platform/kernel/linux-rpi.git
7 years agoiw_cxgb4: drop listen destroy replies if no ep found
Steve Wise [Tue, 5 Sep 2017 18:52:33 +0000 (11:52 -0700)]
iw_cxgb4: drop listen destroy replies if no ep found

If the thread waiting for a CLOSE_LISTSRV_RPL times out and bails,
then we need to handle a subsequent CPL if it arrives and the stid has
been released.  In this case silently drop it.

Cc: stable@vger.kernel.org
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoiw_cxgb4: put ep reference in pass_accept_req()
Steve Wise [Wed, 13 Sep 2017 16:52:32 +0000 (09:52 -0700)]
iw_cxgb4: put ep reference in pass_accept_req()

The listening endpoint should always be dereferenced at the end of
pass_accept_req().

Fixes: f86fac79afec ("RDMA/iw_cxgb4: atomic find and reference for listening endpoints")

Cc: stable@vger.kernel.org
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoMerge tag 'for-linus-4.14b-rc2-tag' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 22 Sep 2017 16:40:47 +0000 (06:40 -1000)]
Merge tag 'for-linus-4.14b-rc2-tag' of git://git./linux/kernel/git/xen/tip

Pull xen fixes from Juergen Gross:
 "A fix for a missing __init annotation and two cleanup patches"

* tag 'for-linus-4.14b-rc2-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
  xen, arm64: drop dummy lookup_address()
  xen: don't compile pv-specific parts if XEN_PV isn't configured
  xen: x86: mark xen_find_pt_base as __init

7 years agoMerge tag 'powerpc-4.14-3' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc...
Linus Torvalds [Fri, 22 Sep 2017 16:39:09 +0000 (06:39 -1000)]
Merge tag 'powerpc-4.14-3' of git://git./linux/kernel/git/powerpc/linux

Pull powerpc fixes from Michael Ellerman:
 "It turns out our single-fix pull from last week was too good to be
  true. I missed a few fixes in that pull that had already come in
  because I was on leave, but also we hadn't found the bugs yet. So this
  week it's a bit bigger, though not ridiculous. Hopefully things will
  settle down from here on.

  Four fixes for the new instruction emulation code. A fix for CPU
  offline on bare metal machines when certain idle states are not
  supported, and a fix for a device_node refcounting oops during CPU
  hotplug, caused by recent changes.

  Going to stable are a fix for an oops during core dump on machines
  that have TM (Transactional Memory) disabled. Reordering some EEH
  initialisation to avoid trashing memory, and another device_node
  refcounting fix.

  And a few other minor things.

 Thanks to: Anton Blanchard, Benjamin Herrenschmidt, Cyril Bur, Gautham
 R. Shenoy, Gustavo Romero, Kamalesh Babulal, Matthew Weber, Matt Weber,
 Naveen N. Rao, Nicholas Piggin, Pavithra Prakash, Ravi Bangoria, Ronak
 Desai, Scott Wood, Tyrel Datwyler"

* tag 'powerpc-4.14-3' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  powerpc/pseries: Fix parent_dn reference leak in add_dt_node()
  powerpc/pseries: Fix "OF: ERROR: Bad of_node_put() on /cpus" during DLPAR
  powerpc/eeh: Create PHB PEs after EEH is initialized
  powerpc/kprobes: Update optprobes to use emulate_update_regs()
  powerpc/powernv: Clear LPCR[PECE1] via stop-api only for deep state offline
  powerpc/sstep: mullw should calculate a 64 bit signed result
  powerpc/sstep: Fix issues with mcrf
  powerpc/sstep: Fix issues with set_cr0()
  powerpc/tm: Flush TM only if CPU has TM feature
  powerpc/sysrq: Fix oops whem ppmu is not registered
  powerpc/configs: Update for CONFIG_SND changes
  powerpc/e6500: Update machine check for L1D cache err

7 years agoMerge branch '4.14-fixes' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus
Linus Torvalds [Fri, 22 Sep 2017 16:37:50 +0000 (06:37 -1000)]
Merge branch '4.14-fixes' of git://git.linux-mips.org/ralf/upstream-linus

Pull MIPS fixes from Ralf Baechle:

 - Fix a build error on MSP71xx which used to rely on somehow magically
   <asm/setup.h> being pulled in which no longer happens.

 - Fix the __write_64bit_c0_split inline assembler where there was the
   theoretical possibility of GCC interpret the constraints such that
   bad code could result.

 - A __init was causing section mismatch errors on Alchemy. Just to be
   on the safe side, Manuel's patch does away with all of them.

 - Fix perf event init.

* '4.14-fixes' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus:
  MIPS: PCI: fix pcibios_map_irq section mismatch
  MIPS: Fix input modify in __write_64bit_c0_split()
  MIPS: MSP71xx: Include asm/setup.h
  MIPS: Fix perf event init

7 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux
Linus Torvalds [Fri, 22 Sep 2017 16:32:07 +0000 (06:32 -1000)]
Merge branch 'for-linus' of git://git./linux/kernel/git/s390/linux

Pull s390 fixes from Martin Schwidefsky:

 - A couple of bug fixes: memory management, perf, cio, dasd and
   scm_blk.

 - A larger change in regard to the CPU topology to improve performance
   for systems running under z/VM or KVM.

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
  s390/topology: enable / disable topology dynamically
  s390/topology: alternative topology for topology-less machines
  s390/mm: fix write access check in gup_huge_pmd()
  s390/mm: make pmdp_invalidate() do invalidation only
  s390/cio: recover from bad paths
  s390/scm_blk: consistently use blk_status_t as error type
  s390/dasd: fix race during dasd initialization
  s390/perf: fix bug when creating per-thread event

7 years agoMerge branch 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6
Linus Torvalds [Fri, 22 Sep 2017 16:15:27 +0000 (06:15 -1000)]
Merge branch 'linus' of git://git./linux/kernel/git/herbert/crypto-2.6

Pull crypto fixes from Herbert Xu:
 - Fix compiler warnings in inside-secure
 - Fix LS1021A support in caam
 - Avoid using RBP in x86 crypto code
 - Fix bug in talitos that prevents hashing with algif
 - Fix bugs talitos hashing code that cause incorrect hash result
 - Fix memory freeing path bug in drbg
 - Fix af_alg crash when two SG lists are chained

* 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6:
  crypto: af_alg - update correct dst SGL entry
  crypto: caam - fix LS1021A support on ARMv7 multiplatform kernel
  crypto: inside-secure - fix gcc-4.9 warnings
  crypto: talitos - Don't provide setkey for non hmac hashing algs
  crypto: talitos - fix hashing
  crypto: talitos - fix sha224
  crypto: x86/twofish - Fix RBP usage
  crypto: sha512-avx2 - Fix RBP usage
  crypto: x86/sha256-ssse3 - Fix RBP usage
  crypto: x86/sha256-avx2 - Fix RBP usage
  crypto: x86/sha256-avx - Fix RBP usage
  crypto: x86/sha1-ssse3 - Fix RBP usage
  crypto: x86/sha1-avx2 - Fix RBP usage
  crypto: x86/des3_ede - Fix RBP usage
  crypto: x86/cast6 - Fix RBP usage
  crypto: x86/cast5 - Fix RBP usage
  crypto: x86/camellia - Fix RBP usage
  crypto: x86/blowfish - Fix RBP usage
  crypto: drbg - fix freeing of resources

7 years agoIB/core: Fix for core panic
Alex Estrin [Thu, 31 Aug 2017 16:30:34 +0000 (09:30 -0700)]
IB/core: Fix for core panic

Build with the latest patches resulted in panic:
11384.486289] BUG: unable to handle kernel NULL pointer dereference at
         (null)
[11384.486293] IP:           (null)
[11384.486295] PGD 0
[11384.486295] P4D 0
[11384.486296]
[11384.486299] Oops: 0010 [#1] SMP
......... snip ......
[11384.486401] CPU: 0 PID: 968 Comm: kworker/0:1H Tainted: G        W  O
    4.13.0-a-stream-20170825 #1
[11384.486402] Hardware name: Intel Corporation S2600WT2R/S2600WT2R,
BIOS SE5C610.86B.01.01.0014.121820151719 12/18/2015
[11384.486418] Workqueue: ib-comp-wq ib_cq_poll_work [ib_core]
[11384.486419] task: ffff880850579680 task.stack: ffffc90007fec000
[11384.486420] RIP: 0010:          (null)
[11384.486420] RSP: 0018:ffffc90007fef970 EFLAGS: 00010206
[11384.486421] RAX: ffff88084cfe8000 RBX: ffff88084dce4000 RCX:
ffffc90007fef978
[11384.486422] RDX: 0000000000000000 RSI: 0000000000000001 RDI:
ffff88084cfe8000
[11384.486422] RBP: ffffc90007fefab0 R08: 0000000000000000 R09:
ffff88084dce4080
[11384.486423] R10: ffffffffa02d7f60 R11: 0000000000000000 R12:
ffff88105af65a00
[11384.486423] R13: ffff88084dce4000 R14: 000000000000c000 R15:
000000000000c000
[11384.486424] FS:  0000000000000000(0000) GS:ffff88085f400000(0000)
knlGS:0000000000000000
[11384.486425] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[11384.486425] CR2: 0000000000000000 CR3: 0000000001c09000 CR4:
00000000001406f0
[11384.486426] Call Trace:
[11384.486431]  ? is_valid_mcast_lid.isra.21+0xfb/0x110 [ib_core]
[11384.486436]  ib_attach_mcast+0x6f/0xa0 [ib_core]
[11384.486441]  ipoib_mcast_attach+0x81/0x190 [ib_ipoib]
[11384.486443]  ipoib_mcast_join_complete+0x354/0xb40 [ib_ipoib]
[11384.486448]  mcast_work_handler+0x330/0x6c0 [ib_core]
[11384.486452]  join_handler+0x101/0x220 [ib_core]
[11384.486455]  ib_sa_mcmember_rec_callback+0x54/0x80 [ib_core]
[11384.486459]  recv_handler+0x3a/0x60 [ib_core]
[11384.486462]  ib_mad_recv_done+0x423/0x9b0 [ib_core]
[11384.486466]  __ib_process_cq+0x5d/0xb0 [ib_core]
[11384.486469]  ib_cq_poll_work+0x20/0x60 [ib_core]
[11384.486472]  process_one_work+0x149/0x360
[11384.486474]  worker_thread+0x4d/0x3c0
[11384.486487]  kthread+0x109/0x140
[11384.486488]  ? rescuer_thread+0x380/0x380
[11384.486489]  ? kthread_park+0x60/0x60
[11384.486490]  ? kthread_park+0x60/0x60
[11384.486493]  ret_from_fork+0x25/0x30
[11384.486493] Code:  Bad RIP value.
[11384.486493] Code:  Bad RIP value.
[11384.486496] RIP:           (null) RSP: ffffc90007fef970
[11384.486497] CR2: 0000000000000000
[11384.486531] ---[ end trace b1acec6fb4ff6e75 ]---
[11384.532133] Kernel panic - not syncing: Fatal exception
[11384.536541] Kernel Offset: disabled
[11384.969491] ---[ end Kernel panic - not syncing: Fatal exception
[11384.976875] sched: Unexpected reschedule of offline CPU#1!
[11384.983646] ------------[ cut here ]------------

Rdma device driver may not have implemented (*get_link_layer)()
so it can not be called directly. Should use appropriate helper function.

Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
Fixes: 523633359224 ("IB/core: Fix the validations of a multicast LID in attach or detach operations")
Cc: stable@kernel.org # 4.13
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Alex Estrin <alex.estrin@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agodrivers/perf: arm_pmu_acpi: Release memory obtained by kasprintf
Arvind Yadav [Wed, 20 Sep 2017 06:56:38 +0000 (12:26 +0530)]
drivers/perf: arm_pmu_acpi: Release memory obtained by kasprintf

Free memory region, if arm_pmu_acpi_probe is not successful.

Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
7 years agoiommu/of: Remove PCI host bridge node check
Robin Murphy [Thu, 21 Sep 2017 10:20:58 +0000 (11:20 +0100)]
iommu/of: Remove PCI host bridge node check

of_pci_iommu_init() tries to be clever and stop its alias walk at the
device represented by master_np, in case of weird PCI topologies where
the bridge to the IOMMU and the rest of the system is not at the root.
It turns out this is a bit short-sighted, since there are plenty of
other callers of pci_for_each_dma_alias() which would also need the same
behaviour in that situation, and the only platform so far with such a
topology (Cavium ThunderX2) already solves it more generally via a PCI
quirk. As this check is effectively redundant, and returning a boolean
value as an int is a bit broken anyway, let's just get rid of it.

Reported-by: Jean-Philippe Brucker <jean-philippe.brucker@arm.com>
Fixes: d87beb749281 ("iommu/of: Handle PCI aliases properly")
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Tested-by: Jean-Philippe Brucker <jean-philippe.brucker@arm.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
7 years agonet: prevent dst uses after free
Eric Dumazet [Thu, 21 Sep 2017 16:15:46 +0000 (09:15 -0700)]
net: prevent dst uses after free

In linux-4.13, Wei worked hard to convert dst to a traditional
refcounted model, removing GC.

We now want to make sure a dst refcount can not transition from 0 back
to 1.

The problem here is that input path attached a not refcounted dst to an
skb. Then later, because packet is forwarded and hits skb_dst_force()
before exiting RCU section, we might try to take a refcount on one dst
that is about to be freed, if another cpu saw 1 -> 0 transition in
dst_release() and queued the dst for freeing after one RCU grace period.

Lets unify skb_dst_force() and skb_dst_force_safe(), since we should
always perform the complete check against dst refcount, and not assume
it is not zero.

Bugzilla : https://bugzilla.kernel.org/show_bug.cgi?id=197005

[  989.919496]  skb_dst_force+0x32/0x34
[  989.919498]  __dev_queue_xmit+0x1ad/0x482
[  989.919501]  ? eth_header+0x28/0xc6
[  989.919502]  dev_queue_xmit+0xb/0xd
[  989.919504]  neigh_connected_output+0x9b/0xb4
[  989.919507]  ip_finish_output2+0x234/0x294
[  989.919509]  ? ipt_do_table+0x369/0x388
[  989.919510]  ip_finish_output+0x12c/0x13f
[  989.919512]  ip_output+0x53/0x87
[  989.919513]  ip_forward_finish+0x53/0x5a
[  989.919515]  ip_forward+0x2cb/0x3e6
[  989.919516]  ? pskb_trim_rcsum.part.9+0x4b/0x4b
[  989.919518]  ip_rcv_finish+0x2e2/0x321
[  989.919519]  ip_rcv+0x26f/0x2eb
[  989.919522]  ? vlan_do_receive+0x4f/0x289
[  989.919523]  __netif_receive_skb_core+0x467/0x50b
[  989.919526]  ? tcp_gro_receive+0x239/0x239
[  989.919529]  ? inet_gro_receive+0x226/0x238
[  989.919530]  __netif_receive_skb+0x4d/0x5f
[  989.919532]  netif_receive_skb_internal+0x5c/0xaf
[  989.919533]  napi_gro_receive+0x45/0x81
[  989.919536]  ixgbe_poll+0xc8a/0xf09
[  989.919539]  ? kmem_cache_free_bulk+0x1b6/0x1f7
[  989.919540]  net_rx_action+0xf4/0x266
[  989.919543]  __do_softirq+0xa8/0x19d
[  989.919545]  irq_exit+0x5d/0x6b
[  989.919546]  do_IRQ+0x9c/0xb5
[  989.919548]  common_interrupt+0x93/0x93
[  989.919548]  </IRQ>

Similarly dst_clone() can use dst_hold() helper to have additional
debugging, as a follow up to commit 44ebe79149ff ("net: add debug
atomic_inc_not_zero() in dst_hold()")

In net-next we will convert dst atomic_t to refcount_t for peace of
mind.

Fixes: a4c2fd7f7891 ("net: remove DST_NOCACHE flag")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Wei Wang <weiwan@google.com>
Reported-by: Paweł Staszewski <pstaszewski@itcare.pl>
Bisected-by: Paweł Staszewski <pstaszewski@itcare.pl>
Acked-by: Wei Wang <weiwan@google.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: phy: Fix truncation of large IRQ numbers in phy_attached_print()
Geert Uytterhoeven [Thu, 21 Sep 2017 11:27:02 +0000 (13:27 +0200)]
net: phy: Fix truncation of large IRQ numbers in phy_attached_print()

Given NR_IRQS is 2048 on sparc64, and even 32784 on alpha, 3 digits is
not enough to represent interrupt numbers on all architectures.  Hence
PHY interrupt numbers may be truncated during printing.

Increase the buffer size from 4 to 8 bytes to fix this.

Fixes: 5e369aefdce4818c ("net: stmmac: Delete dead code for MDIO registration")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoMerge branch 'libnvdimm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdim...
Linus Torvalds [Fri, 22 Sep 2017 01:51:19 +0000 (15:51 -1000)]
Merge branch 'libnvdimm-fixes' of git://git./linux/kernel/git/nvdimm/nvdimm

Pull libnvdimm fixes from Dan Williams:
 "A crash fix and corresponding regression test enabling for the crash
  scenario. The unit test for this crash is available in ndctl-v58.2.

  This branch has received a build success notification from the
  0day-kbuild robot over 148 configs. The fix is tagged for -stable /
  backport to 4.13"

* 'libnvdimm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm:
  libnvdimm, namespace: fix btt claim class crash
  tools/testing/nvdimm: disable labels for nfit_test.1

7 years agodt-bindings: clk: stm32h7: fix clock-cell size
Gabriel Fernandez [Tue, 19 Sep 2017 11:44:06 +0000 (13:44 +0200)]
dt-bindings: clk: stm32h7: fix clock-cell size

The clock-cell size is 1 on stm32h7 plaform.

Signed-off-by: Gabriel Fernandez <gabriel.fernandez@st.com>
Fixes: 3e4d618b0722 ("clk: stm32h7: Add stm32h743 clock driver")
Signed-off-by: Rob Herring <robh@kernel.org>
7 years agoInput: uinput - avoid crash when sending FF request to device going away
Dmitry Torokhov [Wed, 6 Sep 2017 23:22:59 +0000 (16:22 -0700)]
Input: uinput - avoid crash when sending FF request to device going away

If FF request comes in while uinput device is going away,
uinput_request_send() will fail with -ENODEV, and uinput_request_submit()
will attempt to mark the slot as unused by calling uinput_request_done().
Unfortunately in this case we haven't initialized request->done completion
yet, and we get a crash:

[   39.402036] BUG: spinlock bad magic on CPU#1, fftest/3108
[   39.402046]  lock: 0xffff88006a93bb00, .magic: 00000000, .owner: /39, .owner_cpu: 1217155072
[   39.402055] CPU: 1 PID: 3108 Comm: fftest Tainted: G        W 4.13.0+ #15
[   39.402059] Hardware name: LENOVO 20HQS0EG02/20HQS0EG02, BIOS N1MET37W (1.22 ) 07/04/2017
[   39.402064]  0000000000000086 f0fad82f3ceaa120 ffff88006a93b9a0 ffffffff9de941bb
[   39.402077]  ffff88026df8ae00 ffff88006a93bb00 ffff88006a93b9c0 ffffffff9dca62b7
[   39.402088]  ffff88006a93bb00 ffff88006a93baf8 ffff88006a93b9e0 ffffffff9dca62e7
[   39.402099] Call Trace:
[   39.402112]  [<ffffffff9de941bb>] dump_stack+0x4d/0x63
[   39.402123]  [<ffffffff9dca62b7>] spin_dump+0x97/0x9c
[   39.402130]  [<ffffffff9dca62e7>] spin_bug+0x2b/0x2d
[   39.402138]  [<ffffffff9dca6373>] do_raw_spin_lock+0x28/0xfd
[   39.402147]  [<ffffffff9e3055cd>] _raw_spin_lock_irqsave+0x19/0x1f
[   39.402154]  [<ffffffff9dca05b7>] complete+0x1d/0x48
[   39.402162]  [<ffffffffc04f30af>] 0xffffffffc04f30af
[   39.402167]  [<ffffffffc04f468c>] 0xffffffffc04f468c
[   39.402177]  [<ffffffff9dd59c16>] ? __slab_free+0x22f/0x359
[   39.402184]  [<ffffffff9dcc13e9>] ? tk_clock_read+0xc/0xe
[   39.402189]  [<ffffffffc04f471f>] 0xffffffffc04f471f
[   39.402195]  [<ffffffff9dc9ffe5>] ? __wake_up+0x44/0x4b
[   39.402200]  [<ffffffffc04f3240>] ? 0xffffffffc04f3240
[   39.402207]  [<ffffffff9e0f57f3>] erase_effect+0xa1/0xd2
[   39.402214]  [<ffffffff9e0f58c6>] input_ff_flush+0x43/0x5c
[   39.402219]  [<ffffffffc04f32ad>] 0xffffffffc04f32ad
[   39.402227]  [<ffffffff9e0f174f>] input_flush_device+0x3d/0x51
[   39.402234]  [<ffffffff9e0f69ae>] evdev_flush+0x49/0x5c
[   39.402243]  [<ffffffff9dd62d6e>] filp_close+0x3f/0x65
[   39.402253]  [<ffffffff9dd7dcf7>] put_files_struct+0x66/0xc1
[   39.402261]  [<ffffffff9dd7ddeb>] exit_files+0x47/0x4e
[   39.402270]  [<ffffffff9dc6b329>] do_exit+0x483/0x969
[   39.402278]  [<ffffffff9dc73211>] ? recalc_sigpending_tsk+0x3d/0x44
[   39.402285]  [<ffffffff9dc6c7a2>] do_group_exit+0x42/0xb0
[   39.402293]  [<ffffffff9dc767e1>] get_signal+0x58d/0x5bf
[   39.402300]  [<ffffffff9dc03701>] do_signal+0x37/0x53e
[   39.402307]  [<ffffffff9e0f8401>] ? evdev_ioctl_handler+0xac8/0xb04
[   39.402314]  [<ffffffff9e0f8464>] ? evdev_ioctl+0x10/0x12
[   39.402321]  [<ffffffff9dd74cfa>] ? do_vfs_ioctl+0x42e/0x501
[   39.402328]  [<ffffffff9dc0170e>] prepare_exit_to_usermode+0x66/0x90
[   39.402333]  [<ffffffff9dc0181b>] syscall_return_slowpath+0xe3/0xec
[   39.402339]  [<ffffffff9e305b7b>] int_ret_from_sys_call+0x25/0x8f

While we could solve this by simply initializing the completion earlier, we
are better off rearranging the code a bit so we avoid calling complete() on
requests that we did not send out. This patch consolidates marking request
slots as free in one place (in uinput_request_submit(), the same place
where we acquire them) and having everyone else simply signal completion
of the requests.

Fixes: 00ce756ce53a ("Input: uinput - mark failed submission requests as free")
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
7 years agoInput: uinput - avoid FF flush when destroying device
Dmitry Torokhov [Sat, 2 Sep 2017 00:13:43 +0000 (17:13 -0700)]
Input: uinput - avoid FF flush when destroying device

Normally, when input device supporting force feedback effects is being
destroyed, we try to "flush" currently playing effects, so that the
physical device does not continue vibrating (or executing other effects).
Unfortunately this does not work well for uinput as flushing of the effects
deadlocks with the destroy action:

- if device is being destroyed because the file descriptor is being closed,
  then there is noone to even service FF requests;

- if device is being destroyed because userspace sent UI_DEV_DESTROY,
  while theoretically it could be possible to service FF requests,
  userspace is unlikely to do so (they'd need to make sure FF handling
  happens on a separate thread) even if kernel solves the issue with FF
  ioctls deadlocking with UI_DEV_DESTROY ioctl on udev->mutex.

To avoid lockups like the one below, let's install a custom input device
flush handler, and avoid trying to flush force feedback effects when we
destroying the device, and instead rely on uinput to shut off the device
properly.

NMI watchdog: Watchdog detected hard LOCKUP on cpu 3
...
 <<EOE>>  [<ffffffff817a0307>] _raw_spin_lock_irqsave+0x37/0x40
 [<ffffffff810e633d>] complete+0x1d/0x50
 [<ffffffffa00ba08c>] uinput_request_done+0x3c/0x40 [uinput]
 [<ffffffffa00ba587>] uinput_request_submit.part.7+0x47/0xb0 [uinput]
 [<ffffffffa00bb62b>] uinput_dev_erase_effect+0x5b/0x76 [uinput]
 [<ffffffff815d91ad>] erase_effect+0xad/0xf0
 [<ffffffff815d929d>] flush_effects+0x4d/0x90
 [<ffffffff815d4cc0>] input_flush_device+0x40/0x60
 [<ffffffff815daf1c>] evdev_cleanup+0xac/0xc0
 [<ffffffff815daf5b>] evdev_disconnect+0x2b/0x60
 [<ffffffff815d74ac>] __input_unregister_device+0xac/0x150
 [<ffffffff815d75f7>] input_unregister_device+0x47/0x70
 [<ffffffffa00bac45>] uinput_destroy_device+0xb5/0xc0 [uinput]
 [<ffffffffa00bb2de>] uinput_ioctl_handler.isra.9+0x65e/0x740 [uinput]
 [<ffffffff811231ab>] ? do_futex+0x12b/0xad0
 [<ffffffffa00bb3f8>] uinput_ioctl+0x18/0x20 [uinput]
 [<ffffffff81241248>] do_vfs_ioctl+0x298/0x480
 [<ffffffff81337553>] ? security_file_ioctl+0x43/0x60
 [<ffffffff812414a9>] SyS_ioctl+0x79/0x90
 [<ffffffff817a04ee>] entry_SYSCALL_64_fastpath+0x12/0x71

Reported-by: Rodrigo Rivas Costa <rodrigorivascosta@gmail.com>
Reported-by: Clément VUCHENER <clement.vuchener@gmail.com>
Fixes: https://bugzilla.kernel.org/show_bug.cgi?id=193741
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
7 years agoMerge branch 'smc-bug-fixes'
David S. Miller [Thu, 21 Sep 2017 22:31:03 +0000 (15:31 -0700)]
Merge branch 'smc-bug-fixes'

Ursula Braun says:

====================
net/smc: bug fixes 2017-09-20

here is a collection of small smc-patches built for net fixing
smc problems in different areas.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet/smc: no close wait in case of process shut down
Ursula Braun [Thu, 21 Sep 2017 07:16:34 +0000 (09:16 +0200)]
net/smc: no close wait in case of process shut down

Usually socket closing is delayed if there is still data available in
the send buffer to be transmitted. If a process is killed, the delay
should be avoided.

Signed-off-by: Ursula Braun <ubraun@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet/smc: introduce a delay
Ursula Braun [Thu, 21 Sep 2017 07:16:33 +0000 (09:16 +0200)]
net/smc: introduce a delay

The number of outstanding work requests is limited. If all work
requests are in use, tx processing is postponed to another scheduling
of the tx worker. Switch to a delayed worker to have a gap for tx
completion queue events before the next retry.

Signed-off-by: Ursula Braun <ubraun@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet/smc: terminate link group if out-of-sync is received
Ursula Braun [Thu, 21 Sep 2017 07:16:32 +0000 (09:16 +0200)]
net/smc: terminate link group if out-of-sync is received

An out-of-sync condition can just be detected by the client.
If the server receives a CLC DECLINE message indicating an out-of-sync
condition for the link groups, the server must clean up the out-of-sync
link group.
There is no need for an extra third parameter in smc_clc_send_decline().

Signed-off-by: Ursula Braun <ubraun@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet/smc: longer delay for client link group removal
Ursula Braun [Thu, 21 Sep 2017 07:16:31 +0000 (09:16 +0200)]
net/smc: longer delay for client link group removal

Client link group creation always follows the server linkgroup creation.
If peer creates a new server link group, client has to create a new
client link group. If peer reuses a server link group for a new
connection, client has to reuse its client link group as well. This
patch introduces a longer delay for client link group removal to make
sure this link group still exists, once the peer decides to reuse a
server link group. This avoids out-of-sync conditions for link groups.
If already scheduled, modify the delay.

Signed-off-by: Ursula Braun <ubraun@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet/smc: adapt send request completion notification
Ursula Braun [Thu, 21 Sep 2017 07:16:30 +0000 (09:16 +0200)]
net/smc: adapt send request completion notification

The solicited flag is meaningful for the receive completion queue.
Ask for next work completion of any type on the send queue.

Signed-off-by: Ursula Braun <ubraun@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet/smc: adjust net_device refcount
Ursula Braun [Thu, 21 Sep 2017 07:16:29 +0000 (09:16 +0200)]
net/smc: adjust net_device refcount

smc_pnet_fill_entry() uses dev_get_by_name() adding a refcount to ndev.
The following smc_pnet_enter() has to reduce the refcount if the entry
to be added exists already in the pnet table.

Signed-off-by: Ursula Braun <ubraun@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet/smc: take RCU read lock for routing cache lookup
Ursula Braun [Thu, 21 Sep 2017 07:16:28 +0000 (09:16 +0200)]
net/smc: take RCU read lock for routing cache lookup

smc_netinfo_by_tcpsk() looks up the routing cache. Such a lookup requires
protection by an RCU read lock.

Signed-off-by: Ursula Braun <ubraun@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet/smc: add receive timeout check
Hans Wippel [Thu, 21 Sep 2017 07:16:27 +0000 (09:16 +0200)]
net/smc: add receive timeout check

The SMC receive function currently lacks a timeout check under the
condition that no data were received and no data are available. This
patch adds such a check.

Signed-off-by: Hans Wippel <hwippel@linux.vnet.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet/smc: add missing dev_put
Hans Wippel [Thu, 21 Sep 2017 07:16:26 +0000 (09:16 +0200)]
net/smc: add missing dev_put

In the infiniband part, SMC currently uses get_netdev which calls
dev_hold on the returned net device. However, the SMC code never calls
dev_put on that net device resulting in a wrong reference count.

This patch adds a dev_put after the usage of the net device to fix the
issue.

Signed-off-by: Hans Wippel <hwippel@linux.vnet.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: stmmac: Cocci spatch "of_table"
Thomas Meyer [Thu, 21 Sep 2017 06:24:27 +0000 (08:24 +0200)]
net: stmmac: Cocci spatch "of_table"

Make sure (of/i2c/platform)_device_id tables are NULL terminated.
Found by coccinelle spatch "misc/of_table.cocci"

Signed-off-by: Thomas Meyer <thomas@m3y3r.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoMerge branch 'lan78xx-fixes'
David S. Miller [Thu, 21 Sep 2017 22:22:54 +0000 (15:22 -0700)]
Merge branch 'lan78xx-fixes'

Nisar Sayed says:

====================
lan78xx: This series of patches are for lan78xx driver.

This series of patches are for lan78xx driver.

These patches fixes potential issues associated with lan78xx driver.

v5
- Updated changes as per comments

v4
- Updated changes to handle return values as per comments
- Updated EEPROM write handling as per comments

v3
- Updated chagnes as per comments

v2
- Added patch version information
- Added fixes tag
- Updated patch description
- Updated chagnes as per comments

v1
- Splitted patches as per comments
- Dropped "fixed_phy device support" and "Fix for system suspend" changes
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agolan78xx: Use default values loaded from EEPROM/OTP after reset
Nisar Sayed [Wed, 20 Sep 2017 21:06:38 +0000 (02:36 +0530)]
lan78xx: Use default values loaded from EEPROM/OTP after reset

Use default value of auto duplex and auto speed values loaded
from EEPROM/OTP after reset. The LAN78xx allows platform
configurations to be loaded from EEPROM/OTP.
Ex: When external phy is connected, the MAC can be configured to
have correct auto speed, auto duplex, auto polarity configured
from the EEPROM/OTP.

Fixes: 55d7de9de6c3 ("Microchip's LAN7800 family USB 2/3 to 10/100/1000 Ethernet device driver")
Signed-off-by: Nisar Sayed <Nisar.Sayed@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agolan78xx: Allow EEPROM write for less than MAX_EEPROM_SIZE
Nisar Sayed [Wed, 20 Sep 2017 21:06:37 +0000 (02:36 +0530)]
lan78xx: Allow EEPROM write for less than MAX_EEPROM_SIZE

Allow EEPROM write for less than MAX_EEPROM_SIZE

Fixes: 55d7de9de6c3 ("Microchip's LAN7800 family USB 2/3 to 10/100/1000 Ethernet device driver")
Signed-off-by: Nisar Sayed <Nisar.Sayed@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agolan78xx: Fix for eeprom read/write when device auto suspend
Nisar Sayed [Wed, 20 Sep 2017 21:06:36 +0000 (02:36 +0530)]
lan78xx: Fix for eeprom read/write when device auto suspend

Fix for eeprom read/write when device auto suspend

Fixes: 55d7de9de6c3 ("Microchip's LAN7800 family USB 2/3 to 10/100/1000 Ethernet device driver")
Signed-off-by: Nisar Sayed <Nisar.Sayed@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoMerge branch 'phylib-xcvr-type'
David S. Miller [Thu, 21 Sep 2017 22:20:41 +0000 (15:20 -0700)]
Merge branch 'phylib-xcvr-type'

Florian Fainelli says:

====================
net: Bring back transceiver type for PHYLIB

With the introduction of the xLINKSETTINGS ethtool APIs, the transceiver type
was deprecated, but in that process we lost some useful information that PHYLIB
was consistently reporting about internal vs. external PHYs.

This brings back transceiver as a read-only field that is only consumed in the
legacy path where ETHTOOL_GET is called but the underlying drivers implement the
new style klink_settings API.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: phy: Keep reporting transceiver type
Florian Fainelli [Wed, 20 Sep 2017 22:52:14 +0000 (15:52 -0700)]
net: phy: Keep reporting transceiver type

With commit 2d55173e71b0 ("phy: add generic function to support
ksetting support"), we lost the ability to report the transceiver type
like we used to. Now that we have added back the transceiver type to
ethtool_link_settings, we can report it back like we used to and have no
loss of information.

Fixes: 3f1ac7a700d0 ("net: ethtool: add new ETHTOOL_xLINKSETTINGS API")
Fixes: 2d55173e71b0 ("phy: add generic function to support ksetting support")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: ethtool: Add back transceiver type
Florian Fainelli [Wed, 20 Sep 2017 22:52:13 +0000 (15:52 -0700)]
net: ethtool: Add back transceiver type

Commit 3f1ac7a700d0 ("net: ethtool: add new ETHTOOL_xLINKSETTINGS API")
deprecated the ethtool_cmd::transceiver field, which was fine in
premise, except that the PHY library was actually using it to report the
type of transceiver: internal or external.

Use the first word of the reserved field to put this __u8 transceiver
field back in. It is made read-only, and we don't expect the
ETHTOOL_xLINKSETTINGS API to be doing anything with this anyway, so this
is mostly for the legacy path where we do:

ethtool_get_settings()
-> dev->ethtool_ops->get_link_ksettings()
   -> convert_link_ksettings_to_legacy_settings()

to have no information loss compared to the legacy get_settings API.

Fixes: 3f1ac7a700d0 ("net: ethtool: add new ETHTOOL_xLINKSETTINGS API")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: qcom/emac: add software control for pause frame mode
Timur Tabi [Wed, 20 Sep 2017 20:32:53 +0000 (15:32 -0500)]
net: qcom/emac: add software control for pause frame mode

The EMAC has the option of sending only a single pause frame when
flow control is enabled and the RX queue is full.  Although sending
only one pause frame has little value, this would allow admins to
enable automatic flow control without having to worry about the EMAC
flooding nearby switches with pause frames if the kernel hangs.

The option is enabled by using the single-pause-mode private flag.

Signed-off-by: Timur Tabi <timur@codeaurora.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agohv_netvsc: fix send buffer failure on MTU change
Alex Ng [Wed, 20 Sep 2017 18:17:35 +0000 (11:17 -0700)]
hv_netvsc: fix send buffer failure on MTU change

If MTU is changed the host would reject the send buffer change.
This problem is result of recent change to allow changing send
buffer size.

Every time we change the MTU, we store the previous net_device section
count before destroying the buffer, but we don’t store the previous
section size. When we reinitialize the buffer, its size is calculated
by multiplying the previous count and previous size. Since we
continuously increase the MTU, the host returns us a decreasing count
value while the section size is reinitialized to 1728 bytes every
time.

This eventually leads to a condition where the calculated buf_size is
so small that the host rejects it.

Fixes: 8b5327975ae1 ("netvsc: allow controlling send/recv buffer size")
Signed-off-by: Alex Ng <alexng@microsoft.com>
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet_sched: remove cls_flower idr on failure
Cong Wang [Wed, 20 Sep 2017 16:18:45 +0000 (09:18 -0700)]
net_sched: remove cls_flower idr on failure

Fixes: c15ab236d69d ("net/sched: Change cls_flower to use IDR")
Cc: Chris Mi <chrism@mellanox.com>
Cc: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet_sched/hfsc: fix curve activation in hfsc_change_class()
Konstantin Khlebnikov [Wed, 20 Sep 2017 12:46:11 +0000 (15:46 +0300)]
net_sched/hfsc: fix curve activation in hfsc_change_class()

If real-time or fair-share curves are enabled in hfsc_change_class()
class isn't inserted into rb-trees yet. Thus init_ed() and init_vf()
must be called in place of update_ed() and update_vf().

Remove isn't required because for now curves cannot be disabled.

Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet_sched: always reset qdisc backlog in qdisc_reset()
Konstantin Khlebnikov [Wed, 20 Sep 2017 12:45:36 +0000 (15:45 +0300)]
net_sched: always reset qdisc backlog in qdisc_reset()

SKB stored in qdisc->gso_skb also counted into backlog.

Some qdiscs don't reset backlog to zero in ->reset(),
for example sfq just dequeue and free all queued skb.

Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Fixes: 2ccccf5fb43f ("net_sched: update hierarchical backlog too")
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoMerge tag 'kbuild-fixes-v4.14' of git://git.kernel.org/pub/scm/linux/kernel/git/masah...
Linus Torvalds [Thu, 21 Sep 2017 16:01:03 +0000 (06:01 -1000)]
Merge tag 'kbuild-fixes-v4.14' of git://git./linux/kernel/git/masahiroy/linux-kbuild

Pull Kbuild fixes from Masahiro Yamada:
 "Here are some early Kbuild fixes.

  The in-kernel firmware was removed during the previous merge window.
  Since then, some bug reports of broken rpm building are flying in ML.
  We need to fix it now.

  Summary:

   - remove firmware install from rpm-pkg / deb-pkg

   - fix mismatch between release number and UTS_VERSION for rpm-pkg"

* tag 'kbuild-fixes-v4.14' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild:
  kbuild: rpm-pkg: fix version number handling
  kbuild: deb-pkg: remove firmware package support
  kbuild: rpm-pkg: delete firmware_install to fix build error

7 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
Linus Torvalds [Thu, 21 Sep 2017 15:57:50 +0000 (05:57 -1000)]
Merge branch 'for-linus' of git://git./linux/kernel/git/viro/vfs

Pull misc fixes from Al Viro:
 "A couple of regression fixes, one for this merge window, one for the
  previous cycle"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
  ipc/shm: Fix order of parameters when calling copy_compat_shmid_to_user
  iov_iter: fix page_copy_sane for compound pages

7 years agoMerge tag 'mtd/fixes-for-4.14-rc2' of git://git.infradead.org/linux-mtd
Linus Torvalds [Thu, 21 Sep 2017 15:55:20 +0000 (05:55 -1000)]
Merge tag 'mtd/fixes-for-4.14-rc2' of git://git.infradead.org/linux-mtd

Pull mtd fixes from Boris Brezillon:
 "SPI NOR:
   - Fix the SFDP parsing code (bugs reported by Geert Uytterhoeven)

  NAND:
   - Fix a resource leak in the lpc32xx_mlc driver
   - Fix a build warning in the core"

* tag 'mtd/fixes-for-4.14-rc2' of git://git.infradead.org/linux-mtd:
  mtd: nand: remove unused blockmask variable
  mtd: nand: lpc32xx_mlc: Fix an error handling path in lpc32xx_nand_probe()
  mtd: spi-nor: fix DMA unsafe buffer issue in spi_nor_read_sfdp()
  mtd: spi-nor: Check consistency of the memory size extracted from the SFDP

7 years agoMerge tag 'drm-fixes-for-v4.14-rc2' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Thu, 21 Sep 2017 15:52:36 +0000 (05:52 -1000)]
Merge tag 'drm-fixes-for-v4.14-rc2' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
 "amdkfd, i915 and exynos fixes.

  I've ended up on unplanned + planned leave this week, but there were
  some fixes I decided to dequeue, some amdkfd bits missed the next pull
  but they are pretty trivial, so I included them.

  I'm not sure I'll see much else for rc2, lots of people are at XDC"

* tag 'drm-fixes-for-v4.14-rc2' of git://people.freedesktop.org/~airlied/linux:
  drm/exynos/hdmi: Fix unsafe list iteration
  drm: exynos: include linux/irq.h
  drm/exynos: Fix suspend/resume support
  drm/exynos: Fix locking in the suspend/resume paths
  drm/i915: Remove unused 'in_vbl' from i915_get_crtc_scanoutpos()
  drm/i915/cnp: set min brightness from VBT
  Revert "drm/i915/bxt: Disable device ready before shutdown command"
  drm/i915/bxt: set min brightness from VBT
  drm/i915: Fix an error handling in 'intel_framebuffer_init()'
  drm/i915/gvt: Fix incorrect PCI BARs reporting
  drm/amdkfd: pass queue's mqd when destroying mqd
  drm/amdkfd: remove memset before memcpy
  uapi linux/kfd_ioctl.h: only use __u32 and __u64

7 years agoMerge tag 'dma-mapping-4.14-2' of git://git.infradead.org/users/hch/dma-mapping
Linus Torvalds [Thu, 21 Sep 2017 15:51:03 +0000 (05:51 -1000)]
Merge tag 'dma-mapping-4.14-2' of git://git.infradead.org/users/hch/dma-mapping

Pull dma mapping fix from Christoph Hellwig:
 "A fix for a fix that went in this merge window from Arnd"

* tag 'dma-mapping-4.14-2' of git://git.infradead.org/users/hch/dma-mapping:
  dma-coherent: fix rmem_dma_device_init regression

7 years agoMIPS: PCI: fix pcibios_map_irq section mismatch
Manuel Lauss [Tue, 12 Sep 2017 18:36:28 +0000 (20:36 +0200)]
MIPS: PCI: fix pcibios_map_irq section mismatch

Drop  the __init from pcibios_map_irq() to make this section mis-
match go away:

WARNING: vmlinux.o(.text+0x56acd4): Section mismatch in reference from the function pcibios_scanbus() to the function .init.text:pcibios_map_irq()
The function pcibios_scanbus() references
the function __init pcibios_map_irq().
This is often because pcibios_scanbus lacks a __init
annotation or the annotation of pcibios_map_irq is wrong.

Run-Tested only on Alchemy.

Signed-off-by: Manuel Lauss <manuel.lauss@gmail.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/17267/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
7 years agoMIPS: Fix input modify in __write_64bit_c0_split()
James Hogan [Tue, 19 Sep 2017 13:11:22 +0000 (14:11 +0100)]
MIPS: Fix input modify in __write_64bit_c0_split()

The inline asm in __write_64bit_c0_split() modifies the 64-bit input
operand by shifting the high register left by 32, and constructing the
full 64-bit value in the low register (even on a 32-bit kernel), so if
that value is used again it could cause breakage as GCC would assume the
registers haven't changed when they have.

To quote the GCC extended asm documentation:
> Warning: Do not modify the contents of input-only operands (except for
> inputs tied to outputs). The compiler assumes that on exit from the
> asm statement these operands contain the same values as they had
> before executing the statement.

Avoid modifying the input by using a temporary variable as an output
which is modified instead of the input and not otherwise used. The asm
is always __volatile__ so GCC shouldn't optimise it out. The low
register of the temporary output is written before the high register of
the input is read, so we have two constraint alternatives, one where
both use the same registers (for when the input value isn't subsequently
used), and one with an early clobber on the output in case the low
output uses the same register as the high input. This allows the
resulting assembly to remain mostly unchanged.

A diff of a MIPS32r6 kernel reveals only three differences, two in
relation to write_c0_r10k_diag() in cpu_probe() (register allocation
rearranged slightly but otherwise identical), and one in relation to
write_c0_cvmmemctl2() in kvm_vz_local_flush_guesttlb_all(), but the
octeon CPU is only supported on 64-bit kernels where
__write_64bit_c0_split() isn't used so that shouldn't matter in
practice. So there currently doesn't appear to be anything broken by
this bug.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/17315/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
7 years agoMIPS: MSP71xx: Include asm/setup.h
Arnd Bergmann [Mon, 18 Sep 2017 15:55:24 +0000 (17:55 +0200)]
MIPS: MSP71xx: Include asm/setup.h

msp71xx_defconfig can not be built at the in v4.14-rc1

arch/mips/pmcs-msp71xx/msp_smp.c:72:2: error: implicit declaration of function 'set_vi_handler' [-Werror=implicit-function-declaration]

I don't know what caused the regression, but including the right
header is the obvious fix.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: linux-mips@linux-mips.org
Cc: linux-kernel@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/17309/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
7 years agoMerge tag 'irqchip-4.14-2' of git://git.kernel.org/pub/scm/linux/kernel/git/maz/arm...
Thomas Gleixner [Thu, 21 Sep 2017 09:57:08 +0000 (11:57 +0200)]
Merge tag 'irqchip-4.14-2' of git://git./linux/kernel/git/maz/arm-platforms into irq/urgent

Pull irqchip updates from Marc Zyngier

- A GICv3 initialisation fix when some CPUs fail to be brought up
- A GICv4 compile fix for GCC 4.5 (!)
- A MIPS-GIC fix for the PCIe support

7 years agoRevert "genirq: Restrict effective affinity to interrupts actually using it"
Thomas Gleixner [Thu, 21 Sep 2017 09:54:44 +0000 (11:54 +0200)]
Revert "genirq: Restrict effective affinity to interrupts actually using it"

This reverts commit 74def747bcd09692bdbf8c6a15350795b0f11ca8.

The change to the helper function is only correct for the /proc/irq/
readout usage, but breaks the existing x86 usage of that function.

Reported-by: Yanko Kaneti <yaneti@declera.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Marc Zyngier <marc.zyngier@arm.com>
7 years agopowerpc/pseries: Fix parent_dn reference leak in add_dt_node()
Tyrel Datwyler [Wed, 20 Sep 2017 21:02:52 +0000 (17:02 -0400)]
powerpc/pseries: Fix parent_dn reference leak in add_dt_node()

A reference to the parent device node is held by add_dt_node() for the
node to be added. If the call to dlpar_configure_connector() fails
add_dt_node() returns ENOENT and that reference is not freed.

Add a call to of_node_put(parent_dn) prior to bailing out after a
failed dlpar_configure_connector() call.

Fixes: 8d5ff320766f ("powerpc/pseries: Make dlpar_configure_connector parent node aware")
Cc: stable@vger.kernel.org # v3.12+
Signed-off-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
7 years agopowerpc/pseries: Fix "OF: ERROR: Bad of_node_put() on /cpus" during DLPAR
Tyrel Datwyler [Wed, 20 Sep 2017 21:02:51 +0000 (17:02 -0400)]
powerpc/pseries: Fix "OF: ERROR: Bad of_node_put() on /cpus" during DLPAR

Commit 215ee763f8cb ("powerpc: pseries: remove dlpar_attach_node
dependency on full path") reworked dlpar_attach_node() to no longer
look up the parent node "/cpus", but instead to have the parent node
passed by the caller in the function parameter list.

As a result dlpar_attach_node() is no longer responsible for freeing
the reference to the parent node. However, commit 215ee763f8cb failed
to remove the of_node_put(parent) call in dlpar_attach_node(), or to
take into account that the reference to the parent in the caller
dlpar_cpu_add() needs to be held until after dlpar_attach_node()
returns.

As a result doing repeated cpu add/remove dlpar operations will
eventually result in the following error:

  OF: ERROR: Bad of_node_put() on /cpus
  CPU: 0 PID: 10896 Comm: drmgr Not tainted 4.13.0-autotest #1
  Call Trace:
   dump_stack+0x15c/0x1f8 (unreliable)
   of_node_release+0x1a4/0x1c0
   kobject_put+0x1a8/0x310
   kobject_del+0xbc/0xf0
   __of_detach_node_sysfs+0x144/0x210
   of_detach_node+0xf0/0x180
   dlpar_detach_node+0xc4/0x120
   dlpar_cpu_remove+0x280/0x560
   dlpar_cpu_release+0xbc/0x1b0
   arch_cpu_release+0x6c/0xb0
   cpu_release_store+0xa0/0x100
   dev_attr_store+0x68/0xa0
   sysfs_kf_write+0xa8/0xf0
   kernfs_fop_write+0x2cc/0x400
   __vfs_write+0x5c/0x340
   vfs_write+0x1a8/0x3d0
   SyS_write+0xa8/0x1a0
   system_call+0x58/0x6c

Fix the issue by removing the of_node_put(parent) call from
dlpar_attach_node(), and ensuring that the reference to the parent
node is properly held and released by the caller dlpar_cpu_add().

Fixes: 215ee763f8cb ("powerpc: pseries: remove dlpar_attach_node dependency on full path")
Signed-off-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>
Reported-by: Abdul Haleem <abdhalee@linux.vnet.ibm.com>
[mpe: Add a comment in the code and frob the change log slightly]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
7 years agopowerpc/eeh: Create PHB PEs after EEH is initialized
Benjamin Herrenschmidt [Thu, 7 Sep 2017 06:35:40 +0000 (16:35 +1000)]
powerpc/eeh: Create PHB PEs after EEH is initialized

Otherwise we end up not yet having computed the right diag data size
on powernv where EEH initialization is delayed, thus causing memory
corruption later on when calling OPAL.

Fixes: 5cb1f8fdddb7 ("powerpc/powernv/pci: Dynamically allocate PHB diag data")
Cc: stable@vger.kernel.org # v4.13+
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Russell Currey <ruscur@russell.cc>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
7 years agoipc/shm: Fix order of parameters when calling copy_compat_shmid_to_user
Will Deacon [Mon, 18 Sep 2017 16:47:38 +0000 (17:47 +0100)]
ipc/shm: Fix order of parameters when calling copy_compat_shmid_to_user

Commit 553f770ef71b ("ipc: move compat shmctl to native") moved the
compat IPC syscall handling into ipc/shm.c and refactored the struct
accessors in the process. Unfortunately, the call to
copy_compat_shmid_to_user when handling a compat {IPC,SHM}_STAT command
gets the arguments the wrong way round, passing a kernel stack address
as the user buffer (destination) and the user buffer as the kernel stack
address (source).

This patch fixes the parameter ordering so the buffers are accessed
correctly.

Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
7 years agoiov_iter: fix page_copy_sane for compound pages
Petar Penkov [Tue, 29 Aug 2017 18:20:32 +0000 (11:20 -0700)]
iov_iter: fix page_copy_sane for compound pages

Issue is that if the data crosses a page boundary inside a compound
page, this check will incorrectly trigger a WARN_ON.

To fix this, compute the order using the head of the compound page and
adjust the offset to be relative to that head.

Fixes: 72e809ed81ed ("iov_iter: sanity checks for copy to/from page
primitives")

Signed-off-by: Petar Penkov <ppenkov@google.com>
CC: Al Viro <viro@zeniv.linux.org.uk>
CC: Eric Dumazet <edumazet@google.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
7 years agoSMB: Validate negotiate (to protect against downgrade) even if signing off
Steve French [Thu, 21 Sep 2017 00:57:18 +0000 (19:57 -0500)]
SMB: Validate negotiate (to protect against downgrade) even if signing off

As long as signing is supported (ie not a guest user connection) and
connection is SMB3 or SMB3.02, then validate negotiate (protect
against man in the middle downgrade attacks).  We had been doing this
only when signing was required, not when signing was just enabled,
but this more closely matches recommended SMB3 behavior and is
better security.  Suggested by Metze.

Signed-off-by: Steve French <smfrench@gmail.com>
Reviewed-by: Jeremy Allison <jra@samba.org>
Acked-by: Stefan Metzmacher <metze@samba.org>
Reviewed-by: Ronnie Sahlberg <lsahlber@redhat.com>
CC: Stable <stable@vger.kernel.org>
7 years agocifs: release auth_key.response for reconnect.
Shu Wang [Fri, 8 Sep 2017 10:48:33 +0000 (18:48 +0800)]
cifs: release auth_key.response for reconnect.

There is a race that cause cifs reconnect in cifs_mount,
- cifs_mount
  - cifs_get_tcp_session
    - [ start thread cifs_demultiplex_thread
      - cifs_read_from_socket: -ECONNABORTED
        - DELAY_WORK smb2_reconnect_server ]
  - cifs_setup_session
  - [ smb2_reconnect_server ]

auth_key.response was allocated in cifs_setup_session, and
will release when the session destoried. So when session re-
connect, auth_key.response should be check and released.

Tested with my system:
CIFS VFS: Free previous auth_key.response = ffff8800320bbf80

A simple auth_key.response allocation call trace:
- cifs_setup_session
- SMB2_sess_setup
- SMB2_sess_auth_rawntlmssp_authenticate
- build_ntlmssp_auth_blob
- setup_ntlmv2_rsp

Signed-off-by: Shu Wang <shuwang@redhat.com>
Signed-off-by: Steve French <smfrench@gmail.com>
CC: Stable <stable@vger.kernel.org>
Reviewed-by: Ronnie Sahlberg <lsahlber@redhat.com>
7 years agocifs: release cifs root_cred after exit_cifs
Shu Wang [Thu, 7 Sep 2017 08:03:27 +0000 (16:03 +0800)]
cifs: release cifs root_cred after exit_cifs

memory leak was found by kmemleak. exit_cifs_spnego
should be called before cifs module removed, or
cifs root_cred will not be released.

kmemleak report:
unreferenced object 0xffff880070a3ce40 (size 192):
  backtrace:
     kmemleak_alloc+0x4a/0xa0
     kmem_cache_alloc+0xc7/0x1d0
     prepare_kernel_cred+0x20/0x120
     init_cifs_spnego+0x2d/0x170 [cifs]
     0xffffffffc07801f3
     do_one_initcall+0x51/0x1b0
     do_init_module+0x60/0x1fd
     load_module+0x161e/0x1b60
     SYSC_finit_module+0xa9/0x100
     SyS_finit_module+0xe/0x10

Signed-off-by: Shu Wang <shuwang@redhat.com>
Signed-off-by: Steve French <smfrench@gmail.com>
Reviewed-by: Ronnie Sahlberg <lsahlber@redhat.com>
CC: Stable <stable@vger.kernel.org>
7 years agoCIFS: make arrays static const, reduces object code size
Colin Ian King [Tue, 19 Sep 2017 15:27:39 +0000 (16:27 +0100)]
CIFS: make arrays static const, reduces object code size

Don't populate the read-only arrays types[] on the stack, instead make
them both static const.  Makes the object code smaller by over 200 bytes:

Before:
   text    data     bss     dec     hex filename
 111503   37696     448  149647   2488f fs/cifs/file.o

After:
   text    data     bss     dec     hex filename
 111140   37856     448  149444   247c4 fs/cifs/file.o

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Steve French <smfrench@gmail.com>
Reviewed-by: Ronnie Sahlberg <lsahlber@redhat.com>
7 years agoMerge branch 'hns3-tm-fixes'
David S. Miller [Wed, 20 Sep 2017 23:15:40 +0000 (16:15 -0700)]
Merge branch 'hns3-tm-fixes'

Yunsheng Lin says:

====================
TM related bugfixes for the HNS3 Ethernet Driver

This patch set contains a few bugfixes related to hclge_tm module.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: hns3: Fix for pri to tc mapping in TM
Yunsheng Lin [Wed, 20 Sep 2017 10:52:58 +0000 (18:52 +0800)]
net: hns3: Fix for pri to tc mapping in TM

Current mapping between pri and tc is one to one,
so user can't map multi priorities to the same tc.
This patch changes the mapping to many to one.

Fixes: 848440544b41f ("net: hns3: Add support of TX Scheduler & Shaper to HNS3 driver")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: hns3: Fix for setting rss_size incorrectly
Yunsheng Lin [Wed, 20 Sep 2017 10:52:57 +0000 (18:52 +0800)]
net: hns3: Fix for setting rss_size incorrectly

rss_size is 1, 2, 4, 8, 16, 32, 64, 128, but acutal tc queue
size can be any u16 less than 128. If tc queue size is 5, we
set the rss_size to 8, indirection table will be used to limit
the size of actual queue size.
It may cause dropping of receiving packet in hardware if
rss_size is not set correctly.
For now, each TC has the same rss size.

Fixes: 46a3df9f9718 ("net: hns3: Add HNS3 Acceleration Engine & Compatibility Layer Support")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: hns3: Fix typo error for feild in hclge_tm
Yunsheng Lin [Wed, 20 Sep 2017 10:52:56 +0000 (18:52 +0800)]
net: hns3: Fix typo error for feild in hclge_tm

This patch fixes a typo error for feild, which should be field.

Fixes: 848440544b41f ("net: hns3: Add support of TX Scheduler & Shaper to HNS3 driver")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: hns3: Fix for rx priv buf allocation when DCB is not supported
Yunsheng Lin [Wed, 20 Sep 2017 10:52:55 +0000 (18:52 +0800)]
net: hns3: Fix for rx priv buf allocation when DCB is not supported

When hdev doesn't support DCB, rx private buffer is not allocated,
otherwise there is not enough buffer for rx shared buffer, causing
buffer allocation process to fail.
This patch fixes by checking the dcb capability in
hclge_rx_buffer_calc.

Fixes: 46a3df9f9718 ("net: hns3: Add HNS3 Acceleration Engine & Compatibility Layer Support")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: hns3: Fix for rx_priv_buf_alloc not setting rx shared buffer
Yunsheng Lin [Wed, 20 Sep 2017 10:52:54 +0000 (18:52 +0800)]
net: hns3: Fix for rx_priv_buf_alloc not setting rx shared buffer

rx_priv_buf_alloc is used to tell hardware how much buffer is
used for rx direction, right now only the private buffer is
assigned.
For ae_dev that doesn't support DCB, private rx buffer is assigned
to zero, only shared rx buffer is used. So not setting the shared
rx buffer cause dropping of packet in SSU.

Fixes: 46a3df9f9718 ("net: hns3: Add HNS3 Acceleration Engine & Compatibility Layer Support")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: hns3: Fix for not setting rx private buffer size to zero
Yunsheng Lin [Wed, 20 Sep 2017 10:52:53 +0000 (18:52 +0800)]
net: hns3: Fix for not setting rx private buffer size to zero

When rx private buffer is disabled, there may be some case that
the rx private buffer is not set to zero, which may cause buffer
allocation process to fail.
This patch fixes this problem by setting priv->enable to 0 and
priv->buf_size to zero when rx private buffer is disabled.

Fixes: 46a3df9f9718 ("net: hns3: Add HNS3 Acceleration Engine & Compatibility Layer Support")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: hns3: Fix for DEFAULT_DV when dev doesn't support DCB
Yunsheng Lin [Wed, 20 Sep 2017 10:52:52 +0000 (18:52 +0800)]
net: hns3: Fix for DEFAULT_DV when dev doesn't support DCB

When ae_dev doesn't support DCB, DEFAULT_DV must be set to
a lower value, otherwise the buffer allocation process will
fail.
This patch fix it by setting it to 30K bytes.

Fixes: 46a3df9f9718 ("net: hns3: Add HNS3 Acceleration Engine & Compatibility Layer Support")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: hns3: Fix initialization when cmd is not supported
Yunsheng Lin [Wed, 20 Sep 2017 10:52:51 +0000 (18:52 +0800)]
net: hns3: Fix initialization when cmd is not supported

When ae_dev doesn't support DCB, rx_priv_wl_config,
common_thrd_config and tm_qs_bp_cfg can't be called, otherwise
cmd return fail, which causes the hclge module initialization
process to fail.
This patch fix it by adding a DCB capability flag to check if
the ae_dev support DCB.

Fixes: 46a3df9f9718 ("net: hns3: Add HNS3 Acceleration Engine & Compatibility Layer Support")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: hns3: Cleanup for ROCE capability flag in ae_dev
Yunsheng Lin [Wed, 20 Sep 2017 10:52:50 +0000 (18:52 +0800)]
net: hns3: Cleanup for ROCE capability flag in ae_dev

This patch add the ROCE supported flag in the driver_data
field of pci_device_id, delete roce_pci_tbl and change
HNAE_DEV_SUPPORT_ROCE_B to HNAE3_DEV_SUPPORT_ROCE_B.
This cleanup is done in order to support adding capability
in pci_device_id and to fix initialization failure when
cmd is not supported.

Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf
David S. Miller [Wed, 20 Sep 2017 23:08:23 +0000 (16:08 -0700)]
Merge git://git./pub/scm/linux/kernel/git/pablo/nf

Pablo Neira Ayuso says:

====================
Netfilter fixes for net

The following patchset contains two Netfilter fixes for your net tree,
they are:

1) Fix NAt compilation with UP, from Geert Uytterhoeven.

2) Fix incorrect number of entries when dumping a set, from
   Vishwanath Pai.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoisdn/i4l: fetch the ppp_write buffer in one shot
Meng Xu [Wed, 20 Sep 2017 01:49:55 +0000 (21:49 -0400)]
isdn/i4l: fetch the ppp_write buffer in one shot

In isdn_ppp_write(), the header (i.e., protobuf) of the buffer is
fetched twice from userspace. The first fetch is used to peek at the
protocol of the message and reset the huptimer if necessary; while the
second fetch copies in the whole buffer. However, given that buf resides
in userspace memory, a user process can race to change its memory content
across fetches. By doing so, we can either avoid resetting the huptimer
for any type of packets (by first setting proto to PPP_LCP and later
change to the actual type) or force resetting the huptimer for LCP
packets.

This patch changes this double-fetch behavior into two single fetches
decided by condition (lp->isdn_device < 0 || lp->isdn_channel <0).
A more detailed discussion can be found at
https://marc.info/?l=linux-kernel&m=150586376926123&w=2

Signed-off-by: Meng Xu <mengxu.gatech@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: fec: return IRQ_HANDLED if fec_ptp_check_pps_event handled it
Troy Kisky [Wed, 20 Sep 2017 00:33:09 +0000 (17:33 -0700)]
net: fec: return IRQ_HANDLED if fec_ptp_check_pps_event handled it

fec_ptp_check_pps_event will return 1 if FEC_T_TF_MASK caused
an interrupt. Don't return IRQ_NONE in this case.

Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
Acked-by: Fugang Duan <fugang.duan@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: fec: remove unused interrupt FEC_ENET_TS_TIMER
Troy Kisky [Wed, 20 Sep 2017 00:33:08 +0000 (17:33 -0700)]
net: fec: remove unused interrupt FEC_ENET_TS_TIMER

FEC_ENET_TS_TIMER is not checked in the interrupt routine
so there is no need to enable it.

Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
Acked-by: Fugang Duan <fugang.duan@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: fec: only check queue 0 if RXF_0/TXF_0 interrupt is set
Troy Kisky [Wed, 20 Sep 2017 00:33:07 +0000 (17:33 -0700)]
net: fec: only check queue 0 if RXF_0/TXF_0 interrupt is set

Before queue 0 was always checked if any queue caused an interrupt.
It is better to just mark queue 0 if queue 0 has caused an interrupt.

Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
Acked-by: Fugang Duan <Fugang.duan@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: change skb->mac_header when Generic XDP calls adjust_head
Edward Cree [Tue, 19 Sep 2017 17:45:56 +0000 (18:45 +0100)]
net: change skb->mac_header when Generic XDP calls adjust_head

Since XDP's view of the packet includes the MAC header, moving the start-
 of-packet with bpf_xdp_adjust_head needs to also update the offset of the
 MAC header (which is relative to skb->head, not to the skb->data that was
 changed).
Without this, tcpdump sees packets starting from the old MAC header rather
 than the new one, at least in my tests on the loopback device.

Fixes: b5cdae3291f7 ("net: Generic XDP")
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: compat: assert the size of cmsg copied in is as expected
Meng Xu [Tue, 19 Sep 2017 17:19:13 +0000 (13:19 -0400)]
net: compat: assert the size of cmsg copied in is as expected

The actual length of cmsg fetched in during the second loop
(i.e., kcmsg - kcmsg_base) could be different from what we
get from the first loop (i.e., kcmlen).

The main reason is that the two get_user() calls in the two
loops (i.e., get_user(ucmlen, &ucmsg->cmsg_len) and
__get_user(ucmlen, &ucmsg->cmsg_len)) could cause ucmlen
to have different values even they fetch from the same userspace
address, as user can race to change the memory content in
&ucmsg->cmsg_len across fetches.

Although in the second loop, the sanity check
if ((char *)kcmsg_base + kcmlen - (char *)kcmsg < CMSG_ALIGN(tmp))
is inplace, it only ensures that the cmsg fetched in during the
second loop does not exceed the length of kcmlen, but not
necessarily equal to kcmlen. But indicated by the assignment
kmsg->msg_controllen = kcmlen, we should enforce that.

This patch adds this additional sanity check and ensures that
what is recorded in kmsg->msg_controllen is the actual cmsg length.

Signed-off-by: Meng Xu <mengxu.gatech@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoMerge tag 'exynos-drm-fixes-for-v4.14-rc1' of git://git.kernel.org/pub/scm/linux...
Dave Airlie [Wed, 20 Sep 2017 22:02:39 +0000 (08:02 +1000)]
Merge tag 'exynos-drm-fixes-for-v4.14-rc1' of git://git./linux/kernel/git/daeinki/drm-exynos into drm-fixes

- fix suspend/resume issues.
- fix memory corruption detected by kasan.
- fix build error on x86.

* tag 'exynos-drm-fixes-for-v4.14-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/daeinki/drm-exynos:
  drm/exynos/hdmi: Fix unsafe list iteration
  drm: exynos: include linux/irq.h
  drm/exynos: Fix suspend/resume support
  drm/exynos: Fix locking in the suspend/resume paths

7 years agoMerge tag 'drm-intel-fixes-2017-09-20' of git://anongit.freedesktop.org/git/drm-intel...
Dave Airlie [Wed, 20 Sep 2017 22:02:21 +0000 (08:02 +1000)]
Merge tag 'drm-intel-fixes-2017-09-20' of git://anongit.freedesktop.org/git/drm-intel into drm-fixes

drm/i915 fixes for 4.14-rc1

Couple fixes for stable:

- Fix MIPI panels on BXT.
- Fix PCI BARs information on GVT.

Plus other fixes:

- Fix minimal brightness for BXT, GLK, CFL and CNL.
- Fix compilation warning: unused in_vbl
- Fix error handling in intel_framebuffer_init

* tag 'drm-intel-fixes-2017-09-20' of git://anongit.freedesktop.org/git/drm-intel:
  drm/i915: Remove unused 'in_vbl' from i915_get_crtc_scanoutpos()
  drm/i915/cnp: set min brightness from VBT
  Revert "drm/i915/bxt: Disable device ready before shutdown command"
  drm/i915/bxt: set min brightness from VBT
  drm/i915: Fix an error handling in 'intel_framebuffer_init()'
  drm/i915/gvt: Fix incorrect PCI BARs reporting

7 years ago[SMB3] Update session and share information displayed for debugging SMB2/SMB3
Steve French [Wed, 20 Sep 2017 21:46:49 +0000 (16:46 -0500)]
[SMB3] Update session and share information displayed for debugging SMB2/SMB3

We were not displaying some key fields (session status and capabilities and
whether guest authenticated) for SMB2/SMB3 session in /proc/fs/cifs/DebugData.

This is needed for real world triage of problems with the (now much more
common) SMB3 mounts.

Signed-off-by: Steve French <smfrench@gmail.com>
7 years agobpf: one perf event close won't free bpf program attached by another perf event
Yonghong Song [Mon, 18 Sep 2017 23:38:36 +0000 (16:38 -0700)]
bpf: one perf event close won't free bpf program attached by another perf event

This patch fixes a bug exhibited by the following scenario:
  1. fd1 = perf_event_open with attr.config = ID1
  2. attach bpf program prog1 to fd1
  3. fd2 = perf_event_open with attr.config = ID1
     <this will be successful>
  4. user program closes fd2 and prog1 is detached from the tracepoint.
  5. user program with fd1 does not work properly as tracepoint
     no output any more.

The issue happens at step 4. Multiple perf_event_open can be called
successfully, but only one bpf prog pointer in the tp_event. In the
current logic, any fd release for the same tp_event will free
the tp_event->prog.

The fix is to free tp_event->prog only when the closing fd
corresponds to the one which registered the program.

Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agopacket: hold bind lock when rebinding to fanout hook
Willem de Bruijn [Thu, 14 Sep 2017 21:14:41 +0000 (17:14 -0400)]
packet: hold bind lock when rebinding to fanout hook

Packet socket bind operations must hold the po->bind_lock. This keeps
po->running consistent with whether the socket is actually on a ptype
list to receive packets.

fanout_add unbinds a socket and its packet_rcv/tpacket_rcv call, then
binds the fanout object to receive through packet_rcv_fanout.

Make it hold the po->bind_lock when testing po->running and rebinding.
Else, it can race with other rebind operations, such as that in
packet_set_ring from packet_rcv to tpacket_rcv. Concurrent updates
can result in a socket being added to a fanout group twice, causing
use-after-free KASAN bug reports, among others.

Reported independently by both trinity and syzkaller.
Verified that the syzkaller reproducer passes after this patch.

Fixes: dc99f600698d ("packet: Add fanout support.")
Reported-by: nixioaming <nixiaoming@huawei.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoPCI: endpoint: Use correct "end of test" interrupt
John Keeping [Wed, 20 Sep 2017 18:56:06 +0000 (13:56 -0500)]
PCI: endpoint: Use correct "end of test" interrupt

pci_epf_test_raise_irq() reads the interrupt to use for the response from
reg->command, but this has been cleared at the beginning of the command
handler so the value is always zero at this point.

Instead, extract the interrupt index before handling the command and then
pass the requested interrupt into pci_epf_test_raise_irq().  This allows us
to remove the specific code to extract the interrupt for
COMMAND_RAISE_MSI_IRQ since it is now handled in common code.

Fixes: 3ecf3232c54c ("PCI: endpoint: Do not reset *command* inadvertently")
Signed-off-by: John Keeping <john@metanate.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Kishon Vijay Abraham I <kishon@ti.com>
7 years agoMerge tag 'trace-v4.14-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt...
Linus Torvalds [Wed, 20 Sep 2017 16:38:07 +0000 (06:38 -1000)]
Merge tag 'trace-v4.14-rc1' of git://git./linux/kernel/git/rostedt/linux-trace

Pull tracing fixes from Steven Rostedt:
 "This includes three minor fixes.

    - Have writing to trace file clear the irqsoff (and friends) tracer

    - trace_pipe behavior for instance buffers was different than top
      buffer

    - Show a message of why mmiotrace doesn't start from commandline"

* tag 'trace-v4.14-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  tracing: Fix trace_pipe behavior for instance traces
  tracing: Ignore mmiotrace from kernel commandline
  tracing: Erase irqsoff trace with empty write

7 years agoscripts/dtc: dtx_diff - 2nd update of include dts paths to match build
Frank Rowand [Tue, 19 Sep 2017 00:18:44 +0000 (17:18 -0700)]
scripts/dtc: dtx_diff - 2nd update of include dts paths to match build

Update dtx_diff include paths in the same manner as:
commit b12869a8d519 ("of: remove drivers/of/testcase-data from
include search path for CPP"), commit 5ffa2aed389c ("of: remove
arch/$(SRCARCH)/boot/dts from include search path for CPP"), and
commit 50f9ddaf64e1 ("of: search scripts/dtc/include-prefixes path
for both CPP and DTC").

Remove proposed include path kernel/dts/, which was never implemented
for the dtb build.

For the diff case, each source file is compiled separately.  For
each of those compiles, provide the location of the source file
as an include path, not the location of both source files.

Signed-off-by: Frank Rowand <frank.rowand@sony.com>
Signed-off-by: Rob Herring <robh@kernel.org>
7 years agokbuild: rpm-pkg: fix version number handling
Masahiro Yamada [Wed, 20 Sep 2017 13:01:26 +0000 (22:01 +0900)]
kbuild: rpm-pkg: fix version number handling

The "Release:" field of the spec file is determined based on the
.version file.

However, the .version file is not copied to the source tar file.
So, when we build the kernel from the source package, the UTS_VERSION
always indicates #1.  This does not match with "rpm -q".

The kernel UTS_VERSION and "rpm -q" do not agree for binrpm-pkg, either.
Please note the kernel has already been built before the spec file is
created.  Currently, mkspec invokes mkversion.  This script returns an
incremented version.  So, the "Release:" field of the spec file is
greater than the version in the kernel by one.

For the source package build (where .version file is missing), we can
give KBUILD_BUILD_VERSION=%{release} to the build command.

For the binary package build, we can simply read out the .version file
because it contains the version number that was used for building the
kernel image.

We can remove scripts/mkversion because scripts/package/Makefile need
not touch the .version file.

Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
7 years agokbuild: deb-pkg: remove firmware package support
Masahiro Yamada [Mon, 18 Sep 2017 09:22:19 +0000 (18:22 +0900)]
kbuild: deb-pkg: remove firmware package support

Commit 5620a0d1aacd ("firmware: delete in-kernel firmware") deleted
in-kernel firmware support, including the firmware install command.

So, the firmware package does not make sense any more.  Remove it.

Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Reviewed-by: Riku Voipio <riku.voipio@linaro.org>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agokbuild: rpm-pkg: delete firmware_install to fix build error
Masahiro Yamada [Mon, 18 Sep 2017 09:12:36 +0000 (18:12 +0900)]
kbuild: rpm-pkg: delete firmware_install to fix build error

Commit 5620a0d1aacd ("firmware: delete in-kernel firmware") deleted
in-kernel firmware support, including "make firmware_install".

Since then, "make rpm-pkg" / "make binrpm-pkg" fails to build with
the error:

  make[2]: *** No rule to make target `firmware_install'.  Stop.

Commit df85b2d767aa ("firmware: Restore support for built-in firmware")
restored the build infrastructure for CONFIG_EXTRA_FIRMWARE, but this
is out of the scope of "make firmware_install".  So, the right thing to
do is to kill the use of "make firmware_install".

Fixes: 5620a0d1aacd ("firmware: delete in-kernel firmware")
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agos390/topology: enable / disable topology dynamically
Heiko Carstens [Thu, 14 Sep 2017 12:42:32 +0000 (14:42 +0200)]
s390/topology: enable / disable topology dynamically

Add a new sysctl file /proc/sys/s390/topology which displays if
topology is on (1) or off (0) as specified by the "topology=" kernel
parameter.

This allows to change topology information during runtime and
configuring it via /etc/sysctl.conf instead of using the kernel line
parameter.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
7 years agos390/topology: alternative topology for topology-less machines
Heiko Carstens [Tue, 19 Sep 2017 10:52:22 +0000 (12:52 +0200)]
s390/topology: alternative topology for topology-less machines

If running on machines that do not provide topology information we
currently generate a "fake" topology which defines the maximum
distance between each cpu: each cpu will be put into an own drawer.

Historically this used to be the best option for (virtual) machines in
overcommited hypervisors.

For some workloads however it is better to generate a different
topology where all cpus are siblings within a package (all cpus are
core siblings). This shows performance improvements of up to 10%,
depending on the workload.

In order to keep the current behaviour, but also allow to switch to
the different core sibling topology use the existing "topology="
kernel parameter:

Specifying "topology=on" on machines without topology information will
generate the core siblings (fake) topology information, instead of the
default topology information where all cpus have the maximum distance.

On machines which provide topology information specifying
"topology=on" does not have any effect.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
7 years agopowerpc/kprobes: Update optprobes to use emulate_update_regs()
Naveen N. Rao [Fri, 15 Sep 2017 10:08:21 +0000 (15:38 +0530)]
powerpc/kprobes: Update optprobes to use emulate_update_regs()

Optprobes depended on an updated regs->nip from analyse_instr() to
identify the location to branch back from the optprobes trampoline.
However, since commit 3cdfcbfd32b9d ("powerpc: Change analyse_instr so
it doesn't modify *regs"), analyse_instr() doesn't update the registers
anymore.  Due to this, we end up branching back from the optprobes
trampoline to the same branch into the trampoline resulting in a loop.

Fix this by calling out to emulate_update_regs() before using the nip.
Additionally, explicitly compare the return value from analyse_instr()
to 1, rather than just checking for !0 so as to guard against any
future changes to analyse_instr() that may result in -1 being returned
in more scenarios.

Fixes: 3cdfcbfd32b9d ("powerpc: Change analyse_instr so it doesn't modify *regs")
Signed-off-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
7 years agoMerge branch 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/scottwood/linux...
Michael Ellerman [Wed, 20 Sep 2017 10:05:24 +0000 (20:05 +1000)]
Merge branch 'next' of git://git./linux/kernel/git/scottwood/linux into fixes

Merge one commit from Scott which I missed while away.

7 years agocrypto: af_alg - update correct dst SGL entry
Stephan Mueller [Wed, 30 Aug 2017 07:17:39 +0000 (09:17 +0200)]
crypto: af_alg - update correct dst SGL entry

When two adjacent TX SGL are processed and parts of both TX SGLs
are pulled into the per-request TX SGL, the wrong per-request
TX SGL entries were updated.

This fixes a NULL pointer dereference when a cipher implementation walks
the TX SGL where some of the SGL entries were NULL.

Fixes: e870456d8e7c ("crypto: algif_skcipher - overhaul memory...")
Signed-off-by: Stephan Mueller <smueller@chronox.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - fix LS1021A support on ARMv7 multiplatform kernel
Horia Geantă [Fri, 1 Sep 2017 14:12:59 +0000 (17:12 +0300)]
crypto: caam - fix LS1021A support on ARMv7 multiplatform kernel

When built using multi_v7_defconfig, driver does not work on LS1021A:
[...]
caam 1700000.crypto: can't identify CAAM ipg clk: -2
caam: probe of 1700000.crypto failed with error -2
[...]

It turns out we have to detect at runtime whether driver is running
on an i.MX platform or not.

Cc: <stable@vger.kernel.org>
Fixes: 6c3af9559352 ("crypto: caam - add support for LS1021A")
Signed-off-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: inside-secure - fix gcc-4.9 warnings
Arnd Bergmann [Tue, 12 Sep 2017 10:12:16 +0000 (12:12 +0200)]
crypto: inside-secure - fix gcc-4.9 warnings

All older compiler versions up to gcc-4.9 produce these
harmless warnings:

drivers/crypto/inside-secure/safexcel_cipher.c:389:9: warning: missing braces around initializer [-Wmissing-braces]
drivers/crypto/inside-secure/safexcel_cipher.c:389:9: warning: (near initialization for ‘result.completion’) [-Wmissing-braces]
drivers/crypto/inside-secure/safexcel_hash.c:422:9: warning: missing braces around initializer [-Wmissing-braces]
drivers/crypto/inside-secure/safexcel_hash.c:422:9: warning: (near initialization for ‘result.completion’) [-Wmissing-braces]

This changes the syntax to something that works on all versions
without warnings.

Fixes: 1b44c5a60c13 ("crypto: inside-secure - add SafeXcel EIP197 crypto engine driver")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Antoine Tenart <antoine.tenart@free-electrons.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: talitos - Don't provide setkey for non hmac hashing algs.
LEROY Christophe [Tue, 12 Sep 2017 09:03:39 +0000 (11:03 +0200)]
crypto: talitos - Don't provide setkey for non hmac hashing algs.

Today, md5sum fails with error -ENOKEY because a setkey
function is set for non hmac hashing algs, see strace output below:

mmap(NULL, 378880, PROT_READ, MAP_SHARED, 6, 0) = 0x77f50000
accept(3, 0, NULL)                      = 7
vmsplice(5, [{"bin/\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 378880}], 1, SPLICE_F_MORE|SPLICE_F_GIFT) = 262144
splice(4, NULL, 7, NULL, 262144, SPLICE_F_MORE) = -1 ENOKEY (Required key not available)
write(2, "Generation of hash for file kcap"..., 50) = 50
munmap(0x77f50000, 378880)              = 0

This patch ensures that setkey() function is set only
for hmac hashing.

Cc: <stable@vger.kernel.org>
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: talitos - fix hashing
LEROY Christophe [Wed, 13 Sep 2017 10:44:57 +0000 (12:44 +0200)]
crypto: talitos - fix hashing

md5sum on some files gives wrong result

Exemple:

With the md5sum from libkcapi:
c15115c05bad51113f81bdaee735dd09  test

With the original md5sum:
bbdf41d80ba7e8b2b7be3a0772be76cb  test

This patch fixes this issue

Cc: <stable@vger.kernel.org>
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: talitos - fix sha224
LEROY Christophe [Wed, 13 Sep 2017 10:44:51 +0000 (12:44 +0200)]
crypto: talitos - fix sha224

Kernel crypto tests report the following error at startup

[    2.752626] alg: hash: Test 4 failed for sha224-talitos
[    2.757907] 00000000: 30 e2 86 e2 e7 8a dd 0d d7 eb 9f d5 83 fe f1 b0
00000010: 2d 5a 6c a5 f9 55 ea fd 0e 72 05 22

This patch fixes it

Cc: <stable@vger.kernel.org>
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: x86/twofish - Fix RBP usage
Josh Poimboeuf [Mon, 18 Sep 2017 19:42:11 +0000 (14:42 -0500)]
crypto: x86/twofish - Fix RBP usage

Using RBP as a temporary register breaks frame pointer convention and
breaks stack traces when unwinding from an interrupt in the crypto code.

Use R13 instead of RBP.  Both are callee-saved registers, so the
substitution is straightforward.

Reported-by: Eric Biggers <ebiggers@google.com>
Reported-by: Peter Zijlstra <peterz@infradead.org>
Tested-by: Eric Biggers <ebiggers@google.com>
Acked-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: sha512-avx2 - Fix RBP usage
Josh Poimboeuf [Mon, 18 Sep 2017 19:42:10 +0000 (14:42 -0500)]
crypto: sha512-avx2 - Fix RBP usage

Using RBP as a temporary register breaks frame pointer convention and
breaks stack traces when unwinding from an interrupt in the crypto code.

Mix things up a little bit to get rid of the RBP usage, without hurting
performance too much.  Use RDI instead of RBP for the TBL pointer.  That
will clobber CTX, so spill CTX onto the stack and use R12 to read it in
the outer loop.  R12 is used as a non-persistent temporary variable
elsewhere, so it's safe to use.

Also remove the unused y4 variable.

Reported-by: Eric Biggers <ebiggers3@gmail.com>
Reported-by: Peter Zijlstra <peterz@infradead.org>
Tested-by: Eric Biggers <ebiggers@google.com>
Acked-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: x86/sha256-ssse3 - Fix RBP usage
Josh Poimboeuf [Mon, 18 Sep 2017 19:42:09 +0000 (14:42 -0500)]
crypto: x86/sha256-ssse3 - Fix RBP usage

Using RBP as a temporary register breaks frame pointer convention and
breaks stack traces when unwinding from an interrupt in the crypto code.

Swap the usages of R12 and RBP.  Use R12 for the TBL register, and use
RBP to store the pre-aligned stack pointer.

Reported-by: Eric Biggers <ebiggers@google.com>
Reported-by: Peter Zijlstra <peterz@infradead.org>
Tested-by: Eric Biggers <ebiggers@google.com>
Acked-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: x86/sha256-avx2 - Fix RBP usage
Josh Poimboeuf [Mon, 18 Sep 2017 19:42:08 +0000 (14:42 -0500)]
crypto: x86/sha256-avx2 - Fix RBP usage

Using RBP as a temporary register breaks frame pointer convention and
breaks stack traces when unwinding from an interrupt in the crypto code.

There's no need to use RBP as a temporary register for the TBL value,
because it always stores the same value: the address of the K256 table.
Instead just reference the address of K256 directly.

Reported-by: Eric Biggers <ebiggers@google.com>
Reported-by: Peter Zijlstra <peterz@infradead.org>
Tested-by: Eric Biggers <ebiggers@google.com>
Acked-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>