, plus a smattering of minor fixes
  (notably a couple for ftrace)"

* git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux-tile:
  tile: nohz: warn if nohz_full uses hypervisor shared cores
  tile: ftrace: fix function_graph tracer issues
  tile: map data region shadow of kernel as R/W
  tile: support CONTEXT_TRACKING and thus NOHZ_FULL
  tile: support arch_irq_work_raise
  arch: tile: fix null pointer dereference on pt_regs pointer
  tile/elf: reorganize notify_exec()
  tile: use si_int instead of si_ptr for compat_siginfo

9 years agonet: dsa: use DEVICE_ATTR_RW to declare temp1_max
Vivien Didelot [Fri, 17 Apr 2015 19:12:25 +0000 (15:12 -0400)]
net: dsa: use DEVICE_ATTR_RW to declare temp1_max

Since commit da4759c (sysfs: Use only return value from is_visible for
the file mode), it is possible to reduce the permissions of a file.

So declare temp1_max with the DEVICE_ATTR_RW macro and remove the write
permission in dsa_hwmon_attrs_visible if set_temp_limit isn't provided.

Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Reviewed-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus
Linus Torvalds [Fri, 17 Apr 2015 19:50:54 +0000 (15:50 -0400)]
Merge branch 'upstream' of git://git.linux-mips.org/ralf/upstream-linus

Pull MIPS updates from Ralf Baechle:
 "This is the main pull request for MIPS for Linux 4.1.  Most
  noteworthy:

   - Add more Octeon-optimized crypto functions
   - Octeon crypto preemption and locking fixes
   - Little endian support for Octeon
   - Use correct CSR to soft reset Octeons
   - Support LEDs on the Octeon-based DSR-1000N
   - Fix PCI interrupt mapping for the Octeon-based DSR-1000N
   - Mark prom_free_prom_memory() as __init for a number of systems
   - Support for Imagination's Pistachio SOC.  This includes arch and
     CLK bits.  I'd like to merge pinctrl bits later
   - Improve parallelism of csum_partial for certain pipelines
   - Organize DTB files in subdirs like other architectures
   - Implement read_sched_clock for all MIPS platforms other than
     Octeon
   - Massive series of 38 fixes and cleanups for the FPU emulator /
     kernel
   - Further FPU remulator work to support new features.  This sits on a
     separate branch which also has been pulled into the 4.1 KVM branch
   - Clean up and fixes for the SEAD3 eval board; remove unused file
   - Various updates for Netlogic platforms
   - A number of small updates for Loongson 3 platforms
   - Increase the memory limit for ATH79 platforms to 256MB
   - A fair number of fixes and updates for BCM47xx platforms
   - Finish the implementation of XPA support
   - MIPS FDC support.  No, not floppy controller but Fast Debug Channel :)
   - Detect the R16000 used in SGI legacy platforms
   - Fix Kconfig dependencies for the SSB bus support"

* 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus: (265 commits)
  MIPS: Makefile: Fix MIPS ASE detection code
  MIPS: asm: elf: Set O32 default FPU flags
  MIPS: BCM47XX: Fix detecting Microsoft MN-700 & Asus WL500G
  MIPS: Kconfig: Disable SMP/CPS for 64-bit
  MIPS: Hibernate: flush TLB entries earlier
  MIPS: smp-cps: cpu_set FPU mask if FPU present
  MIPS: lose_fpu(): Disable FPU when MSA enabled
  MIPS: ralink: add missing symbol for RALINK_ILL_ACC
  MIPS: ralink: Fix bad config symbol in PCI makefile.
  SSB: fix Kconfig dependencies
  MIPS: Malta: Detect and fix bad memsize values
  Revert "MIPS: Avoid pipeline stalls on some MIPS32R2 cores."
  MIPS: Octeon: Delete override of cpu_has_mips_r2_exec_hazard.
  MIPS: Fix cpu_has_mips_r2_exec_hazard.
  MIPS: kernel: entry.S: Set correct ISA level for mips_ihb
  MIPS: asm: spinlock: Fix addiu instruction for R10000_LLSC_WAR case
  MIPS: r4kcache: Use correct base register for MIPS R6 cache flushes
  MIPS: Kconfig: Fix typo for the r2-to-r6 emulator kernel parameter
  MIPS: unaligned: Fix regular load/store instruction emulation for EVA
  MIPS: unaligned: Surround load/store macros in do {} while statements
  ...

9 years agoMerge tag 'xtensa-20150416' of git://github.com/czankel/xtensa-linux
Linus Torvalds [Fri, 17 Apr 2015 19:32:30 +0000 (15:32 -0400)]
Merge tag 'xtensa-20150416' of git://github.com/czankel/xtensa-linux

Pull Xtensa updates from Chris Zankel:

 - fix linker script transformation for .text / .text.fixup

 - wire bpf and execveat syscalls

 - provide __NR_sync_file_range2 instead of __NR_sync_file_range, as
   that's what xtensa uses.

 - make xtfpgs LCD driver functional and configurable.  This fixes
   hardware lockup on KC705/ML605 boot

 - add audio subsystem bits to xtfpga DTS and provide sample KC705
   config with audio features enabled

 - add CY7C67300 USB controller support to XTFPGA

 - fix locking issues in ISS network driver

 - document PIC and MX interrupt distributor device tree bindings

* tag 'xtensa-20150416' of git://github.com/czankel/xtensa-linux:
  xtensa: xtfpga: add CY7C67300 USB controller support
  irqchip: xtensa-pic: xtensa-mx: document DT bindings
  xtensa: ISS: fix locking in TAP network adapter
  xtensa: Fix fix linker script transformation for .text / .text.fixup
  xtensa: provide __NR_sync_file_range2 instead of __NR_sync_file_range
  xtensa: wire bpf and execveat syscalls
  xtensa: xtfpga: fix hardware lockup caused by LCD driver
  xtensa: xtfpga: provide defconfig with audio subsystem
  xtensa: xtfpga: add audio card to xtfpga DTS

9 years agoiommu-common: Fix PARISC compile-time warnings
Sowmini Varadhan [Fri, 17 Apr 2015 02:28:04 +0000 (22:28 -0400)]
iommu-common: Fix PARISC compile-time warnings

Fixes warnings due to
- no DMA_ERROR_CODE on PARISC,
- sizeof (unsigned long) == 4 bytes on PARISC.

Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonetns: remove BUG_ONs from net_generic()
Denys Vlasenko [Fri, 17 Apr 2015 17:06:30 +0000 (19:06 +0200)]
netns: remove BUG_ONs from net_generic()

This inline has ~500 callsites.

On 04/14/2015 08:37 PM, David Miller wrote:
> That BUG_ON() was added 7 years ago, and I don't remember it ever
> triggering or helping us diagnose something, so just remove it and
> keep the function inlined.

On x86 allyesconfig build:

    text     data      bss       dec     hex filename
82447071 22255384 20627456 125329911 77861f7 vmlinux4
82441375 22255384 20627456 125324215 7784bb7 vmlinux5prime

Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
CC: Eric W. Biederman <ebiederm@xmission.com>
CC: David S. Miller <davem@davemloft.net>
CC: Jan Engelhardt <jengelh@medozas.de>
CC: Jiri Pirko <jpirko@redhat.com>
CC: linux-kernel@vger.kernel.org
CC: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoIB/ipoib: Fix ndo_get_iflink
Erez Shitrit [Thu, 16 Apr 2015 13:34:34 +0000 (16:34 +0300)]
IB/ipoib: Fix ndo_get_iflink

Currently, iflink of the parent interface was always accessed, even
when interface didn't have a parent and hence we crashed there.

Handle the interface types properly: for a child interface, return
the ifindex of the parent, for parent interface, return its ifindex.

For child devices, make sure to set the parent pointer prior to
invoking register_netdevice(), this allows the new ndo to be called
by the stack immediately after the child device is registered.

Fixes: 5aa7add8f14b ('infiniband/ipoib: implement ndo_get_iflink')
Reported-by: Honggang Li <honli@redhat.com>
Signed-off-by: Erez Shitrit <erezsh@mellanox.com>
Signed-off-by: Honggang Li <honli@redhat.com>
Reviewed-By: Jason Gunthorpe <jgunthorpe@obsidianresearch.com>+
Acked-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agosfc: Fix memcpy() with const destination compiler warning.
David S. Miller [Fri, 17 Apr 2015 19:15:40 +0000 (15:15 -0400)]
sfc: Fix memcpy() with const destination compiler warning.

drivers/net/ethernet/sfc/selftest.c: In function â€˜efx_iterate_state’:
drivers/net/ethernet/sfc/selftest.c:388:9: warning: passing argument 1 of â€˜memcpy’ discards â€˜const’ qualifier from pointer target type [-Wdiscarded-array-qualifiers]

This is because the msg[] member of struct efx_loopback_payload
is marked as 'const'.  Remove that.

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoaltera tse: Fix network-delays and -retransmissions after high throughput.
Andreas Oetken [Thu, 16 Apr 2015 21:48:08 +0000 (23:48 +0200)]
altera tse: Fix network-delays and -retransmissions after high throughput.

Fix bug which occurs when more than <limit> packets are available during
napi-poll, leading to "delays" and retransmissions on the network.

Check for (count < limit) before checking the get_rx_status in tse_rx-function.
Function get_rx_status is reading from the response-fifo.
If there is currently a response in the fifo,
reading the last byte of the response pops the value from the fifo.
If the limit is checked as second condition
and the limit is reached the fifo is popped but the packet is not processed.

Signed-off-by: Andreas Oetken <ennoerlangen@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-block
Linus Torvalds [Fri, 17 Apr 2015 19:09:51 +0000 (15:09 -0400)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-block

Pull block core fix from Jens Axboe:
 "A commit in the previous pull request introduce a regression.  So far
  only observed on qemu-sparc64, but it's a general bug.  Please pull
  this single fix to rectify that, thanks"

[ And it turns out that it's been seen outside of that qemu-sparc64
  case, and is easy to trigger with small number of CPUs and blk-mq
  enabled by default - Linus ]

* 'for-linus' of git://git.kernel.dk/linux-block:
  blk-mq: fix iteration of busy bitmap

9 years agoMerge tag 'acpica-4.1-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
Linus Torvalds [Fri, 17 Apr 2015 19:01:29 +0000 (15:01 -0400)]
Merge tag 'acpica-4.1-rc1' of git://git./linux/kernel/git/rafael/linux-pm

Pull ACPICA updates from Rafael Wysocki:
 "This updates the kernel's ACPICA code to upstream revision 20150410
  and adds a fix for a GPE handling regression introduced during the
  3.19 cycle on top of that.

  Included are two stable-candidate bug fixes (one of them fixing a 3.16
  regression), multiple other fixes and a bunch of cleanups.

  Specifics:

   - Fix for a GPE handling regression on Dell Latitude D600 that caused
     GPE signaling to stop working on that machine, which appears to be
     due to a hardware glitch, but it used to work and it can be made
     work again in a relativly straightforward way (Rafael J Wysocki).

   - Fix for a mutex unlock regression related to the handling of ACPI
     tables introduced during the 3.16 development cycle (Octavian
     Purdila).

   - _REV modification to always return 2 which has been done by all
     versions of Windows since NT and the firmware people started to use
     it to distinguish between OSes in their AML and do some silly and
     wrong things on that basis (Bob Moore).

   - Fixes and cleanups related to the acpi_physicall_address data type
     including one stable-candidate fix for an issue occasionally
     occuring on 64-bit machines running 32-bit kernels where using
     offsets provided by the firmware may lead to address overflows (Lv
     Zheng).

   - External() opcode support infrastructure needed for recompiling
     disassembled ACPI tables in some cases including interpreter
     modification to ignore that opcode (Bob Moore).

   - Support for the "Windows 2015" string in _OSI (Bob Moore).

   - GPE debug interface change to return values read from hardware
     registers (Lv Zheng).

   - Removal of the __DATE__ macro usage in tools (Rasmus Villemoes).

   - Assorted minor fixes and cleanups (Lv Zheng, Rickard Strandqvist,
     Bob Moore)"

* tag 'acpica-4.1-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm: (32 commits)
  ACPICA: Store GPE register enable masks upfront
  ACPICA: Update version to 20150410.
  ACPICA: Fix a couple issues with the local printf module.
  ACPICA: Disassembler: Some cleanup of the table dump module.
  ACPICA: iASL: Add support for MSDM ACPI table.
  ACPICA: Update for SLIC ACPI table.
  ACPICA: Add "//" before ascii output of buffers.
  ACPICA: Remove unused internal AML opcode.
  ACPICA: Permanently set _REV to the value '2'.
  ACPICA: Add "Windows 2015" string to _OSI support.
  ACPICA: Add infrastructure for External() opcode.
  ACPICA: iASL: Enhancement for constant folding.
  ACPICA: iASL/Disassembler: Add option to assume table contains valid AML.
  ACPICA: Update AML Debugger global variables.
  ACPICA: Update Resource descriptor dump module.
  ACPICA: Fix a sscanf format string.
  ACPICA: Casting changes around acpi_physical_address/acpi_size.
  ACPICA: Resources: Correct conditional compilation definitions.
  ACPICA: Utilities: Correct conditional compilation definitions.
  ACPICA: Tables: Move an iasl specific table function to iasl source file.
  ...

9 years agotile: nohz: warn if nohz_full uses hypervisor shared cores
Chris Metcalf [Mon, 30 Mar 2015 20:33:00 +0000 (16:33 -0400)]
tile: nohz: warn if nohz_full uses hypervisor shared cores

The "hypervisor shared" cores are ones that the Tilera hypervisor
uses to receive interrupts to manage hypervisor-owned devices.
It's a bad idea to try to use those cores with nohz_full, since
they will get interrupted unpredictably -- and invisibly to Linux
tracing tools, since the interrupts are delivered at a higher
privilege level to the Tilera hypervisor.

Generate a clear warning at boot up that this doesn't end well
for the nohz_full cores in question.

Signed-off-by: Chris Metcalf <cmetcalf@ezchip.com>
9 years agotile: ftrace: fix function_graph tracer issues
Tony Lu [Fri, 27 Mar 2015 18:46:38 +0000 (14:46 -0400)]
tile: ftrace: fix function_graph tracer issues

- Add support for ARCH_SUPPORTS_FTRACE_OPS
- Replace the instruction in ftrace_call with the bundle {move r10, lr;
jal ftrace_stub}, so that the lr contains the right value after returning
from ftrace_stub.  An alternative fix might be to leave the instruction
in ftrace_call alone when it is being updated with ftrace_stub.

Signed-off-by: Tony Lu <zlu@ezchip.com>
Signed-off-by: Chris Metcalf <cmetcalf@ezchip.com>
9 years agotile: map data region shadow of kernel as R/W
Chris Metcalf [Fri, 27 Mar 2015 18:35:31 +0000 (14:35 -0400)]
tile: map data region shadow of kernel as R/W

This is necessary for things like reading /proc/kcore, doing ftrace,
etc.  It happens by default when using huge pages to map the kernel
data, but not when using small pages.

Signed-off-by: Chris Metcalf <cmetcalf@ezchip.com>
9 years agotile: support CONTEXT_TRACKING and thus NOHZ_FULL
Chris Metcalf [Mon, 23 Mar 2015 18:23:58 +0000 (14:23 -0400)]
tile: support CONTEXT_TRACKING and thus NOHZ_FULL

Add the TIF_NOHZ flag appropriately.

Add call to user_exit() on entry to do_work_pending() and on entry
to syscalls via do_syscall_trace_enter(), and also the top of
do_syscall_trace_exit() just because it's done in x86.

Add call to user_enter() at the bottom of do_work_pending() once we
have no more work to do before returning to userspace.

Wrap all the trap code in exception_enter() / exception_exit().

Signed-off-by: Chris Metcalf <cmetcalf@ezchip.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
9 years agonet: remove unused 'dev' argument from netif_needs_gso()
Johannes Berg [Fri, 17 Apr 2015 13:45:04 +0000 (15:45 +0200)]
net: remove unused 'dev' argument from netif_needs_gso()

In commit 04ffcb255f22 ("net: Add ndo_gso_check") Tom originally
added the 'dev' argument to be able to call ndo_gso_check().

Then later, when generalizing this in commit 5f35227ea34b
("net: Generalize ndo_gso_check to ndo_features_check")
Jesse removed the call to ndo_gso_check() in netif_needs_gso()
by calling the new ndo_features_check() in a different place.
This made the 'dev' argument unused.

Remove the unused argument and go back to the code as before.

Cc: Tom Herbert <therbert@google.com>
Cc: Jesse Gross <jesse@nicira.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoact_mirred: Fix bogus header when redirecting from VLAN
Herbert Xu [Fri, 17 Apr 2015 05:32:09 +0000 (13:32 +0800)]
act_mirred: Fix bogus header when redirecting from VLAN

When you redirect a VLAN device to any device, you end up with
crap in af_packet on the xmit path because hard_header_len is
not equal to skb->mac_len.  So the redirected packet contains
four extra bytes at the start which then gets interpreted as
part of the MAC address.

This patch fixes this by only pushing skb->mac_len.  We also
need to fix ifb because it tries to undo the pushing done by
act_mirred.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoinet_diag: fix access to tcp cc information
Eric Dumazet [Fri, 17 Apr 2015 01:10:35 +0000 (18:10 -0700)]
inet_diag: fix access to tcp cc information

Two different problems are fixed here :

1) inet_sk_diag_fill() might be called without socket lock held.
   icsk->icsk_ca_ops can change under us and module be unloaded.
   -> Access to freed memory.
   Fix this using rcu_read_lock() to prevent module unload.

2) Some TCP Congestion Control modules provide information
   but again this is not safe against icsk->icsk_ca_ops
   change and nla_put() errors were ignored. Some sockets
   could not get the additional info if skb was almost full.

Fix this by returning a status from get_info() handlers and
using rcu protection as well.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agotcp: tcp_get_info() should fetch socket fields once
Eric Dumazet [Thu, 16 Apr 2015 23:12:28 +0000 (16:12 -0700)]
tcp: tcp_get_info() should fetch socket fields once

tcp_get_info() can be called without holding socket lock,
so any socket fields can change under us.

Use READ_ONCE() to fetch sk_pacing_rate and sk_max_pacing_rate

Fixes: 977cb0ecf82e ("tcp: add pacing_rate information into tcp_info")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agotile: support arch_irq_work_raise
Chris Metcalf [Mon, 23 Mar 2015 15:21:23 +0000 (11:21 -0400)]
tile: support arch_irq_work_raise

Tile includes a hypervisor hook to deliver messages to arbitrary
tiles, so we can use that to raise an interrupt as soon as
possible on our own core.  Unfortunately the Tilera hypervisor
disabled that support on principle in previous releases, but
it will be available in MDE 4.3.4 and later.

Signed-off-by: Chris Metcalf <cmetcalf@ezchip.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
9 years agoarch: tile: fix null pointer dereference on pt_regs pointer
Colin Ian King [Mon, 16 Mar 2015 20:14:02 +0000 (16:14 -0400)]
arch: tile: fix null pointer dereference on pt_regs pointer

Cppcheck reports the following issue:
[arch/tile/kernel/stack.c:116]: (error) Possible null
  pointer dereference: p

In this case, on reporting on an odd fault, p is set to NULL
and immediately afterwords p is dereferenced iff
!kbt->profile is false.  Rather than doing this check just
return NULL rather than falling through to the potential
null pointer dereference (since the original intentional
outcome would be to return NULL anyhow) for this odd fault
case.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Chris Metcalf <cmetcalf@ezchip.com> [tweaked lightly]
9 years agotile/elf: reorganize notify_exec()
Davidlohr Bueso [Wed, 25 Feb 2015 21:58:35 +0000 (13:58 -0800)]
tile/elf: reorganize notify_exec()

In the future mm->exe_file will be done without mmap_sem
serialization, thus isolate and reorganize the tile elf
code to make the transition easier. Good users will, make
use of the more standard get_mm_exe_file(), requiring only
holding the mmap_sem to read the value, and relying on reference
counting to make sure that the exe file won't dissappear
underneath us.

The visible effects of this patch are:

   o We now take and drop the mmap_sem more often. Instead of
     just in arch_setup_additional_pages(), we also do it in:

     1) get_mm_exe_file()
     2) to get the mm->vm_file and notify the simulator.

    [Note that 1) will disappear once we change the locking
     rules for exe_file.]

   o We avoid getting a free page and doing d_path() while
     holding the mmap_sem. This requires reordering the checks.

Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: Chris Metcalf <cmetcalf@ezchip.com>
9 years agotile: use si_int instead of si_ptr for compat_siginfo
Chris Metcalf [Mon, 16 Mar 2015 19:04:05 +0000 (15:04 -0400)]
tile: use si_int instead of si_ptr for compat_siginfo

To be compatible with the generic get_compat_sigevent(), the
copy_siginfo_to_user32() and thus copy_siginfo_from_user32()
have to use si_int instead of si_ptr.  Using si_ptr means that
for the case of ILP32 compat code running in big-endian mode,
we would end up copying the high 32 bits of the pointer value
into si_int instead of the desired low 32 bits.

Signed-off-by: Chris Metcalf <cmetcalf@ezchip.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
9 years agoblk-mq: fix iteration of busy bitmap
Jens Axboe [Fri, 17 Apr 2015 14:28:50 +0000 (08:28 -0600)]
blk-mq: fix iteration of busy bitmap

Commit 889fa31f00b2 was a bit too eager in reducing the loop count,
so we ended up missing queues in some configurations. Ensure that
our division rounds up, so that's not the case.

Reported-by: Guenter Roeck <linux@roeck-us.net>
Fixes: 889fa31f00b2 ("blk-mq: reduce unnecessary software queue looping")
Signed-off-by: Jens Axboe <axboe@fb.com>
9 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Fri, 17 Apr 2015 13:04:38 +0000 (09:04 -0400)]
Merge branch 'akpm' (patches from Andrew)

Merge third patchbomb from Andrew Morton:

 - various misc things

 - a couple of lib/ optimisations

 - provide DIV_ROUND_CLOSEST_ULL()

 - checkpatch updates

 - rtc tree

 - befs, nilfs2, hfs, hfsplus, fatfs, adfs, affs, bfs

 - ptrace fixes

 - fork() fixes

 - seccomp cleanups

 - more mmap_sem hold time reductions from Davidlohr

* emailed patches from Andrew Morton <akpm@linux-foundation.org>: (138 commits)
  proc: show locks in /proc/pid/fdinfo/X
  docs: add missing and new /proc/PID/status file entries, fix typos
  drivers/rtc/rtc-at91rm9200.c: make IO endian agnostic
  Documentation/spi/spidev_test.c: fix warning
  drivers/rtc/rtc-s5m.c: allow usage on device type different than main MFD type
  .gitignore: ignore *.tar
  MAINTAINERS: add Mediatek SoC mailing list
  tomoyo: reduce mmap_sem hold for mm->exe_file
  powerpc/oprofile: reduce mmap_sem hold for exe_file
  oprofile: reduce mmap_sem hold for mm->exe_file
  mips: ip32: add platform data hooks to use DS1685 driver
  lib/Kconfig: fix up HAVE_ARCH_BITREVERSE help text
  x86: switch to using asm-generic for seccomp.h
  sparc: switch to using asm-generic for seccomp.h
  powerpc: switch to using asm-generic for seccomp.h
  parisc: switch to using asm-generic for seccomp.h
  mips: switch to using asm-generic for seccomp.h
  microblaze: use asm-generic for seccomp.h
  arm: use asm-generic for seccomp.h
  seccomp: allow COMPAT sigreturn overrides
  ...

9 years agoproc: show locks in /proc/pid/fdinfo/X
Andrey Vagin [Thu, 16 Apr 2015 19:49:38 +0000 (12:49 -0700)]
proc: show locks in /proc/pid/fdinfo/X

Let's show locks which are associated with a file descriptor in
its fdinfo file.

Currently we don't have a reliable way to determine who holds a lock.  We
can find some information in /proc/locks, but PID which is reported there
can be wrong.  For example, a process takes a lock, then forks a child and
dies.  In this case /proc/locks contains the parent pid, which can be
reused by another process.

$ cat /proc/locks
...
6: FLOCK  ADVISORY  WRITE 324 00:13:13431 0 EOF
...

$ ps -C rpcbind
  PID TTY          TIME CMD
  332 ?        00:00:00 rpcbind

$ cat /proc/332/fdinfo/4
pos: 0
flags: 0100000
mnt_id: 22
lock: 1: FLOCK  ADVISORY  WRITE 324 00:13:13431 0 EOF

$ ls -l /proc/332/fd/4
lr-x------ 1 root root 64 Mar  5 14:43 /proc/332/fd/4 -> /run/rpcbind.lock

$ ls -l /proc/324/fd/
total 0
lrwx------ 1 root root 64 Feb 27 14:50 0 -> /dev/pts/0
lrwx------ 1 root root 64 Feb 27 14:50 1 -> /dev/pts/0
lrwx------ 1 root root 64 Feb 27 14:49 2 -> /dev/pts/0

You can see that the process with the 324 pid doesn't hold the lock.

This information is required for proper dumping and restoring file
locks.

Signed-off-by: Andrey Vagin <avagin@openvz.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Acked-by: Jeff Layton <jlayton@poochiereds.net>
Acked-by: "J. Bruce Fields" <bfields@fieldses.org>
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: Pavel Emelyanov <xemul@parallels.com>
Cc: Joe Perches <joe@perches.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agodocs: add missing and new /proc/PID/status file entries, fix typos
Nathan Scott [Thu, 16 Apr 2015 19:49:35 +0000 (12:49 -0700)]
docs: add missing and new /proc/PID/status file entries, fix typos

docs: add missing and new /proc/PID/status file entries, fix typos

Signed-off-by: Nathan Scott <nathans@redhat.com>
Signed-off-by: Chen Hanxiao <chenhanxiao@cn.fujitsu.com>
Cc: Serge Hallyn <serge.hallyn@canonical.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agodrivers/rtc/rtc-at91rm9200.c: make IO endian agnostic
Ben Dooks [Thu, 16 Apr 2015 19:49:32 +0000 (12:49 -0700)]
drivers/rtc/rtc-at91rm9200.c: make IO endian agnostic

Change the __raw IO calls to readl/write_relaxed which makes the driver
endian agnostic to run properly on big endian systems.

Signed-off-by: Ben Dooks <ben.dooks@codethink.co.uk>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: Andrew Victor <linux@maxim.org.za>
Acked-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Acked-by: Jean-Christophe PLAGNIOL-VILLARD <plagnioj@jcrosoft.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agoDocumentation/spi/spidev_test.c: fix warning
Andrew Morton [Thu, 16 Apr 2015 19:49:29 +0000 (12:49 -0700)]
Documentation/spi/spidev_test.c: fix warning

Documentation/spi/spidev_test.c:83:5: warning: no previous prototype for 'unespcape' [-Wmissing-prototypes]

fix spelling too.

Acked-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agodrivers/rtc/rtc-s5m.c: allow usage on device type different than main MFD type
Krzysztof Kozlowski [Thu, 16 Apr 2015 19:49:27 +0000 (12:49 -0700)]
drivers/rtc/rtc-s5m.c: allow usage on device type different than main MFD type

The RTC driver supports two flavors of S5M devices: S5M8767-like and
S2MPS14-like.

On S2MPS13 and S2MPS14 devices the RTC module is the same so we want to
re-use the existing support of S2MPS14.  However device type was passed
from parent MFD driver in platform data structure.  This way for the
S2MPS13 device the main MFD driver passed device type of 'S2MPS13X'.

Instead decouple detecting of device type between main MFD and RTC driver.
 This allows adding support for other S2MPS14 variations (like S2MPS11 and
S2MPS13) easily by adding to mfd/sec-core.c:

static const struct mfd_cell s2mps13_devs[] = {
{ .name = "s2mps14-rtc", }
};

Signed-off-by: Krzysztof Kozlowski <k.kozlowski@samsung.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: Kyungmin Park <kyungmin.park@samsung.com>
Cc: Marek Szyprowski <m.szyprowski@samsung.com>
Cc: Chanwoo Choi <cw00.choi@samsung.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years ago.gitignore: ignore *.tar
Andrey Skvortsov [Thu, 16 Apr 2015 19:49:24 +0000 (12:49 -0700)]
.gitignore: ignore *.tar

Running make tar-pkg results in following:

 # Untracked files:
 #   (use "git add <file>..." to include in what will be committed)
 #
 #       linux-4.0.0-rc3-next-20150313-150225--x86.tar

This patch makes git ignore *.tar files.
Running 'git ls-files -i --exclude-standard' does not show any
tar files excluded from tracking after the change.

Signed-off-by: Andrey Skvortsov <andrej.skvortzov@gmail.com>
Cc: Michal Marek <mmarek@suse.cz>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Boaz Harrosh <boaz@plexistor.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agoMAINTAINERS: add Mediatek SoC mailing list
Matthias Brugger [Thu, 16 Apr 2015 19:49:21 +0000 (12:49 -0700)]
MAINTAINERS: add Mediatek SoC mailing list

Add the new list that Mediatek specific patches should also be
directed to.

Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com>
Cc: Olof Johansson <olof@lixom.net>
Cc: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agotomoyo: reduce mmap_sem hold for mm->exe_file
Davidlohr Bueso [Thu, 16 Apr 2015 19:49:18 +0000 (12:49 -0700)]
tomoyo: reduce mmap_sem hold for mm->exe_file

The mm->exe_file is currently serialized with mmap_sem (shared) in order
to both safely (1) read the file and (2) compute the realpath by calling
tomoyo_realpath_from_path, making it an absolute overkill.  Good users
will, on the other hand, make use of the more standard get_mm_exe_file(),
requiring only holding the mmap_sem to read the value, and relying on
reference

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Acked-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: James Morris <jmorris@namei.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agopowerpc/oprofile: reduce mmap_sem hold for exe_file
Davidlohr Bueso [Thu, 16 Apr 2015 19:49:15 +0000 (12:49 -0700)]
powerpc/oprofile: reduce mmap_sem hold for exe_file

In the future mm->exe_file will be done without mmap_sem serialization,
thus isolate and reorganize the related code to make the transition
easier.  Good users will, make use of the more standard get_mm_exe_file(),
requiring only holding the mmap_sem to read the value, and relying on
reference counting to make sure that the exe file won't dissappear
underneath us while getting the dcookie.

Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Robert Richter <rric@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agooprofile: reduce mmap_sem hold for mm->exe_file
Davidlohr Bueso [Thu, 16 Apr 2015 19:49:12 +0000 (12:49 -0700)]
oprofile: reduce mmap_sem hold for mm->exe_file

sync_buffer() needs the mmap_sem for two distinct operations, both only
occurring upon user context switch handling:

 1) Dealing with the exe_file.

 2) Adding the dcookie data as we need to lookup the vma that
   backs it. This is done via add_sample() and add_data().

This patch isolates 1), for it will no longer need the mmap_sem for
serialization.  However, for now, make of the more standard
get_mm_exe_file(), requiring only holding the mmap_sem to read the value,
and relying on reference counting to make sure that the exe file won't
dissappear underneath us while doing the get dcookie.

As a consequence, for 2) we move the mmap_sem locking into where we really
need it, in lookup_dcookie().  The benefits are twofold: reduce mmap_sem
hold times, and cleaner code.

[akpm@linux-foundation.org: export get_mm_exe_file for arch/x86/oprofile/oprofile.ko]
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Cc: Robert Richter <rric@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agomips: ip32: add platform data hooks to use DS1685 driver
Joshua Kinard [Thu, 16 Apr 2015 19:49:09 +0000 (12:49 -0700)]
mips: ip32: add platform data hooks to use DS1685 driver

This modifies the IP32 (SGI O2) platform and reset code to utilize the new
rtc-ds1685 driver.  The old mc146818rtc.h header is removed and ip32_defconfig
is updated as well.

Signed-off-by: Joshua Kinard <kumba@gentoo.org>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agolib/Kconfig: fix up HAVE_ARCH_BITREVERSE help text
Andrew Morton [Thu, 16 Apr 2015 19:49:07 +0000 (12:49 -0700)]
lib/Kconfig: fix up HAVE_ARCH_BITREVERSE help text

Cc: Yalin Wang <yalin.wang@sonymobile.com>
Cc: Russell King <linux@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agox86: switch to using asm-generic for seccomp.h
Kees Cook [Thu, 16 Apr 2015 19:49:04 +0000 (12:49 -0700)]
x86: switch to using asm-generic for seccomp.h

Switch to using the newly created asm-generic/seccomp.h for the seccomp
strict mode syscall definitions. The obsolete sigreturn syscall override
is retained in 32-bit mode, and the ia32 syscall overrides are used in
the compat case. Remaining definitions were identical.

Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agosparc: switch to using asm-generic for seccomp.h
Kees Cook [Thu, 16 Apr 2015 19:49:01 +0000 (12:49 -0700)]
sparc: switch to using asm-generic for seccomp.h

Switch to using the newly created asm-generic/seccomp.h for the seccomp
strict mode syscall definitions. The obsolete sigreturn in COMPAT mode
is retained as an override. Remaining definitions are identical. Also
corrected missing #define for header reinclusion protection.

Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agopowerpc: switch to using asm-generic for seccomp.h
Kees Cook [Thu, 16 Apr 2015 19:48:58 +0000 (12:48 -0700)]
powerpc: switch to using asm-generic for seccomp.h

Switch to using the newly created asm-generic/seccomp.h for the seccomp
strict mode syscall definitions.  The obsolete sigreturn in COMPAT mode is
retained as an override.  Remaining definitions are identical, though they
incorrectly appeared in uapi, which has been corrected.

Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agoparisc: switch to using asm-generic for seccomp.h
Kees Cook [Thu, 16 Apr 2015 19:48:55 +0000 (12:48 -0700)]
parisc: switch to using asm-generic for seccomp.h

Switch to using the newly created asm-generic/seccomp.h for the seccomp
strict mode syscall definitions. Definitions were identical.

Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: "James E.J. Bottomley" <jejb@parisc-linux.org>
Cc: Helge Deller <deller@gmx.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agomips: switch to using asm-generic for seccomp.h
Kees Cook [Thu, 16 Apr 2015 19:48:53 +0000 (12:48 -0700)]
mips: switch to using asm-generic for seccomp.h

Switch to using the newly created asm-generic/seccomp.h for the seccomp
strict mode syscall definitions.  COMPAT definitions retain their
overrides and the remaining definitions were identical.

Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agomicroblaze: use asm-generic for seccomp.h
Kees Cook [Thu, 16 Apr 2015 19:48:50 +0000 (12:48 -0700)]
microblaze: use asm-generic for seccomp.h

Switch to using the newly created asm-generic/seccomp.h for the seccomp
strict mode syscall definitions. Since microblaze is 32-bit, the COMPAT
seccomp defines are unused and can be dropped. The obsolete sigreturn
for seccomp strict mode is retained as an override. Remaining definitions
are identical.

Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: Michal Simek <monstr@monstr.eu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agoarm: use asm-generic for seccomp.h
Kees Cook [Thu, 16 Apr 2015 19:48:47 +0000 (12:48 -0700)]
arm: use asm-generic for seccomp.h

Switch to using the newly created asm-generic/seccomp.h for the seccomp
strict mode syscall definitions. Definitions were identical.

Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: Russell King <linux@arm.linux.org.uk>
Cc: Laura Abbott <lauraa@codeaurora.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agoseccomp: allow COMPAT sigreturn overrides
Kees Cook [Thu, 16 Apr 2015 19:48:44 +0000 (12:48 -0700)]
seccomp: allow COMPAT sigreturn overrides

Most architectures don't need to do much special for the strict-mode
seccomp syscall entries.  Remove the redundant headers and reduce the
others.

This patch (of 8):

Some architectures may need to override the compat sigreturn definition,
as is already possible in the non-compat case.

Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Russell King <linux@arm.linux.org.uk>
Cc: Michal Simek <monstr@monstr.eu>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: "James E.J. Bottomley" <jejb@parisc-linux.org>
Cc: Helge Deller <deller@gmx.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Daniel Borkmann <dborkman@redhat.com>
Cc: Laura Abbott <lauraa@codeaurora.org>
Cc: James Hogan <james.hogan@imgtec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agoarc: do not export symbols in troubleshoot.c
Davidlohr Bueso [Thu, 16 Apr 2015 19:48:40 +0000 (12:48 -0700)]
arc: do not export symbols in troubleshoot.c

print_task_path_n_nm() is local to this file, its only user being
show_regs().  Mark the function static and avoid the EXPORT_SYMBOL.

Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Acked-by: Vineet Gupta <vgupta@synoipsys.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agoinclude/linux/kconfig.h: ese macros which are already defined
Michal Simek [Thu, 16 Apr 2015 19:48:38 +0000 (12:48 -0700)]
include/linux/kconfig.h: ese macros which are already defined

It is better to use macros which are already available because then there
is just one location which needs to be change.

[akpm@linux-foundation.org: move IS_ENABLED definition to after the IS_BUILTIN and IS_MODULE definitions]
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agomemstick: mspro_block: add missing curly braces
Dan Carpenter [Thu, 16 Apr 2015 19:48:35 +0000 (12:48 -0700)]
memstick: mspro_block: add missing curly braces

Using the indenting we can see the curly braces were obviously intended.
This is a static checker fix, but my guess is that we don't read enough
bytes, because we don't calculate "t_len" correctly.

Fixes: f1d82698029b ('memstick: use fully asynchronous request processing')
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Alex Dubov <oakad@yahoo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agobfs: correct return values
Sanidhya Kashyap [Thu, 16 Apr 2015 19:48:32 +0000 (12:48 -0700)]
bfs: correct return values

In case of failed memory allocation, the return should be ENOMEM instead
of ENOSPC.

Return -EIO when sb_bread() fails.

Signed-off-by: Sanidhya Kashyap <sanidhya.gatech@gmail.com>
Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agobfs: bfad_worker cleanup
Jiri Slaby [Thu, 16 Apr 2015 19:48:29 +0000 (12:48 -0700)]
bfs: bfad_worker cleanup

This kthread is not loop at all due to break at the end of the loop.  Make
that function linear, with no while loop.

And remove an unnecessary cast.

Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Cc: Anil Gurumurthy <anil.gurumurthy@qlogic.com>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agoaffs: kstrdup() memory handling
Sanidhya Kashyap [Thu, 16 Apr 2015 19:48:26 +0000 (12:48 -0700)]
affs: kstrdup() memory handling

There is a possibility of kstrdup() failure upon memory pressure.
Therefore, returning ENOMEM even for new_opts.

[akpm@linux-foundation.org: cleanup]
Signed-off-by: Sanidhya Kashyap <sanidhya.gatech@gmail.com>
Cc: Taesoo kim <taesoo@gatech.edu>
Cc: Fabian Frederick <fabf@skynet.be>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agofs/affs: use affs_test_opt()
Fabian Frederick [Thu, 16 Apr 2015 19:48:24 +0000 (12:48 -0700)]
fs/affs: use affs_test_opt()

Replace mount option test by affs_test_opt().

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agofs/affs/super.c: use affs_set_opt()
Fabian Frederick [Thu, 16 Apr 2015 19:48:21 +0000 (12:48 -0700)]
fs/affs/super.c: use affs_set_opt()

Replace direct mount option assignation by affs_set_opt() macro.

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agofs/affs/affs.h: add mount option manipulation macros
Fabian Frederick [Thu, 16 Apr 2015 19:48:18 +0000 (12:48 -0700)]
fs/affs/affs.h: add mount option manipulation macros

Add clear/set/test affs mount option macros.

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agofs/affs: use AFFS_MOUNT prefix for mount options
Fabian Frederick [Thu, 16 Apr 2015 19:48:15 +0000 (12:48 -0700)]
fs/affs: use AFFS_MOUNT prefix for mount options

Currently, affs still uses direct access on mount_options.  This patch
prepares to use affs_clear/set/test_opt() like other filesystems.

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agoadfs: return correct return values
Sanidhya Kashyap [Thu, 16 Apr 2015 19:48:13 +0000 (12:48 -0700)]
adfs: return correct return values

Fix the wrong values returned by various functions such as EIO and ENOMEM.

Signed-off-by: Sanidhya Kashyap <sanidhya.gatech@gmail.com>
Cc: Fabian Frederick <fabf@skynet.be>
Cc: Joe Perches <joe@perches.com>
Cc: Taesoo kim <taesoo@gatech.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agogcov: fix softlockups
Andrey Ryabinin [Thu, 16 Apr 2015 19:48:10 +0000 (12:48 -0700)]
gcov: fix softlockups

gcov profiling if enabled with other heavy compile-time instrumentation
like KASan could trigger following softlockups:

  NMI watchdog: BUG: soft lockup - CPU#0 stuck for 22s! [swapper/0:1]
  Modules linked in:
  irq event stamp: 22823276
  hardirqs last  enabled at (22823275): [<ffffffff86e8d10d>] mutex_lock_nested+0x7d9/0x930
  hardirqs last disabled at (22823276): [<ffffffff86e9521d>] apic_timer_interrupt+0x6d/0x80
  softirqs last  enabled at (22823172): [<ffffffff811ed969>] __do_softirq+0x4db/0x729
  softirqs last disabled at (22823167): [<ffffffff811edfcf>] irq_exit+0x7d/0x15b
  CPU: 0 PID: 1 Comm: swapper/0 Tainted: G        W       3.19.0-05245-gbb33326-dirty #3
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.7.5.1-0-g8936dbb-20141113_115728-nilsson.home.kraxel.org 04/01/2014
  task: ffff88006cba8000 ti: ffff88006cbb0000 task.ti: ffff88006cbb0000
  RIP: kasan_mem_to_shadow+0x1e/0x1f
  Call Trace:
    strcmp+0x28/0x70
    get_node_by_name+0x66/0x99
    gcov_event+0x4f/0x69e
    gcov_enable_events+0x54/0x7b
    gcov_fs_init+0xf8/0x134
    do_one_initcall+0x1b2/0x288
    kernel_init_freeable+0x467/0x580
    kernel_init+0x15/0x18b
    ret_from_fork+0x7c/0xb0
  Kernel panic - not syncing: softlockup: hung tasks

Fix this by sticking cond_resched() in gcov_enable_events().

Signed-off-by: Andrey Ryabinin <a.ryabinin@samsung.com>
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Cc: Peter Oberparleiter <oberpar@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agokernel/sysctl.c: detect overflows when converting to int
Heinrich Schuchardt [Thu, 16 Apr 2015 19:48:07 +0000 (12:48 -0700)]
kernel/sysctl.c: detect overflows when converting to int

When converting unsigned long to int overflows may occur.  These currently
are not detected when writing to the sysctl file system.

E.g. on a system where int has 32 bits and long has 64 bits
  echo 0x800001234 > /proc/sys/kernel/threads-max
has the same effect as
  echo 0x1234 > /proc/sys/kernel/threads-max

The patch adds the missing check in do_proc_dointvec_conv.

With the patch an overflow will result in an error EINVAL when writing to
the the sysctl file system.

Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agocpumask: don't perform while loop in cpumask_next_and()
Sergey Senozhatsky [Thu, 16 Apr 2015 19:48:04 +0000 (12:48 -0700)]
cpumask: don't perform while loop in cpumask_next_and()

cpumask_next_and() is looking for cpumask_next() in src1 in a loop and
tests if found cpu is also present in src2. remove that loop, perform
cpumask_and() of src1 and src2 first and use that new mask to find
cpumask_next().

Apart from removing while loop, ./bloat-o-meter on x86_64 shows
add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-8 (-8)
function                                     old     new   delta
cpumask_next_and                              62      54      -8

Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Amir Vadai <amirv@mellanox.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agofs/exec.c:de_thread: move notify_count write under lock
Kirill Tkhai [Thu, 16 Apr 2015 19:48:01 +0000 (12:48 -0700)]
fs/exec.c:de_thread: move notify_count write under lock

We set sig->notify_count = -1 between RELEASE and ACQUIRE operations:

spin_unlock_irq(lock);
...
if (!thread_group_leader(tsk)) {
...
                for (;;) {
sig->notify_count = -1;
                        write_lock_irq(&tasklist_lock);

There are no restriction on it so other processors may see this STORE
mixed with other STOREs in both areas limited by the spinlocks.

Probably, it may be reordered with the above

sig->group_exit_task = tsk;
sig->notify_count = zap_other_threads(tsk);

in some way.

Set it under tasklist_lock locked to be sure nothing will be reordered.

Signed-off-by: Kirill Tkhai <ktkhai@parallels.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agoprctl: avoid using mmap_sem for exe_file serialization
Davidlohr Bueso [Thu, 16 Apr 2015 19:47:59 +0000 (12:47 -0700)]
prctl: avoid using mmap_sem for exe_file serialization

Oleg cleverly suggested using xchg() to set the new mm->exe_file instead
of calling set_mm_exe_file() which requires some form of serialization --
mmap_sem in this case.  For archs that do not have atomic rmw instructions
we still fallback to a spinlock alternative, so this should always be
safe.  As such, we only need the mmap_sem for looking up the backing
vm_file, which can be done sharing the lock.  Naturally, this means we
need to manually deal with both the new and old file reference counting,
and we need not worry about the MMF_EXE_FILE_CHANGED bits, which can
probably be deleted in the future anyway.

Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Suggested-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Reviewed-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agomm: rcu-protected get_mm_exe_file()
Konstantin Khlebnikov [Thu, 16 Apr 2015 19:47:56 +0000 (12:47 -0700)]
mm: rcu-protected get_mm_exe_file()

This patch removes mm->mmap_sem from mm->exe_file read side.
Also it kills dup_mm_exe_file() and moves exe_file duplication into
dup_mmap() where both mmap_sems are locked.

[akpm@linux-foundation.org: fix comment typo]
Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Cc: Davidlohr Bueso <dbueso@suse.de>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: "Paul E. McKenney" <paulmck@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agoDoc/sysctl/kernel.txt: document threads-max
Heinrich Schuchardt [Thu, 16 Apr 2015 19:47:53 +0000 (12:47 -0700)]
Doc/sysctl/kernel.txt: document threads-max

File /proc/sys/kernel/threads-max controls the maximum number of threads
that can be created using fork().

[akpm@linux-foundation.org: fix typo, per Guenter]
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agokernel/sysctl.c: threads-max observe limits
Heinrich Schuchardt [Thu, 16 Apr 2015 19:47:50 +0000 (12:47 -0700)]
kernel/sysctl.c: threads-max observe limits

Users can change the maximum number of threads by writing to
/proc/sys/kernel/threads-max.

With the patch the value entered is checked against the same limits that
apply when fork_init is called.

Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agokernel/fork.c: avoid division by zero
Heinrich Schuchardt [Thu, 16 Apr 2015 19:47:47 +0000 (12:47 -0700)]
kernel/fork.c: avoid division by zero

PAGE_SIZE is not guaranteed to be equal to or less than 8 times the
THREAD_SIZE.

E.g.  architecture hexagon may have page size 1M and thread size 4096.
This would lead to a division by zero in the calculation of max_threads.

With 32-bit calculation there is no solution which delivers valid results
for all possible combinations of the parameters.  The code is only called
once.  Hence a 64-bit calculation can be used as solution.

[akpm@linux-foundation.org: use clamp_t(), per Oleg]
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agokernel/fork.c: new function for max_threads
Heinrich Schuchardt [Thu, 16 Apr 2015 19:47:44 +0000 (12:47 -0700)]
kernel/fork.c: new function for max_threads

PAGE_SIZE is not guaranteed to be equal to or less than 8 times the
THREAD_SIZE.

E.g.  architecture hexagon may have page size 1M and thread size 4096.
This would lead to a division by zero in the calculation of max_threads.

With this patch the buggy code is moved to a separate function
set_max_threads.  The error is not fixed.

After fixing the problem in a separate patch the new function can be
reused to adjust max_threads after adding or removing memory.

Argument mempages of function fork_init() is removed as totalram_pages is
an exported symbol.

The creation of separate patches for refactoring to a new function and for
fixing the logic was suggested by Ingo Molnar.

Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agofork_init: update max_threads comment
Jean Delvare [Thu, 16 Apr 2015 19:47:41 +0000 (12:47 -0700)]
fork_init: update max_threads comment

The comment explaining what value max_threads is set to is outdated.  The
maximum memory consumption ratio for thread structures was 1/2 until
February 2002, then it was briefly changed to 1/16 before being set to 1/8
which we still use today.  The comment was never updated to reflect that
change, it's about time.

Signed-off-by: Jean Delvare <jdelvare@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agofork: report pid reservation failure properly
Michal Hocko [Thu, 16 Apr 2015 19:47:38 +0000 (12:47 -0700)]
fork: report pid reservation failure properly

copy_process will report any failure in alloc_pid as ENOMEM currently
which is misleading because the pid allocation might fail not only when
the memory is short but also when the pid space is consumed already.

The current man page even mentions this case:

: EAGAIN
:
:       A system-imposed limit on the number of threads was encountered.
:       There are a number of limits that may trigger this error: the
:       RLIMIT_NPROC soft resource limit (set via setrlimit(2)), which
:       limits the number of processes and threads for a real user ID, was
:       reached; the kernel's system-wide limit on the number of processes
:       and threads, /proc/sys/kernel/threads-max, was reached (see
:       proc(5)); or the maximum number of PIDs, /proc/sys/kernel/pid_max,
:       was reached (see proc(5)).

so the current behavior is also incorrect wrt.  documentation.  POSIX man
page also suggest returing EAGAIN when the process count limit is reached.

This patch simply propagates error code from alloc_pid and makes sure we
return -EAGAIN due to reservation failure.  This will make behavior of
fork closer to both our documentation and POSIX.

alloc_pid might alsoo fail when the reaper in the pid namespace is dead
(the namespace basically disallows all new processes) and there is no
good error code which would match documented ones. We have traditionally
returned ENOMEM for this case which is misleading as well but as per
Eric W. Biederman this behavior is documented in man pid_namespaces(7)

: If the "init" process of a PID namespace terminates, the kernel
: terminates all of the processes in the namespace via a SIGKILL signal.
: This behavior reflects the fact that the "init" process is essential for
: the correct operation of a PID namespace.  In this case, a subsequent
: fork(2) into this PID namespace will fail with the error ENOMEM; it is
: not possible to create a new processes in a PID namespace whose "init"
: process has terminated.

and introducing a new error code would be too risky so let's stick to
ENOMEM for this case.

Signed-off-by: Michal Hocko <mhocko@suse.cz>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Michael Kerrisk <mtk.manpages@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agosignal: remove warning about using SI_TKILL in rt_[tg]sigqueueinfo
Vladimir Davydov [Thu, 16 Apr 2015 19:47:35 +0000 (12:47 -0700)]
signal: remove warning about using SI_TKILL in rt_[tg]sigqueueinfo

Sending SI_TKILL from rt_[tg]sigqueueinfo was deprecated, so now we issue
a warning on the first attempt of doing it.  We use WARN_ON_ONCE, which is
not informative and, what is worse, taints the kernel, making the trinity
syscall fuzzer complain false-positively from time to time.

It does not look like we need this warning at all, because the behaviour
changed quite a long time ago (2.6.39), and if an application relies on
the old API, it gets EPERM anyway and can issue a warning by itself.

So let us zap the warning in kernel.

Signed-off-by: Vladimir Davydov <vdavydov@parallels.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Cc: Richard Weinberger <richard@nod.at>
Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agoptrace: ptrace_detach() can no longer race with SIGKILL
Oleg Nesterov [Thu, 16 Apr 2015 19:47:32 +0000 (12:47 -0700)]
ptrace: ptrace_detach() can no longer race with SIGKILL

ptrace_detach() re-checks ->ptrace under tasklist lock and calls
release_task() if __ptrace_detach() returns true.  This was needed because
the __TASK_TRACED tracee could be killed/untraced, and it could even pass
exit_notify() before we take tasklist_lock.

But this is no longer possible after 9899d11f6544 "ptrace: ensure
arch_ptrace/ptrace_request can never race with SIGKILL".  We can turn
these checks into WARN_ON() and remove release_task().

While at it, document the setting of child->exit_code.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Pavel Labath <labath@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agoptrace: fix race between ptrace_resume() and wait_task_stopped()
Oleg Nesterov [Thu, 16 Apr 2015 19:47:29 +0000 (12:47 -0700)]
ptrace: fix race between ptrace_resume() and wait_task_stopped()

ptrace_resume() is called when the tracee is still __TASK_TRACED.  We set
tracee->exit_code and then wake_up_state() changes tracee->state.  If the
tracer's sub-thread does wait() in between, task_stopped_code(ptrace => T)
wrongly looks like another report from tracee.

This confuses debugger, and since wait_task_stopped() clears ->exit_code
the tracee can miss a signal.

Test-case:

#include <stdio.h>
#include <unistd.h>
#include <sys/wait.h>
#include <sys/ptrace.h>
#include <pthread.h>
#include <assert.h>

int pid;

void *waiter(void *arg)
{
int stat;

for (;;) {
assert(pid == wait(&stat));
assert(WIFSTOPPED(stat));
if (WSTOPSIG(stat) == SIGHUP)
continue;

assert(WSTOPSIG(stat) == SIGCONT);
printf("ERR! extra/wrong report:%x\n", stat);
}
}

int main(void)
{
pthread_t thread;

pid = fork();
if (!pid) {
assert(ptrace(PTRACE_TRACEME, 0,0,0) == 0);
for (;;)
kill(getpid(), SIGHUP);
}

assert(pthread_create(&thread, NULL, waiter, NULL) == 0);

for (;;)
ptrace(PTRACE_CONT, pid, 0, SIGCONT);

return 0;
}

Note for stable: the bug is very old, but without 9899d11f6544 "ptrace:
ensure arch_ptrace/ptrace_request can never race with SIGKILL" the fix
should use lock_task_sighand(child).

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Reported-by: Pavel Labath <labath@google.com>
Tested-by: Pavel Labath <labath@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>
9 years agofs/fat: comment fix, fat_bits can be also 32
Alexander Kuleshov [Thu, 16 Apr 2015 19:47:26 +0000 (12:47 -0700)]
fs/fat: comment fix, fat_bits can be also 32

Signed-off-by: Alexander Kuleshov <kuleshovmail@gmail.com>
Acked-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agofs/fat: remove unnecessary includes
Alexander Kuleshov [Thu, 16 Apr 2015 19:47:24 +0000 (12:47 -0700)]
fs/fat: remove unnecessary includes

'fat.h' includes <linux/buffer_head.h> which includes <linux/fs.h> which
includes all the header files required for all *.c files fat filesystem.

[akpm@linux-foundation.org: fs/fat/iode.c needs seq_file.h]
[sfr@canb.auug.org.au: put one actually necessary include file back]
Signed-off-by: Alexander Kuleshov <kuleshovmail@gmail.com>
Acked-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
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>
9 years agofs/fat: remove unnecessary defintion
Alexander Kuleshov [Thu, 16 Apr 2015 19:47:21 +0000 (12:47 -0700)]
fs/fat: remove unnecessary defintion

'*sb' never used, so let's remote it and pass inode->i_sb directly to the
MSDOS_SB.

Signed-off-by: Alexander Kuleshov <kuleshovmail@gmail.com>
Acked-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agohfsplus: don't store special "osx" xattr prefix on-disk
Thomas Hebb [Thu, 16 Apr 2015 19:47:18 +0000 (12:47 -0700)]
hfsplus: don't store special "osx" xattr prefix on-disk

On Mac OS X, HFS+ extended attributes are not namespaced.  Since we want
to be compatible with OS X filesystems and yet still support the Linux
namespacing system, the hfsplus driver implements a special "osx"
namespace that is reported for any attribute that is not namespaced
on-disk.  However, the current code for getting and setting these
unprefixed attributes is broken.

hfsplus_osx_setattr() and hfsplus_osx_getattr() are passed names that have
already had their "osx." prefixes stripped by the generic functions.  The
functions first, quite correctly, check those names to make sure that they
aren't prefixed with a known namespace, which would allow namespace access
restrictions to be bypassed.  However, the functions then prepend "osx."
to the name they're given before passing it on to hfsplus_getattr() and
hfsplus_setattr().  Not only does this cause the "osx." prefix to be
stored on-disk, defeating its purpose, it also breaks the check for the
special "com.apple.FinderInfo" attribute, which is reported for all files,
and as a consequence makes some userspace applications (e.g.  GNU patch)
fail even when extended attributes are not otherwise in use.

There are five commits which have touched this particular code:

  127e5f5ae51e ("hfsplus: rework functionality of getting, setting and deleting of extended attributes")
  b168fff72109 ("hfsplus: use xattr handlers for removexattr")
  bf29e886b242 ("hfsplus: correct usage of HFSPLUS_ATTR_MAX_STRLEN for non-English attributes")
  fcacbd95e121 ("fs/hfsplus: move xattr_name allocation in hfsplus_getxattr()")
  ec1bbd346f18 ("fs/hfsplus: move xattr_name allocation in hfsplus_setxattr()")

The first commit creates the functions to begin with.  The namespace is
prepended by the original code, which I believe was correct at the time,
since hfsplus_?etattr() stripped the prefix if found.  The second commit
removes this behavior from hfsplus_?etattr() and appears to have been
intended to also remove the prefixing from hfsplus_osx_?etattr().
However, what it actually does is remove a necessary strncpy() call
completely, breaking the osx namespace entirely.  The third commit re-adds
the strncpy() call as it was originally, but doesn't mention it in its
commit message.  The final two commits refactor the code and don't affect
its functionality.

This commit does what b168fff attempted to do (prevent the prefix from
being added), but does it properly, instead of passing in an empty buffer
(which is what b168fff actually did).

Fixes: b168fff72109 ("hfsplus: use xattr handlers for removexattr")
Signed-off-by: Thomas Hebb <tommyhebb@gmail.com>
Cc: Hin-Tak Leung <htl10@users.sourceforge.net>
Cc: Sergei Antonov <saproj@gmail.com>
Cc: Anton Altaparmakov <anton@tuxera.com>
Cc: Fabian Frederick <fabf@skynet.be>
Cc: Christian Kujau <lists@nerdbynature.de>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Viacheslav Dubeyko <slava@dubeyko.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agohfsplus: fix expand when not enough available space
Sergei Antonov [Thu, 16 Apr 2015 19:47:15 +0000 (12:47 -0700)]
hfsplus: fix expand when not enough available space

Fix a bug which is reproduced as follows. Create a file:

 echo abc > test_file

Try to expand the file beyond available space:

 truncate --size=<size exceeding available space> test_file

Since HFS+ does not support file size > allocated size, truncate should
fail.  However, it ends successfully.  The driver returns success despite
having been unable to allocate the requested space for the file.  Also
filesystem check finds an error:

 Checking catalog file.
 Incorrect size for file test_file
 (It should be 469094400 instead of 1000000000)

Add a piece of code analogous to code in the fat driver.  Now a proper
error is returned and filesystem remains consistent.

Signed-off-by: Sergei Antonov <saproj@gmail.com>
Cc: Vyacheslav Dubeyko <slava@dubeyko.com>
Cc: Hin-Tak Leung <htl10@users.sourceforge.net>
Reviewed-by: Anton Altaparmakov <anton@tuxera.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Sougata Santra <sougata@tuxera.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agohfsplus: incorrect return value
Chengyu Song [Thu, 16 Apr 2015 19:47:12 +0000 (12:47 -0700)]
hfsplus: incorrect return value

In case of memory allocation error, the return should be -ENOMEM, instead
of -ENOSPC.

Signed-off-by: Chengyu Song <csong84@gatech.edu>
Reviewed-by: Sergei Antonov <saproj@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agofs/hfsplus: replace if/BUG by BUG_ON
Fabian Frederick [Thu, 16 Apr 2015 19:47:09 +0000 (12:47 -0700)]
fs/hfsplus: replace if/BUG by BUG_ON

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agofs/hfsplus: use bool instead of int for is_known_namespace() return value
Fabian Frederick [Thu, 16 Apr 2015 19:47:07 +0000 (12:47 -0700)]
fs/hfsplus: use bool instead of int for is_known_namespace() return value

is_known_namespace() only returns true/false.  Also remove inline and let
compiler decide what to do with static functions.

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agofs/hfsplus: atomically set inode->i_flags
Fabian Frederick [Thu, 16 Apr 2015 19:47:04 +0000 (12:47 -0700)]
fs/hfsplus: atomically set inode->i_flags

According to commit 5f16f3225b06 ("ext4: atomically set inode->i_flags in
ext4_set_inode_flags()").

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Cc: "Theodore Ts'o" <tytso@mit.edu>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>