GitHub/moto-9609/android_kernel_motorola_exynos9610.git
7 years agoIB/core, opa_vnic, hfi1, mlx5: Properly free rdma_netdev
Niranjana Vishwanathapura [Fri, 30 Jun 2017 20:14:46 +0000 (13:14 -0700)]
IB/core, opa_vnic, hfi1, mlx5: Properly free rdma_netdev

IPOIB is calling free_rdma_netdev even though alloc_rdma_netdev has
returned -EOPNOTSUPP.
Move free_rdma_netdev from ib_device structure to rdma_netdev structure
thus ensuring proper cleanup function is called for the rdma net device.

Fix the following trace:

ib0: Failed to modify QP to ERROR state
BUG: unable to handle kernel paging request at 0000000000001d20
IP: hfi1_vnic_free_rn+0x26/0xb0 [hfi1]
Call Trace:
 ipoib_remove_one+0xbe/0x160 [ib_ipoib]
 ib_unregister_device+0xd0/0x170 [ib_core]
 rvt_unregister_device+0x29/0x90 [rdmavt]
 hfi1_unregister_ib_device+0x1a/0x100 [hfi1]
 remove_one+0x4b/0x220 [hfi1]
 pci_device_remove+0x39/0xc0
 device_release_driver_internal+0x141/0x200
 driver_detach+0x3f/0x80
 bus_remove_driver+0x55/0xd0
 driver_unregister+0x2c/0x50
 pci_unregister_driver+0x2a/0xa0
 hfi1_mod_cleanup+0x10/0xf65 [hfi1]
 SyS_delete_module+0x171/0x250
 do_syscall_64+0x67/0x150
 entry_SYSCALL64_slow_path+0x25/0x25

Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Niranjana Vishwanathapura <niranjana.vishwanathapura@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/uverbs: Check port number supplied by user verbs cmds
Boris Pismenny [Tue, 27 Jun 2017 12:09:13 +0000 (15:09 +0300)]
RDMA/uverbs: Check port number supplied by user verbs cmds

The ib_uverbs_create_ah() ind ib_uverbs_modify_qp() calls receive
the port number from user input as part of its attributes and assumes
it is valid. Down on the stack, that parameter is used to access kernel
data structures.  If the value is invalid, the kernel accesses memory
it should not.  To prevent this, verify the port number before using it.

BUG: KASAN: use-after-free in ib_uverbs_create_ah+0x6d5/0x7b0
Read of size 4 at addr ffff880018d67ab8 by task syz-executor/313

BUG: KASAN: slab-out-of-bounds in modify_qp.isra.4+0x19d0/0x1ef0
Read of size 4 at addr ffff88006c40ec58 by task syz-executor/819

Fixes: 67cdb40ca444 ("[IB] uverbs: Implement more commands")
Fixes: 189aba99e70 ("IB/uverbs: Extend modify_qp and support packet pacing")
Cc: <stable@vger.kernel.org> # v2.6.14+
Cc: <security@kernel.org>
Cc: Yevgeny Kliteynik <kliteyn@mellanox.com>
Cc: Tziporet Koren <tziporet@mellanox.com>
Cc: Alex Polak <alexpo@mellanox.com>
Signed-off-by: Boris Pismenny <borisp@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoIB/opa_vnic: Use spinlock instead of mutex for stats_lock
Vishwanathapura, Niranjana [Wed, 14 Jun 2017 19:34:42 +0000 (12:34 -0700)]
IB/opa_vnic: Use spinlock instead of mutex for stats_lock

Stats can be read from atomic context, hence make stats_lock as
a spinlock.

Fix the following trace with debug kernel.

BUG: sleeping function called from invalid context at
kernel/locking/mutex.c:238
in_atomic(): 1, irqs_disabled(): 0, pid: 6487, name: sadc
Call Trace:
 dump_stack+0x63/0x90
 ___might_sleep+0xda/0x130
 __might_sleep+0x4a/0x90
 mutex_lock+0x20/0x50
 opa_vnic_get_stats64+0x56/0x140 [opa_vnic]
 dev_get_stats+0x74/0x130
 dev_seq_printf_stats+0x37/0x120
 dev_seq_show+0x14/0x30
 seq_read+0x26d/0x3d0

Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Niranjana Vishwanathapura <niranjana.vishwanathapura@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoIB/opa_vnic: Use GFP_ATOMIC while sending trap
Vishwanathapura, Niranjana [Wed, 14 Jun 2017 19:34:41 +0000 (12:34 -0700)]
IB/opa_vnic: Use GFP_ATOMIC while sending trap

Pass GFP_ATOMIC flag to ib_create_send_mad() while sending trap as it
can be triggered from the atomic context.

Fix the following trace with debug kernel.

BUG: sleeping function called from invalid context at mm/slab.h:432
in_atomic(): 1, irqs_disabled(): 0, pid: 1771, name: NetworkManager
Call Trace:
 dump_stack+0x63/0x90
 ___might_sleep+0xda/0x130
 __might_sleep+0x4a/0x90
 __kmalloc+0x19e/0x220
 ? ib_create_send_mad+0xea/0x390 [ib_core]
 ib_create_send_mad+0xea/0x390 [ib_core]
 opa_vnic_vema_send_trap+0x17b/0x460 [opa_vnic]
 opa_vnic_vema_report_event+0x57/0x80 [opa_vnic]
 opa_vnic_mac_send_event+0xaa/0xf0 [opa_vnic]
 opa_vnic_set_rx_mode+0x17/0x30 [opa_vnic]
 __dev_set_rx_mode+0x52/0x90
 dev_set_rx_mode+0x26/0x40
 __dev_open+0xe8/0x140

Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Niranjana Vishwanathapura <niranjana.vishwanathapura@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agordma/cxgb4: Fix memory leaks during module exit
Raju Rangoju [Fri, 9 Jun 2017 16:47:49 +0000 (22:17 +0530)]
rdma/cxgb4: Fix memory leaks during module exit

Fix memory leaks of iw_cxgb4 module in the exit path

Signed-off-by: Raju Rangoju <rajur@chelsio.com>
Reviewed-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoIB/ipoib: Fix memory leak in create child syscall
Feras Daoud [Wed, 14 Jun 2017 06:59:09 +0000 (09:59 +0300)]
IB/ipoib: Fix memory leak in create child syscall

The flow of creating a new child goes through ipoib_vlan_add
which allocates a new interface and checks the rtnl_lock.

If the lock is taken, restart_syscall will be called to restart
the system call again. In this case we are not releasing the
already allocated interface, causing a leak.

Fixes: 9baa0b036410 ("IB/ipoib: Add rtnl_link_ops support")
Signed-off-by: Feras Daoud <ferasda@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoIB/ipoib: Fix access to un-initialized napi struct
Alex Vesker [Wed, 14 Jun 2017 06:59:08 +0000 (09:59 +0300)]
IB/ipoib: Fix access to un-initialized napi struct

There is no need to re-enable napi since we set the initialized
flag before calling ipoib_ib_dev_stop which will disable napi,
disabling napi twice is harmless in case it was already disabled.

One more reason for this fix is that when using IPoIB new device
driver napi is not added to priv, this can lead to kernel panic
when rn_ops ndo_open fails.

[ 289.755840] invalid opcode: 0000 [#1] SMP
[ 289.757111] task: ffff880036964440 ti: ffff880178ee8000 task.ti: ffff880178ee8000
[ 289.757111] RIP: 0010:[<ffffffffa05368d6>] [<ffffffffa05368d6>] napi_enable.part.24+0x4/0x6 [ib_ipoib]
[ 289.757111] RSP: 0018:ffff880178eeb6d8 EFLAGS: 00010246
[ 289.757111] RAX: 0000000000000000 RBX: ffff880177a80010 RCX: 000000007fffffff
[ 289.757111] RDX: ffffffff81d5f118 RSI: 0000000000000000 RDI: ffff880177a80010
[ 289.757111] RBP: ffff880178eeb6d8 R08: 0000000000000082 R09: 0000000000000283
[ 289.757111] R10: 0000000000000000 R11: 0000000000000000 R12: ffff880175a00000
[ 289.757111] R13: ffff880177a80080 R14: 0000000000000000 R15: 0000000000000001
[ 289.757111] FS: 00007fe2ee346880(0000) GS:ffff88017fc00000(0000) knlGS:0000000000000000
[ 289.757111] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 289.757111] CR2: 00007fffca979020 CR3: 00000001792e4000 CR4: 00000000000006f0
[ 289.757111] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 289.757111] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
[ 289.757111] Stack:
[ 289.796027] ffff880178eeb6f0 ffffffffa05251f5 ffff880177a80000 ffff880178eeb718
[ 289.796027] ffffffffa0528505 ffff880175a00000 ffff880177a80000 0000000000000000
[ 289.796027] ffff880178eeb748 ffffffffa051f0ab ffff880175a00000 ffffffffa0537d60
[ 289.796027] Call Trace:
[ 289.796027] [<ffffffffa05251f5>] napi_enable+0x25/0x30 [ib_ipoib]
[ 289.796027] [<ffffffffa0528505>] ipoib_ib_dev_open+0x175/0x190 [ib_ipoib]
[ 289.796027] [<ffffffffa051f0ab>] ipoib_open+0x4b/0x160 [ib_ipoib]
[ 289.796027] [<ffffffff814fe33f>] _dev_open+0xbf/0x130
[ 289.796027] [<ffffffff814fe62d>] __dev_change_flags+0x9d/0x170
[ 289.796027] [<ffffffff814fe729>] dev_change_flags+0x29/0x60
[ 289.796027] [<ffffffff8150caf7>] do_setlink+0x397/0xa40

Fixes: cd565b4b51e5 ('IB/IPoIB: Support acceleration options callbacks')
Signed-off-by: Alex Vesker <valex@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoIB/ipoib: Delete napi in device uninit default
Alex Vesker [Wed, 14 Jun 2017 06:59:07 +0000 (09:59 +0300)]
IB/ipoib: Delete napi in device uninit default

This patch mekas init_default and uninit_default symmetric
with a call to delete napi. Additionally, the uninit_default
gained delete napi call in case of init_default fails.

Fixes: 515ed4f3aab4 ('IB/IPoIB: Separate control and data related initializations')
Signed-off-by: Alex Vesker <valex@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoIB/ipoib: Limit call to free rdma_netdev for capable devices
Alex Vesker [Wed, 14 Jun 2017 06:59:06 +0000 (09:59 +0300)]
IB/ipoib: Limit call to free rdma_netdev for capable devices

Limit calls to free_rdma_netdev() for capable devices only.

Fixes: cd565b4b51e5 ('IB/IPoIB: Support acceleration options callbacks')
Signed-off-by: Alex Vesker <valex@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoIB/ipoib: Fix memory leaks for child interfaces priv
Alex Vesker [Wed, 14 Jun 2017 06:59:05 +0000 (09:59 +0300)]
IB/ipoib: Fix memory leaks for child interfaces priv

There is a need to free priv explicitly and not just to release
the device, child priv is freed explicitly on remove flow and this
patch also includes priv free on error flow in P_key creation
and also in add_port.

Fixes: cd565b4b51e5 ('IB/IPoIB: Support acceleration options callbacks')
Signed-off-by: Alex Vesker <valex@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agorxe: Fix a sleep-in-atomic bug in post_one_send
Jia-Ju Bai [Mon, 5 Jun 2017 12:23:40 +0000 (20:23 +0800)]
rxe: Fix a sleep-in-atomic bug in post_one_send

The driver may sleep under a spin lock, and the function call path is:
post_one_send (acquire the lock by spin_lock_irqsave)
  init_send_wqe
    copy_from_user --> may sleep

There is no flow that makes "qp->is_user" true, and copy_from_user may
cause bug when a non-user pointer is used. So the lines of copy_from_user
and check of "qp->is_user" are removed.

Signed-off-by: Jia-Ju Bai <baijiaju1990@163.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Acked-by: Moni Shoua <monis@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/qedr: Add 64KB PAGE_SIZE support to user-space queues
Ram Amrani [Mon, 5 Jun 2017 13:32:27 +0000 (16:32 +0300)]
RDMA/qedr: Add 64KB PAGE_SIZE support to user-space queues

Add 64KB PAGE_SIZE support to user-space CQ, SQ and RQ queues.
De-facto it means that code was added to translate 64KB
pages to smaller 4KB pages that the FW can handle. Otherwise,
the FW would wrap (or jump to the next page)  when reaching 4KB
while the user space library will continue on the same large page.

Note that MR code remains as is since the FW supports larger pages
for MRs.

Signed-off-by: Ram Amrani <Ram.Amrani@cavium.com>
Signed-off-by: Michal Kalderon <Michal.Kalderon@cavium.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/qedr: Initialize byte_len in WC of READ and SEND commands
Michal Kalderon [Mon, 5 Jun 2017 13:32:26 +0000 (16:32 +0300)]
RDMA/qedr: Initialize byte_len in WC of READ and SEND commands

Initialize byte_len in work completion of RDMA_READ and RDMA_SEND.
Exposed by uDAPL application.

Signed-off-by: Michal Kalderon <Michal.Kalderon@cavium.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/bnxt_re: Remove FMR support
Selvin Xavier [Mon, 22 May 2017 10:15:44 +0000 (03:15 -0700)]
RDMA/bnxt_re: Remove FMR support

Some issues observed with FMR implementation
while running stress traffic. So removing the
FMR verbs support for now.

Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/bnxt_re: Fix RQE posting logic
Devesh Sharma [Mon, 22 May 2017 10:15:40 +0000 (03:15 -0700)]
RDMA/bnxt_re: Fix RQE posting logic

This patch adds code to ring RQ Doorbell aggressively
so that the adapter can DMA RQ buffers sooner, instead
of DMA all WQEs in the post_recv WR list together at the
end of the post_recv verb.
Also use spinlock to serialize RQ posting

Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Signed-off-by: Devesh Sharma <devesh.sharma@broadcom.com>
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/bnxt_re: Add HW workaround for avoiding stall for UD QPs
Somnath Kotur [Mon, 22 May 2017 10:15:36 +0000 (03:15 -0700)]
RDMA/bnxt_re: Add HW workaround for avoiding stall for UD QPs

HW stalls out after 0x800000 WQEs are posted for UD QPs.
To workaround this problem, driver will send a modify_qp cmd
to the HW at around the halfway mark(0x400000) so that FW
can accordingly modify the QP context in the HW to prevent this
stall.
This workaround needs to be done for UD, QP1 and Raw Ethertype
packets. Added a counter to keep track of WQEs posted during post_send.

Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/bnxt_re: Dereg MR in FW before freeing the fast_reg_page_list
Selvin Xavier [Mon, 22 May 2017 10:15:34 +0000 (03:15 -0700)]
RDMA/bnxt_re: Dereg MR in FW before freeing the fast_reg_page_list

If the host buffers are freed before destroying MR in HW,
HW could try accessing these buffers. This could cause a host
crash. Fixing the code to avoid this condition.

Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/bnxt_re: HW workarounds for handling specific conditions
Eddie Wai [Wed, 14 Jun 2017 10:26:23 +0000 (03:26 -0700)]
RDMA/bnxt_re: HW workarounds for handling specific conditions

This patch implements the following HW workarounds

1. The SQ depth needs to be augmented  by 128 + 1 to avoid running
   into an Out of order CQE issue
2. Workaround to handle the problem where the HW fast path engine continues
   to access DMA memory in retranmission mode even after the WQE has
   already been completed. If the HW reports this condition, driver detects
   it and posts a Fence WQE. The driver stops reporting the completions
   to stack until it receives completion  for Fence WQE.

Signed-off-by: Eddie Wai <eddie.wai@broadcom.com>
Signed-off-by: Sriharsha Basavapatna <sriharsha.basavapatna@broadcom.com>
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/bnxt_re: Fixing the Control path command and response handling
Devesh Sharma [Mon, 22 May 2017 10:15:31 +0000 (03:15 -0700)]
RDMA/bnxt_re: Fixing the Control path command and response handling

Fixing a concurrency issue with creq handling. Each caller
was given a globally managed crsq element, which was
accessed outside a lock. This could result in corruption,
if lot of applications are simultaneously issuing Control Path
commands. Now, each caller will provide its own response buffer
and the responses will be copied under a lock.
Also, Fixing the queue full condition check for the CMDQ.

As a part of these changes, the control path code is refactored
to remove the code replication in the response status checking.

Signed-off-by: Devesh Sharma <devesh.sharma@broadcom.com>
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoIB/addr: Fix setting source address in addr6_resolve()
Roland Dreier [Tue, 6 Jun 2017 16:22:00 +0000 (09:22 -0700)]
IB/addr: Fix setting source address in addr6_resolve()

Commit eea40b8f624f ("infiniband: call ipv6 route lookup via the stub
interface") introduced a regression in address resolution when connecting
to IPv6 destination addresses.  The old code called ip6_route_output(),
while the new code calls ipv6_stub->ipv6_dst_lookup().  The two are almost
the same, except that ipv6_dst_lookup() also calls ip6_route_get_saddr()
if the source address is in6addr_any.

This means that the test of ipv6_addr_any(&fl6.saddr) now never succeeds,
and so we never copy the source address out.  This ends up causing
rdma_resolve_addr() to fail, because without a resolved source address,
cma_acquire_dev() will fail to find an RDMA device to use.  For me, this
causes connecting to an NVMe over Fabrics target via RoCE / IPv6 to fail.

Fix this by copying out fl6.saddr if ipv6_addr_any() is true for the original
source address passed into addr6_resolve().  We can drop our call to
ipv6_dev_get_saddr() because ipv6_dst_lookup() already does that work.

Fixes: eea40b8f624 ("infiniband: call ipv6 route lookup via the stub interface")
Cc: <stable@vger.kernel.org> # 3.12+
Signed-off-by: Roland Dreier <roland@purestorage.com>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/SA: Fix kernel panic in CMA request handler flow
Majd Dibbiny [Sun, 21 May 2017 16:09:54 +0000 (19:09 +0300)]
RDMA/SA: Fix kernel panic in CMA request handler flow

Commit 9fdca4da4d8c (IB/SA: Split struct sa_path_rec based on IB and
ROCE specific fields) moved the service_id to be specific attribute
for IB and OPA SA Path Record, and thus wasn't assigned for RoCE.

This caused to the following kernel panic in the CMA request handler flow:

[   27.074594] BUG: unable to handle kernel NULL pointer dereference at 0000000000000008
[   27.074731] IP: __radix_tree_lookup+0x1d/0xe0
...
[   27.075356] Workqueue: ib_cm cm_work_handler [ib_cm]
[   27.075401] task: ffff88022e3b8000 task.stack: ffffc90001298000
[   27.075449] RIP: 0010:__radix_tree_lookup+0x1d/0xe0
...
[   27.075979] Call Trace:
[   27.076015]  radix_tree_lookup+0xd/0x10
[   27.076055]  cma_ps_find+0x59/0x70 [rdma_cm]
[   27.076097]  cma_id_from_event+0xd2/0x470 [rdma_cm]
[   27.076144]  ? ib_init_ah_from_path+0x39a/0x590 [ib_core]
[   27.076193]  cma_req_handler+0x25/0x480 [rdma_cm]
[   27.076237]  cm_process_work+0x25/0x120 [ib_cm]
[   27.076280]  ? cm_get_bth_pkey.isra.62+0x3c/0xa0 [ib_cm]
[   27.076350]  cm_req_handler+0xb03/0xd40 [ib_cm]
[   27.076430]  ? sched_clock_cpu+0x11/0xb0
[   27.076478]  cm_work_handler+0x194/0x1588 [ib_cm]
[   27.076525]  process_one_work+0x160/0x410
[   27.076565]  worker_thread+0x137/0x4a0
[   27.076614]  kthread+0x112/0x150
[   27.076684]  ? max_active_store+0x60/0x60
[   27.077642]  ? kthread_park+0x90/0x90
[   27.078530]  ret_from_fork+0x2c/0x40

This patch moves it back to the common SA Path Record structure
and removes the redundant setter and getter.

Tested on Connect-IB and Connect-X4 in Infiniband and RoCE respectively.

Fixes: 9fdca4da4d8c (IB/SA: Split struct sa_path_rec based on IB ands
ROCE specific fields)
Signed-off-by: Majd Dibbiny <majd@mellanox.com>
Reviewed-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/umem: Fix missing mmap_sem in get umem ODP call
Leon Romanovsky [Sun, 21 May 2017 16:08:09 +0000 (19:08 +0300)]
RDMA/umem: Fix missing mmap_sem in get umem ODP call

Add mmap_sem lock around VMA inspection in ib_umem_odp_get().

Fixes: 0008b84ea9af ('IB/umem: Add support to huge ODP')
Signed-off-by: Artemy Kovalyov <artemyko@mellanox.com>
Reviewed-by: Eli Cohen <eli@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/core: not to set page dirty bit if it's already set.
Qing Huang [Thu, 18 May 2017 23:33:53 +0000 (16:33 -0700)]
RDMA/core: not to set page dirty bit if it's already set.

This change will optimize kernel memory deregistration operations.
__ib_umem_release() used to call set_page_dirty_lock() against every
writable page in its memory region. Its purpose is to keep data
synced between CPU and DMA device when swapping happens after mem
deregistration ops. Now we choose not to set page dirty bit if it's
already set by kernel prior to calling __ib_umem_release(). This
reduces memory deregistration time by half or even more when we ran
application simulation test program.

Signed-off-by: Qing Huang <qing.huang@oracle.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/uverbs: Declare local function static and add brackets to sizeof
Leon Romanovsky [Thu, 18 May 2017 04:40:33 +0000 (07:40 +0300)]
RDMA/uverbs: Declare local function static and add brackets to sizeof

Commit 57520751445b ("IB/SA: Add OPA path record type") introduced
new local function __ib_copy_path_rec_to_user, but didn't limit its
scope. This produces the following sparse warning:

drivers/infiniband/core/uverbs_marshall.c:99:6: warning:
symbol '__ib_copy_path_rec_to_user' was not declared. Should it be
static?

In addition, it used sizeof ... notations instead of sizeof(...), which
is correct in C, but a little bit misleading. Let's change it too.

Fixes: 57520751445b ("IB/SA: Add OPA path record type")
Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/netlink: Reduce exposure of RDMA netlink functions
Leon Romanovsky [Sun, 14 May 2017 12:49:57 +0000 (15:49 +0300)]
RDMA/netlink: Reduce exposure of RDMA netlink functions

RDMA netlink is part of ib_core, hence ibnl_chk_listeners(),
ibnl_init() and ibnl_cleanup() don't need to be published
in public header file.

Let's remove EXPORT_SYMBOL from ibnl_chk_listeners() and move all these
functions to private header file.

CC: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/srp: Fix NULL deref at srp_destroy_qp()
Israel Rukshin [Thu, 11 May 2017 15:52:36 +0000 (18:52 +0300)]
RDMA/srp: Fix NULL deref at srp_destroy_qp()

If srp_init_qp() fails at srp_create_ch_ib() then ch->send_cq
may be NULL.
Calling directly to ib_destroy_qp() is sufficient because
no work requests were posted on the created qp.

Fixes: 9294000d6d89 ("IB/srp: Drain the send queue before destroying a QP")
Cc: <stable@vger.kernel.org>
Signed-off-by: Israel Rukshin <israelr@mellanox.com>
Reviewed-by: Max Gurtovoy <maxg@mellanox.com>
Reviewed-by: Bart van Assche <bart.vanassche@sandisk.com>--
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/IPoIB: Limit the ipoib_dev_uninit_default scope
Leon Romanovsky [Sun, 14 May 2017 10:32:06 +0000 (13:32 +0300)]
RDMA/IPoIB: Limit the ipoib_dev_uninit_default scope

ipoib_dev_uninit_default() call is used in ipoib_main.c file only
and it generates the following warning from smatch tool:
drivers/infiniband/ulp/ipoib/ipoib_main.c:1593:6: warning:
symbol 'ipoib_dev_uninit_default' was not declared. Should it
be static?

so let's declare that function as static.

Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/IPoIB: Replace netdev_priv with ipoib_priv for ipoib_get_link_ksettings
Honggang Li [Thu, 11 May 2017 12:14:28 +0000 (20:14 +0800)]
RDMA/IPoIB: Replace netdev_priv with ipoib_priv for ipoib_get_link_ksettings

ipoib_dev_init accesses the wrong private data for the IPoIB device.
Commit cd565b4b51e5 (IB/IPoIB: Support acceleration options callbacks)
changed ipoib_priv from being identical to netdev_priv to being an
area inside of, but not the same pointer as, the netdev_priv pointer.
As such, the struct we want is the ipoib_priv area, not the netdev_priv
area, so use the right accessor, otherwise we kernel panic.

[   27.271938] IPv6: ADDRCONF(NETDEV_CHANGE): mlx5_ib0.8006: link becomes ready
[   28.156790] BUG: unable to handle kernel NULL pointer dereference at 000000000000067c
[   28.166309] IP: ib_query_port+0x30/0x180 [ib_core]
...
[   28.306282] RIP: 0010:ib_query_port+0x30/0x180 [ib_core]
...
[   28.393337] Call Trace:
[   28.397594]  ipoib_get_link_ksettings+0x66/0xe0 [ib_ipoib]
[   28.405274]  __ethtool_get_link_ksettings+0xa0/0x1c0
[   28.412353]  speed_show+0x74/0xa0
[   28.417503]  dev_attr_show+0x20/0x50
[   28.422922]  ? mutex_lock+0x12/0x40
[   28.428179]  sysfs_kf_seq_show+0xbf/0x1a0
[   28.434002]  kernfs_seq_show+0x21/0x30
[   28.439470]  seq_read+0x116/0x3b0
[   28.444445]  ? do_filp_open+0xa5/0x100
[   28.449774]  kernfs_fop_read+0xff/0x180
[   28.455220]  __vfs_read+0x37/0x150
[   28.460167]  ? security_file_permission+0x9d/0xc0
[   28.466560]  vfs_read+0x8c/0x130
[   28.471318]  SyS_read+0x55/0xc0
[   28.475950]  do_syscall_64+0x67/0x150
[   28.481163]  entry_SYSCALL64_slow_path+0x25/0x25
...
[   28.584493] ---[ end trace 3549968a4bf0aa5d ]---

Fixes: cd565b4b51e5 (IB/IPoIB: Support acceleration options callbacks)
Fixes: 0d7e2d2166f6 (IB/ipoib: add get_link_ksettings in ethtool)
Signed-off-by: Honggang Li <honli@redhat.com>
Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/qedr: add null check before pointer dereference
Gustavo A. R. Silva [Mon, 22 May 2017 18:30:15 +0000 (13:30 -0500)]
RDMA/qedr: add null check before pointer dereference

Add null check before dereferencing pointer sgid_attr.ndev
inside function rdma_vlan_dev_vlan_id().

Addresses-Coverity-ID: 1373979
Signed-off-by: Gustavo A. R. Silva <garsilva@embeddedor.com>
Acked-by: Ram Amrani <Ram.Amrani@cavium.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/mlx5: set UMR wqe fence according to HCA cap
Max Gurtovoy [Sun, 28 May 2017 07:53:11 +0000 (10:53 +0300)]
RDMA/mlx5: set UMR wqe fence according to HCA cap

Cache the needed umr_fence and set the wqe ctrl segmennt
accordingly.

Signed-off-by: Max Gurtovoy <maxg@mellanox.com>
Acked-by: Leon Romanovsky <leon@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agonet/mlx5: Define interface bits for fencing UMR wqe
Max Gurtovoy [Sun, 28 May 2017 07:53:10 +0000 (10:53 +0300)]
net/mlx5: Define interface bits for fencing UMR wqe

HW can implement UMR wqe re-transmission in various ways.
Thus, add HCA cap to distinguish the needed fence for UMR to make
sure that the wqe wouldn't fail on mkey checks.

Signed-off-by: Max Gurtovoy <maxg@mellanox.com>
Acked-by: Leon Romanovsky <leon@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/mlx4: Fix MAD tunneling when SRIOV is enabled
Jack Morgenstein [Sun, 21 May 2017 16:11:13 +0000 (19:11 +0300)]
RDMA/mlx4: Fix MAD tunneling when SRIOV is enabled

The cited patch added a type field to structures ib_ah and rdma_ah_attr.

Function mlx4_ib_query_ah() builds an rdma_ah_attr structure from the
data in an mlx4_ib_ah structure (which contains both an ib_ah structure
and an address vector).

For mlx4_ib_query_ah() to work properly, the type field in the contained
ib_ah structure must be set correctly.

In the outgoing MAD tunneling flow, procedure mlx4_ib_multiplex_mad()
paravirtualizes a MAD received from a slave and sends the processed
mad out over the wire. During this processing, it populates an
mlx4_ib_ah structure and calls mlx4_ib_query_ah().

The cited commit overlooked setting the type field in the contained
ib_ah structure before invoking mlx4_ib_query_ah(). As a result, the
type field remained uninitialized, and the rdma_ah_attr structure was
incorrectly built. This resulted in improperly built MADs being sent out
over the wire.

This patch properly initializes the type field in the contained ib_ah
structure before calling mlx4_ib_query_ah(). The rdma_ah_attr structure
is then generated correctly.

Fixes: 44c58487d51a ("IB/core: Define 'ib' and 'roce' rdma_ah_attr types")
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/qib,hfi1: Fix MR reference count leak on write with immediate
Mike Marciniszyn [Fri, 12 May 2017 16:02:00 +0000 (09:02 -0700)]
RDMA/qib,hfi1: Fix MR reference count leak on write with immediate

The handling of IB_RDMA_WRITE_ONLY_WITH_IMMEDIATE will leak a memory
reference when a buffer cannot be allocated for returning the immediate
data.

The issue is that the rkey validation has already occurred and the RNR
nak fails to release the reference that was fruitlessly gotten.  The
the peer will send the identical single packet request when its RNR
timer pops.

The fix is to release the held reference prior to the rnr nak exit.
This is the only sequence the requires both rkey validation and the
buffer allocation on the same packet.

Cc: Stable <stable@vger.kernel.org> # 4.7+
Tested-by: Tadeusz Struk <tadeusz.struk@intel.com>
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/hfi1: Defer setting VL15 credits to link-up interrupt
Byczkowski, Jakub [Fri, 12 May 2017 16:01:37 +0000 (09:01 -0700)]
RDMA/hfi1: Defer setting VL15 credits to link-up interrupt

Keep VL15 credits at 0 during LNI, before link-up. Store
VL15 credits value during verify cap interrupt and set
in after link-up. This addresses an issue where VL15 MAD
packets could be sent by one side of the link before
the other side is ready to receive them.

Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Reviewed-by: Dean Luick <dean.luick@intel.com>
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jakub Byczkowski <jakub.byczkowski@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/hfi1: change PCI bar addr assignments to Linux API functions
Steven L. Roberts [Wed, 10 May 2017 19:58:13 +0000 (14:58 -0500)]
RDMA/hfi1: change PCI bar addr assignments to Linux API functions

The Omni-Path adapter driver fails to load on the ppc64le platform
due to invalid PCI setup.

This patch makes the PCI configuration more robust and will
fix 64 bit addressing for ppc64le.

Signed-off-by: Steven L Roberts <robers97@gmail.com>
Acked-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/hfi1: fix array termination by appending NULL to attr array
Steven L. Roberts [Wed, 10 May 2017 15:54:12 +0000 (10:54 -0500)]
RDMA/hfi1: fix array termination by appending NULL to attr array

This fixes a kernel panic when loading the hfi driver as a dynamic module.

Signed-off-by: Steven L Roberts <robers97@gmail.com>
Reviewed-by: Leon Romanovsky <leon@kernel.org>
Acked-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/iw_cxgb4: fix the calculation of ipv6 header size
Raju Rangoju [Wed, 31 May 2017 06:36:58 +0000 (12:06 +0530)]
RDMA/iw_cxgb4: fix the calculation of ipv6 header size

Take care of ipv6 checks while computing header length for deducing mtu
size of ipv6 servers. Due to the incorrect header length computation for
ipv6 servers, wrong mss is reported to the peer (client).

Signed-off-by: Raju Rangoju <rajur@chelsio.com>
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
Acked-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/iw_cxgb4: calculate t4_eq_status_entries properly
Ganesh Goudar [Fri, 19 May 2017 09:18:42 +0000 (14:48 +0530)]
RDMA/iw_cxgb4: calculate t4_eq_status_entries properly

use egrstatuspagesize to calculate t4_eq_status_entries.

Fixes: bb58d07964f2 ("cxgb4: Update IngPad and IngPack values")
Reported-by: Logan Gunthorpe <logang@deltatee.com>
Signed-off-by: Potnuri Bharat Teja <bharat@chelsio.com>
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
Acked-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/iw_cxgb4: Avoid touch after free error in ARP failure handlers
Raju Rangoju [Mon, 15 May 2017 06:40:39 +0000 (06:40 +0000)]
RDMA/iw_cxgb4: Avoid touch after free error in ARP failure handlers

The patch 761e19a504af (RDMA/iw_cxgb4: Handle return value of
c4iw_ofld_send() in abort_arp_failure()) from May 6, 2016
leads to the following static checker warning:
drivers/infiniband/hw/cxgb4/cm.c:575 abort_arp_failure()
warn: passing freed memory 'skb'

Also fixes skb leak when l2t resolution fails

Fixes: 761e19a504afa55 (RDMA/iw_cxgb4: Handle return value of
c4iw_ofld_send() in abort_arp_failure())
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Raju Rangoju <rajur@chelsio.com>
Reviewed-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/nes: ACK MPA Reply frame
Tatyana Nikolova [Wed, 10 May 2017 21:46:40 +0000 (16:46 -0500)]
RDMA/nes: ACK MPA Reply frame

Explicitly ACK the MPA Reply frame so the peer
does not retransmit the frame.

Signed-off-by: Tatyana Nikolova <tatyana.e.nikolova@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/nes: Don't set 0-length FULPDU RTR indication control flag
Tatyana Nikolova [Wed, 10 May 2017 21:46:39 +0000 (16:46 -0500)]
RDMA/nes: Don't set 0-length FULPDU RTR indication control flag

Don't set control flag for 0-length FULPDU (Send) RTR indication
in the enhanced MPA Request/Reply frames, because it isn't supported.

Signed-off-by: Tatyana Nikolova <tatyana.e.nikolova@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/i40iw: fix duplicated code for different branches
Gustavo A. R. Silva [Thu, 18 May 2017 18:11:17 +0000 (13:11 -0500)]
RDMA/i40iw: fix duplicated code for different branches

Refactor code to avoid identical code for different branches.

Addresses-Coverity-ID: 1357356
Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Gustavo A. R. Silva <garsilva@embeddedor.com>
Acked-by: Shiraz Saleem <shiraz.saleem@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/i40iw: Remove MSS change support
Shiraz Saleem [Fri, 19 May 2017 21:14:02 +0000 (16:14 -0500)]
RDMA/i40iw: Remove MSS change support

MSS change on active QPs is not supported. Store new MSS
value for new QPs only. Remove code to modify MSS on the fly.
This also resolves a crash on QP modify to QP 0.

BUG: unable to handle kernel NULL pointer dereference at 0000000000000008
IP: i40iw_sc_qp_modify+0x22/0x280 [i40iw]
Oops: 0000 [#1] SMP KASAN
CPU: 2 PID: 1236 Comm: kworker/u16:4 Not tainted 4.12.0-rc1 #5
Hardware name: Gigabyte Technology Co., Ltd. To be filled by O.E.M./Q87M-D2H,
BIOS F7 01/17/2014
Workqueue: l2params i40iw_l2params_worker [i40iw]
task: ffff88070f5a9b40 task.stack: ffff88070f5a0000
RIP: 0010:i40iw_sc_qp_modify+0x22/0x280 [i40iw]
...
Call Trace:
i40iw_exec_cqp_cmd+0x2ce/0x410 [i40iw]
? _raw_spin_lock_irqsave+0x6f/0x80
? i40iw_process_cqp_cmd+0x1d/0x80 [i40iw]
i40iw_process_cqp_cmd+0x7c/0x80 [i40iw]
i40iw_handle_cqp_op+0x2f/0x200 [i40iw]
? trace_hardirqs_off+0xd/0x10
? _raw_spin_unlock_irqrestore+0x46/0x50
i40iw_hw_modify_qp+0x5e/0x90 [i40iw]
i40iw_qp_mss_modify+0x52/0x60 [i40iw]
i40iw_change_l2params+0x145/0x160 [i40iw]
i40iw_l2params_worker+0x1f/0x40 [i40iw]
process_one_work+0x1f5/0x650
? process_one_work+0x161/0x650
worker_thread+0x48/0x3b0
kthread+0x112/0x150
? process_one_work+0x650/0x650
? kthread_create_on_node+0x40/0x40
ret_from_fork+0x2e/0x40
Code: 2e 0f 1f 84 00 00 00 00 00 55 48 89 e5 41 56 41 55 41 89 cd 41 54 49 89 fc
53 48 89 f3 48 89 d6 48 83 ec 08 48 8b 87 10 01 00 00 <48> 8b 40 08 4c 8b b0 40 04
00 00 4c 89 f7 e8 1b e5 ff ff 48 85
RIP: i40iw_sc_qp_modify+0x22/0x280 [i40iw] RSP: ffff88070f5a7c28
CR2: 0000000000000008
---[ end trace 77a405931e296060 ]---

Reported-by: Stefan Assmann <sassmann@redhat.com>
Signed-off-by: Shiraz Saleem <shiraz.saleem@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/i40iw: Fix device initialization error path
Mustafa Ismail [Thu, 11 May 2017 04:32:14 +0000 (23:32 -0500)]
RDMA/i40iw: Fix device initialization error path

Some error paths in i40iw_initialize_dev are doing
additional and unnecessary work before exiting.
Correctly free resources allocated prior to error
and return with correct status code.

Signed-off-by: Mustafa Ismail <mustafa.ismail@intelcom>
Signed-off-by: Henry Orosco <henry.orosco@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/i40iw: ACK MPA Reject frame
Tatyana Nikolova [Fri, 5 May 2017 17:05:17 +0000 (12:05 -0500)]
RDMA/i40iw: ACK MPA Reject frame

Explicitly ACK the MPA Reject frame so the peer does
not retransmit the frame.

Signed-off-by: Tatyana Nikolova <tatyana.e.nikolova@intel.com>
Signed-off-by: Henry Orosco <henry.orosco@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoRDMA/i40iw: Don't set 0-length FULPDU RTR indication control flag
Tatyana Nikolova [Fri, 5 May 2017 17:05:16 +0000 (12:05 -0500)]
RDMA/i40iw: Don't set 0-length FULPDU RTR indication control flag

Don't set control flag for 0-length FULPDU (Send)
RTR indication in the enhanced MPA Request/Reply
frames, because it isn't supported.

Signed-off-by: Tatyana Nikolova <tatyana.e.nikolova@intel.com>
Signed-off-by: Henry Orosco <henry.orosco@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
7 years agoLinux 4.12-rc3
Linus Torvalds [Mon, 29 May 2017 00:20:53 +0000 (17:20 -0700)]
Linux 4.12-rc3

7 years agoMerge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/evalenti/linux...
Linus Torvalds [Sun, 28 May 2017 23:18:27 +0000 (16:18 -0700)]
Merge branch 'fixes' of git://git./linux/kernel/git/evalenti/linux-soc-thermal

Pull thermal SoC management fixes from Eduardo Valentin:

 - fixes to TI SoC driver, Broadcom, qoriq

 - small sparse warning fix on thermal core

* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/evalenti/linux-soc-thermal:
  thermal: broadcom: ns-thermal: default on iProc SoCs
  ti-soc-thermal: Fix a typo in a comment line
  ti-soc-thermal: Delete error messages for failed memory allocations in ti_bandgap_build()
  ti-soc-thermal: Use devm_kcalloc() in ti_bandgap_build()
  thermal: core: make thermal_emergency_poweroff static
  thermal: qoriq: remove useless call for of_thermal_get_trip_points()

7 years agoMerge tag 'tty-4.12-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
Linus Torvalds [Sat, 27 May 2017 16:39:09 +0000 (09:39 -0700)]
Merge tag 'tty-4.12-rc3' of git://git./linux/kernel/git/gregkh/tty

Pull tty/serial fixes from Greg KH:
 "Here are some serial and tty fixes for 4.12-rc3. They are a bit bigger
  than normal, which is why I had them bake in linux-next for a few
  weeks and didn't send them to you for -rc2.

  They revert a few of the serdev patches from 4.12-rc1, and bring
  things back to how they were in 4.11, to try to make things a bit more
  stable there. Rob and Johan both agree that this is the way forward,
  so this isn't people squabbling over semantics. Other than that, just
  a few minor serial driver fixes that people have had problems with.

  All of these have been in linux-next for a few weeks with no reported
  issues"

* tag 'tty-4.12-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  serial: altera_uart: call iounmap() at driver remove
  serial: imx: ensure UCR3 and UFCR are setup correctly
  MAINTAINERS/serial: Change maintainer of jsm driver
  serial: enable serdev support
  tty/serdev: add serdev registration interface
  serdev: Restore serdev_device_write_buf for atomic context
  serial: core: fix crash in uart_suspend_port
  tty: fix port buffer locking
  tty: ehv_bytechan: clean up init error handling
  serial: ifx6x60: fix use-after-free on module unload
  serial: altera_jtaguart: adding iounmap()
  serial: exar: Fix stuck MSIs
  serial: efm32: Fix parity management in 'efm32_uart_console_get_options()'
  serdev: fix tty-port client deregistration
  Revert "tty_port: register tty ports with serdev bus"
  drivers/tty: 8250: only call fintek_8250_probe when doing port I/O

7 years agoMerge tag 'powerpc-4.12-4' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc...
Linus Torvalds [Sat, 27 May 2017 16:28:34 +0000 (09:28 -0700)]
Merge tag 'powerpc-4.12-4' of git://git./linux/kernel/git/powerpc/linux

Pull powerpc fixes from Michael Ellerman:
 "Fix running SPU programs on Cell, and a few other minor fixes.

  Thanks to Alistair Popple, Jeremy Kerr, Michael Neuling, Nicholas
  Piggin"

* tag 'powerpc-4.12-4' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  powerpc: Add PPC_FEATURE userspace bits for SCV and DARN instructions
  powerpc/spufs: Fix hash faults for kernel regions
  powerpc: Fix booting P9 hash with CONFIG_PPC_RADIX_MMU=N
  powerpc/powernv/npu-dma.c: Fix opal_npu_destroy_context() call
  selftests/powerpc: Fix TM resched DSCR test with some compilers

7 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 27 May 2017 16:17:58 +0000 (09:17 -0700)]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Thomas Gleixner:
 "A series of fixes for X86:

   - The final fix for the end-of-stack issue in the unwinder
   - Handle non PAT systems gracefully
   - Prevent access to uninitiliazed memory
   - Move early delay calaibration after basic init
   - Fix Kconfig help text
   - Fix a cross compile issue
   - Unbreak older make versions"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/timers: Move simple_udelay_calibration past init_hypervisor_platform
  x86/alternatives: Prevent uninitialized stack byte read in apply_alternatives()
  x86/PAT: Fix Xorg regression on CPUs that don't support PAT
  x86/watchdog: Fix Kconfig help text file path reference to lockup watchdog documentation
  x86/build: Permit building with old make versions
  x86/unwind: Add end-of-stack check for ftrace handlers
  Revert "x86/entry: Fix the end of the stack for newly forked tasks"
  x86/boot: Use CROSS_COMPILE prefix for readelf

7 years agoMerge branch 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 27 May 2017 16:14:24 +0000 (09:14 -0700)]
Merge branch 'timers-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull timer fixlet from Thomas Gleixner:
 "Silence dmesg spam by making the posix cpu timer printks depend on
  print_fatal_signals"

* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  posix-timers: Make signal printks conditional

7 years agoMerge branch 'ras-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 27 May 2017 16:06:43 +0000 (09:06 -0700)]
Merge branch 'ras-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull RAS fixes from Thomas Gleixner:
 "Two fixlets for RAS:

   - Export memory_error() so the NFIT module can utilize it

   - Handle memory errors in NFIT correctly"

* 'ras-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  acpi, nfit: Fix the memory error check in nfit_handle_mce()
  x86/MCE: Export memory_error()

7 years agoMerge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 27 May 2017 16:02:41 +0000 (09:02 -0700)]
Merge branch 'perf-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull perf tooling fixes from Thomas Gleixner:

 - Synchronization of tools and kernel headers

 - A series of fixes for perf report addressing various failures:
    * Handle invalid maps proper
    * Plug a memory leak
    * Handle frames and callchain order correctly

 - Fixes for handling inlines and children mode

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  tools/include: Sync kernel ABI headers with tooling headers
  perf tools: Put caller above callee in --children mode
  perf report: Do not drop last inlined frame
  perf report: Always honor callchain order for inlined nodes
  perf script: Add --inline option for debugging
  perf report: Fix off-by-one for non-activation frames
  perf report: Fix memory leak in addr2line when called by addr2inlines
  perf report: Don't crash on invalid maps in `-g srcline` mode

7 years agoMerge branch 'locking-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 27 May 2017 15:59:37 +0000 (08:59 -0700)]
Merge branch 'locking-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull locking fix from Thomas Gleixner:
 "A fix for a state leak which was introduced in the recent rework of
  futex/rtmutex interaction"

* 'locking-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  futex,rt_mutex: Fix rt_mutex_cleanup_proxy_lock()

7 years agoMerge branch 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 27 May 2017 15:52:27 +0000 (08:52 -0700)]
Merge branch 'core-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull kthread fix from Thomas Gleixner:
 "A single fix which prevents a use after free when kthread fork fails"

* 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  kthread: Fix use-after-free if kthread fork fails

7 years agoMerge tag 'trace-v4.12-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt...
Linus Torvalds [Sat, 27 May 2017 15:30:30 +0000 (08:30 -0700)]
Merge tag 'trace-v4.12-rc2' of git://git./linux/kernel/git/rostedt/linux-trace

Pull ftrace fixes from Steven Rostedt:
 "There's been a few memory issues found with ftrace.

  One was simply a memory leak where not all was being freed that should
  have been in releasing a file pointer on set_graph_function.

  Then Thomas found that the ftrace trampolines were marked for
  read/write as well as execute. To shrink the possible attack surface,
  he added calls to set them to ro. Which also uncovered some other
  issues with freeing module allocated memory that had its permissions
  changed.

  Kprobes had a similar issue which is fixed and a selftest was added to
  trigger that issue again"

* tag 'trace-v4.12-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  x86/ftrace: Make sure that ftrace trampolines are not RWX
  x86/mm/ftrace: Do not bug in early boot on irqs_disabled in cpu_flush_range()
  selftests/ftrace: Add a testcase for many kprobe events
  kprobes/x86: Fix to set RWX bits correctly before releasing trampoline
  ftrace: Fix memory leak in ftrace_graph_release()

7 years agox86/ftrace: Make sure that ftrace trampolines are not RWX
Thomas Gleixner [Thu, 25 May 2017 08:57:51 +0000 (10:57 +0200)]
x86/ftrace: Make sure that ftrace trampolines are not RWX

ftrace use module_alloc() to allocate trampoline pages. The mapping of
module_alloc() is RWX, which makes sense as the memory is written to right
after allocation. But nothing makes these pages RO after writing to them.

Add proper set_memory_rw/ro() calls to protect the trampolines after
modification.

Link: http://lkml.kernel.org/r/alpine.DEB.2.20.1705251056410.1862@nanos
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
7 years agox86/mm/ftrace: Do not bug in early boot on irqs_disabled in cpu_flush_range()
Steven Rostedt (VMware) [Fri, 26 May 2017 14:14:11 +0000 (10:14 -0400)]
x86/mm/ftrace: Do not bug in early boot on irqs_disabled in cpu_flush_range()

With function tracing starting in early bootup and having its trampoline
pages being read only, a bug triggered with the following:

kernel BUG at arch/x86/mm/pageattr.c:189!
invalid opcode: 0000 [#1] SMP
Modules linked in:
CPU: 0 PID: 0 Comm: swapper Not tainted 4.12.0-rc2-test+ #3
Hardware name: MSI MS-7823/CSM-H87M-G43 (MS-7823), BIOS V1.6 02/22/2014
task: ffffffffb4222500 task.stack: ffffffffb4200000
RIP: 0010:change_page_attr_set_clr+0x269/0x302
RSP: 0000:ffffffffb4203c88 EFLAGS: 00010046
RAX: 0000000000000046 RBX: 0000000000000000 RCX: 00000001b6000000
RDX: ffffffffb4203d40 RSI: 0000000000000000 RDI: ffffffffb4240d60
RBP: ffffffffb4203d18 R08: 00000001b6000000 R09: 0000000000000001
R10: ffffffffb4203aa8 R11: 0000000000000003 R12: ffffffffc029b000
R13: ffffffffb4203d40 R14: 0000000000000001 R15: 0000000000000000
FS:  0000000000000000(0000) GS:ffff9a639ea00000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: ffff9a636b384000 CR3: 00000001ea21d000 CR4: 00000000000406b0
Call Trace:
 change_page_attr_clear+0x1f/0x21
 set_memory_ro+0x1e/0x20
 arch_ftrace_update_trampoline+0x207/0x21c
 ? ftrace_caller+0x64/0x64
 ? 0xffffffffc029b000
 ftrace_startup+0xf4/0x198
 register_ftrace_function+0x26/0x3c
 function_trace_init+0x5e/0x73
 tracer_init+0x1e/0x23
 tracing_set_tracer+0x127/0x15a
 register_tracer+0x19b/0x1bc
 init_function_trace+0x90/0x92
 early_trace_init+0x236/0x2b3
 start_kernel+0x200/0x3f5
 x86_64_start_reservations+0x29/0x2b
 x86_64_start_kernel+0x17c/0x18f
 secondary_startup_64+0x9f/0x9f
 ? secondary_startup_64+0x9f/0x9f

Interrupts should not be enabled at this early in the boot process. It is
also fine to leave interrupts enabled during this time as there's only one
CPU running, and on_each_cpu() means to only run on the current CPU.

If early_boot_irqs_disabled is set, it is safe to run cpu_flush_range() with
interrupts disabled. Don't trigger a BUG_ON() in that case.

Link: http://lkml.kernel.org/r/20170526093717.0be3b849@gandalf.local.home
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
7 years agoselftests/ftrace: Add a testcase for many kprobe events
Masami Hiramatsu [Fri, 26 May 2017 04:44:54 +0000 (13:44 +0900)]
selftests/ftrace: Add a testcase for many kprobe events

Add a testcase to test kprobes via ftrace interface
with many concurrent kprobe events.

This tries to add many kprobe events (up to 256) on
kernel functions. To avoid making ftrace-based
kprobes (kprobes on fentry), it skips first N bytes
(on x86 N=5, on ppc or arm N=4) of function entry.
After that, it enables all those events, disable it,
and remove it.

Since the unoptimization buffer reclaiming will
be delayed, after removing events, it will wait
enough time.

Link: http://lkml.kernel.org/r/149577388470.11702.11832460851769204511.stgit@devbox
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Suggested-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
7 years agokprobes/x86: Fix to set RWX bits correctly before releasing trampoline
Masami Hiramatsu [Thu, 25 May 2017 10:38:17 +0000 (19:38 +0900)]
kprobes/x86: Fix to set RWX bits correctly before releasing trampoline

Fix kprobes to set(recover) RWX bits correctly on trampoline
buffer before releasing it. Releasing readonly page to
module_memfree() crash the kernel.

Without this fix, if kprobes user register a bunch of kprobes
in function body (since kprobes on function entry usually
use ftrace) and unregister it, kernel hits a BUG and crash.

Link: http://lkml.kernel.org/r/149570868652.3518.14120169373590420503.stgit@devbox
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Fixes: d0381c81c2f7 ("kprobes/x86: Set kprobes pages read-only")
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
7 years agoftrace: Fix memory leak in ftrace_graph_release()
Luis Henriques [Thu, 25 May 2017 15:20:38 +0000 (16:20 +0100)]
ftrace: Fix memory leak in ftrace_graph_release()

ftrace_hash is being kfree'ed in ftrace_graph_release(), however the
->buckets field is not.  This results in a memory leak that is easily
captured by kmemleak:

unreferenced object 0xffff880038afe000 (size 8192):
  comm "trace-cmd", pid 238, jiffies 4294916898 (age 9.736s)
  hex dump (first 32 bytes):
    00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
    00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
  backtrace:
    [<ffffffff815f561e>] kmemleak_alloc+0x4e/0xb0
    [<ffffffff8113964d>] __kmalloc+0x12d/0x1a0
    [<ffffffff810bf6d1>] alloc_ftrace_hash+0x51/0x80
    [<ffffffff810c0523>] __ftrace_graph_open.isra.39.constprop.46+0xa3/0x100
    [<ffffffff810c05e8>] ftrace_graph_open+0x68/0xa0
    [<ffffffff8114003d>] do_dentry_open.isra.1+0x1bd/0x2d0
    [<ffffffff81140df7>] vfs_open+0x47/0x60
    [<ffffffff81150f95>] path_openat+0x2a5/0x1020
    [<ffffffff81152d6a>] do_filp_open+0x8a/0xf0
    [<ffffffff811411df>] do_sys_open+0x12f/0x200
    [<ffffffff811412ce>] SyS_open+0x1e/0x20
    [<ffffffff815fa6e0>] entry_SYSCALL_64_fastpath+0x13/0x94
    [<ffffffffffffffff>] 0xffffffffffffffff

Link: http://lkml.kernel.org/r/20170525152038.7661-1-lhenriques@suse.com
Cc: stable@vger.kernel.org
Fixes: b9b0c831bed2 ("ftrace: Convert graph filter to use hash tables")
Signed-off-by: Luis Henriques <lhenriques@suse.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
7 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Fri, 26 May 2017 23:45:13 +0000 (16:45 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

Pull input layer fixes from Dmitry Torokhov:
 "Just a few fixups to a couple of drivers"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: elan_i2c - ignore signals when finishing updating firmware
  Input: elan_i2c - clear INT before resetting controller
  Input: atmel_mxt_ts - add T100 as a readable object
  Input: edt-ft5x06 - increase allowed data range for threshold parameter

7 years agoMerge tag 'led_fixes_for_4-12-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 26 May 2017 21:02:30 +0000 (14:02 -0700)]
Merge tag 'led_fixes_for_4-12-rc3' of git://git./linux/kernel/git/j.anaszewski/linux-leds

Pull LED fix from Jacek Anaszewski:
 "A single LED fix for 4.12-rc3.

  leds-pca955x driver uses only i2c_smbus API and thus it should pass
  I2C_FUNC_SMBUS_BYTE_DATA flag to i2c_check_functionality"

* tag 'led_fixes_for_4-12-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/j.anaszewski/linux-leds:
  leds: pca955x: Correct I2C Functionality

7 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Fri, 26 May 2017 20:51:01 +0000 (13:51 -0700)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) Fix state pruning in bpf verifier wrt. alignment, from Daniel
    Borkmann.

 2) Handle non-linear SKBs properly in SCTP ICMP parsing, from Davide
    Caratti.

 3) Fix bit field definitions for rss_hash_type of descriptors in mlx5
    driver, from Jesper Brouer.

 4) Defer slave->link updates until bonding is ready to do a full commit
    to the new settings, from Nithin Sujir.

 5) Properly reference count ipv4 FIB metrics to avoid use after free
    situations, from Eric Dumazet and several others including Cong Wang
    and Julian Anastasov.

 6) Fix races in llc_ui_bind(), from Lin Zhang.

 7) Fix regression of ESP UDP encapsulation for TCP packets, from
    Steffen Klassert.

 8) Fix mdio-octeon driver Kconfig deps, from Randy Dunlap.

 9) Fix regression in setting DSCP on ipv6/GRE encapsulation, from Peter
    Dawson.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (43 commits)
  ipv4: add reference counting to metrics
  net: ethernet: ax88796: don't call free_irq without request_irq first
  ip6_tunnel, ip6_gre: fix setting of DSCP on encapsulated packets
  sctp: fix ICMP processing if skb is non-linear
  net: llc: add lock_sock in llc_ui_bind to avoid a race condition
  bonding: Don't update slave->link until ready to commit
  test_bpf: Add a couple of tests for BPF_JSGE.
  bpf: add various verifier test cases
  bpf: fix wrong exposure of map_flags into fdinfo for lpm
  bpf: add bpf_clone_redirect to bpf_helper_changes_pkt_data
  bpf: properly reset caller saved regs after helper call and ld_abs/ind
  bpf: fix incorrect pruning decision when alignment must be tracked
  arp: fixed -Wuninitialized compiler warning
  tcp: avoid fastopen API to be used on AF_UNSPEC
  net: move somaxconn init from sysctl code
  net: fix potential null pointer dereference
  geneve: fix fill_info when using collect_metadata
  virtio-net: enable TSO/checksum offloads for Q-in-Q vlans
  be2net: Fix offload features for Q-in-Q packets
  vlan: Fix tcp checksum offloads in Q-in-Q vlans
  ...

7 years agoMerge tag 'xfs-4.12-fixes-2' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux
Linus Torvalds [Fri, 26 May 2017 19:13:08 +0000 (12:13 -0700)]
Merge tag 'xfs-4.12-fixes-2' of git://git./fs/xfs/xfs-linux

Pull XFS fixes from Darrick Wong:
 "A few miscellaneous bug fixes & cleanups:

   - Fix indlen block reservation accounting bug when splitting delalloc
     extent

   - Fix warnings about unused variables that appeared in -rc1.

   - Don't spew errors when bmapping a local format directory

   - Fix an off-by-one error in a delalloc eof assertion

   - Make fsmap only return inode information for CAP_SYS_ADMIN

   - Fix a potential mount time deadlock recovering cow extents

   - Fix unaligned memory access in _btree_visit_blocks

   - Fix various SEEK_HOLE/SEEK_DATA bugs"

* tag 'xfs-4.12-fixes-2' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
  xfs: Move handling of missing page into one place in xfs_find_get_desired_pgoff()
  xfs: Fix off-by-in in loop termination in xfs_find_get_desired_pgoff()
  xfs: Fix missed holes in SEEK_HOLE implementation
  xfs: fix off-by-one on max nr_pages in xfs_find_get_desired_pgoff()
  xfs: fix unaligned access in xfs_btree_visit_blocks
  xfs: avoid mount-time deadlock in CoW extent recovery
  xfs: only return detailed fsmap info if the caller has CAP_SYS_ADMIN
  xfs: bad assertion for delalloc an extent that start at i_size
  xfs: fix warnings about unused stack variables
  xfs: BMAPX shouldn't barf on inline-format directories
  xfs: fix indlen accounting error on partial delalloc conversion

7 years agoipv4: add reference counting to metrics
Eric Dumazet [Thu, 25 May 2017 21:27:35 +0000 (14:27 -0700)]
ipv4: add reference counting to metrics

Andrey Konovalov reported crashes in ipv4_mtu()

I could reproduce the issue with KASAN kernels, between
10.246.7.151 and 10.246.7.152 :

1) 20 concurrent netperf -t TCP_RR -H 10.246.7.152 -l 1000 &

2) At the same time run following loop :
while :
do
 ip ro add 10.246.7.152 dev eth0 src 10.246.7.151 mtu 1500
 ip ro del 10.246.7.152 dev eth0 src 10.246.7.151 mtu 1500
done

Cong Wang attempted to add back rt->fi in commit
82486aa6f1b9 ("ipv4: restore rt->fi for reference counting")
but this proved to add some issues that were complex to solve.

Instead, I suggested to add a refcount to the metrics themselves,
being a standalone object (in particular, no reference to other objects)

I tried to make this patch as small as possible to ease its backport,
instead of being super clean. Note that we believe that only ipv4 dst
need to take care of the metric refcount. But if this is wrong,
this patch adds the basic infrastructure to extend this to other
families.

Many thanks to Julian Anastasov for reviewing this patch, and Cong Wang
for his efforts on this problem.

Fixes: 2860583fe840 ("ipv4: Kill rt->fi")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Andrey Konovalov <andreyknvl@google.com>
Reviewed-by: Julian Anastasov <ja@ssi.bg>
Acked-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: ethernet: ax88796: don't call free_irq without request_irq first
Uwe Kleine-König [Thu, 25 May 2017 20:54:53 +0000 (22:54 +0200)]
net: ethernet: ax88796: don't call free_irq without request_irq first

The function ax_init_dev (which is called only from the driver's .probe
function) calls free_irq in the error path without having requested the
irq in the first place. So drop the free_irq call in the error path.

Fixes: 825a2ff1896e ("AX88796 network driver")
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoip6_tunnel, ip6_gre: fix setting of DSCP on encapsulated packets
Peter Dawson [Thu, 25 May 2017 20:35:18 +0000 (06:35 +1000)]
ip6_tunnel, ip6_gre: fix setting of DSCP on encapsulated packets

This fix addresses two problems in the way the DSCP field is formulated
 on the encapsulating header of IPv6 tunnels.
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=195661

1) The IPv6 tunneling code was manipulating the DSCP field of the
 encapsulating packet using the 32b flowlabel. Since the flowlabel is
 only the lower 20b it was incorrect to assume that the upper 12b
 containing the DSCP and ECN fields would remain intact when formulating
 the encapsulating header. This fix handles the 'inherit' and
 'fixed-value' DSCP cases explicitly using the extant dsfield u8 variable.

2) The use of INET_ECN_encapsulate(0, dsfield) in ip6_tnl_xmit was
 incorrect and resulted in the DSCP value always being set to 0.

Commit 90427ef5d2a4 ("ipv6: fix flow labels when the traffic class
 is non-0") caused the regression by masking out the flowlabel
 which exposed the incorrect handling of the DSCP portion of the
 flowlabel in ip6_tunnel and ip6_gre.

Fixes: 90427ef5d2a4 ("ipv6: fix flow labels when the traffic class is non-0")
Signed-off-by: Peter Dawson <peter.a.dawson@boeing.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agosctp: fix ICMP processing if skb is non-linear
Davide Caratti [Thu, 25 May 2017 17:14:56 +0000 (19:14 +0200)]
sctp: fix ICMP processing if skb is non-linear

sometimes ICMP replies to INIT chunks are ignored by the client, even if
the encapsulated SCTP headers match an open socket. This happens when the
ICMP packet is carried by a paged skb: use skb_header_pointer() to read
packet contents beyond the SCTP header, so that chunk header and initiate
tag are validated correctly.

v2:
- don't use skb_header_pointer() to read the transport header, since
  icmp_socket_deliver() already puts these 8 bytes in the linear area.
- change commit message to make specific reference to INIT chunks.

Signed-off-by: Davide Caratti <dcaratti@redhat.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Vlad Yasevich <vyasevich@gmail.com>
Reviewed-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: llc: add lock_sock in llc_ui_bind to avoid a race condition
linzhang [Thu, 25 May 2017 06:07:18 +0000 (14:07 +0800)]
net: llc: add lock_sock in llc_ui_bind to avoid a race condition

There is a race condition in llc_ui_bind if two or more processes/threads
try to bind a same socket.

If more processes/threads bind a same socket success that will lead to
two problems, one is this action is not what we expected, another is
will lead to kernel in unstable status or oops(in my simple test case,
cause llc2.ko can't unload).

The current code is test SOCK_ZAPPED bit to avoid a process to
bind a same socket twice but that is can't avoid more processes/threads
try to bind a same socket at the same time.

So, add lock_sock in llc_ui_bind like others, such as llc_ui_connect.

Signed-off-by: Lin Zhang <xiaolou4617@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-block
Linus Torvalds [Fri, 26 May 2017 18:05:22 +0000 (11:05 -0700)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-block

Pull block fixes from Jens Axboe:
 "A collection of fixes that should go into this series. This contains:

   - A set of NVMe fixes, pulled from Christoph. This includes a set of
     fixes for the fiber channel bits from James Smart, rdma queue depth
     fix from Marta, controller removal fixes from Ming, and some more
     APST quirk updates from Andy.

   - A blk-mq debugfs fix from Bart, fixing a problem with the
     untangling of the sysfs and debugfs blk-mq bits that was added in
     this series.

   - Error code fix in add_partition() from Dan.

   - A small series of fixes for the new blk-throttle code from Shaohua"

* 'for-linus' of git://git.kernel.dk/linux-block: (21 commits)
  blk-mq: Only register debugfs attributes for blk-mq queues
  nvme: Quirk APST on Intel 600P/P3100 devices
  nvme: only setup block integrity if supported by the driver
  nvme: replace is_flags field in nvme_ctrl_ops with a flags field
  nvme-pci: consistencly use ctrl->device for logging
  partitions/msdos: FreeBSD UFS2 file systems are not recognized
  block: fix an error code in add_partition()
  blk-throttle: force user to configure all settings for io.low
  blk-throttle: respect 0 bps/iops settings for io.low
  blk-throttle: output some debug info in trace
  blk-throttle: add hierarchy support for latency target and idle time
  nvme_fc: remove extra controller reference taken on reconnect
  nvme_fc: correct nvme status set on abort
  nvme_fc: set logging level on resets/deletes
  nvme_fc: revise comment on teardown
  nvme_fc: Support ctrl_loss_tmo
  nvme_fc: get rid of local reconnect_delay
  blk-mq: remove blk_mq_abort_requeue_list()
  nvme: avoid to use blk_mq_abort_requeue_list()
  nvme: use blk_mq_start_hw_queues() in nvme_kill_queues()
  ...

7 years agoMerge tag 'pci-v4.12-fixes-1' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaa...
Linus Torvalds [Fri, 26 May 2017 17:51:18 +0000 (10:51 -0700)]
Merge tag 'pci-v4.12-fixes-1' of git://git./linux/kernel/git/helgaas/pci

Pull PCI fixes from Bjorn Helgaas:

 - fix PCI_ENDPOINT build error (merged for v4.12)

 - fix Switchtec driver (merged for v4.12)

 - fix imx6 config read timeouts, fallout from changing to non-postable
   reads

 - add PM "needs_resume" flag for i915 suspend issue

* tag 'pci-v4.12-fixes-1' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci:
  PCI/PM: Add needs_resume flag to avoid suspend complete optimization
  PCI: imx6: Fix config read timeout handling
  switchtec: Fix minor bug with partition ID register
  switchtec: Use new cdev_device_add() helper function
  PCI: endpoint: Make PCI_ENDPOINT depend on HAS_DMA

7 years agoMerge tag 'ceph-for-4.12-rc3' of git://github.com/ceph/ceph-client
Linus Torvalds [Fri, 26 May 2017 16:35:22 +0000 (09:35 -0700)]
Merge tag 'ceph-for-4.12-rc3' of git://github.com/ceph/ceph-client

Pul ceph fixes from Ilya Dryomov:
 "A bunch of make W=1 and static checker fixups, a RECONNECT_SEQ
  messenger patch from Zheng and Luis' fallocate fix"

* tag 'ceph-for-4.12-rc3' of git://github.com/ceph/ceph-client:
  ceph: check that the new inode size is within limits in ceph_fallocate()
  libceph: cleanup old messages according to reconnect seq
  libceph: NULL deref on crush_decode() error path
  libceph: fix error handling in process_one_ticket()
  libceph: validate blob_struct_v in process_one_ticket()
  libceph: drop version variable from ceph_monmap_decode()
  libceph: make ceph_msg_data_advance() return void
  libceph: use kbasename() and kill ceph_file_part()

7 years agoMerge tag 'mmc-v4.12-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc
Linus Torvalds [Fri, 26 May 2017 16:05:35 +0000 (09:05 -0700)]
Merge tag 'mmc-v4.12-rc2' of git://git./linux/kernel/git/ulfh/mmc

Pull MMC fixes from Ulf Hansson:
 "This contains fixes to make the WiFi work again for the ARM64 Hikey
  board.

  Together with a couple of DTS updates for the Hikey board we have also
  extended the mmc pwrseq_simple, to support a new power-off-delay-us DT
  property, as that was required to enable a graceful power off sequence
  for the WiFi chip"

* tag 'mmc-v4.12-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc:
  arm64: dts: hikey: Fix WiFi support
  arm64: dts: hi6220: Move board data from the dwmmc nodes to hikey dts
  arm64: dts: hikey: Add the SYS_5V and the VDD_3V3 regulators
  arm64: dts: hi6220: Move the fixed_5v_hub regulator to the hikey dts
  arm64: dts: hikey: Add clock for the pmic mfd
  mfd: dts: hi655x: Add clock binding for the pmic
  mmc: pwrseq_simple: Parse DTS for the power-off-delay-us property
  mmc: dt: pwrseq-simple: Invent power-off-delay-us

7 years agoMerge tag 'sound-4.12-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
Linus Torvalds [Fri, 26 May 2017 16:03:09 +0000 (09:03 -0700)]
Merge tag 'sound-4.12-rc3' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "This contains a few HD-audio device-specific quirks and an endianess
  fix for USB-audio, as well as the update of quirk model list document.
  All fixes are small and trivial.

  The document update could have been postponed, but it's a good thing
  for user and has absolutely zero risk of breakage, so included here"

* tag 'sound-4.12-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: hda - apply STAC_9200_DELL_M22 quirk for Dell Latitude D430
  ALSA: hda - Update the list of quirk models
  ALSA: hda - Provide dual-codecs model option for a few Realtek codecs
  ALSA: hda - Apply dual-codec quirk for MSI Z270-Gaming mobo
  ALSA: hda - No loopback on ALC299 codec
  ALSA: usb-audio: fix Amanero Combo384 quirk on big-endian hosts

7 years agoMerge tag 'drm-fixes-for-v4.12-rc3' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Fri, 26 May 2017 15:54:06 +0000 (08:54 -0700)]
Merge tag 'drm-fixes-for-v4.12-rc3' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
 "Not a whole lot happening here, a set of amdgpu fixes and one core
  deadlock fix, and some misc drivers fixes"

* tag 'drm-fixes-for-v4.12-rc3' of git://people.freedesktop.org/~airlied/linux:
  drm/amdgpu: fix null point error when rmmod amdgpu.
  drm/amd/powerplay: fix a signedness bugs
  drm/amdgpu: fix NULL pointer panic of emit_gds_switch
  drm/radeon: Unbreak HPD handling for r600+
  drm/amd/powerplay/smu7: disable mclk switching for high refresh rates
  drm/amd/powerplay/smu7: add vblank check for mclk switching (v2)
  drm/radeon/ci: disable mclk switching for high refresh rates (v2)
  drm/amdgpu/ci: disable mclk switching for high refresh rates (v2)
  drm/amdgpu: fix fundamental suspend/resume issue
  drm/gma500/psb: Actually use VBT mode when it is found
  drm: Fix deadlock retry loop in page_flip_ioctl
  drm: qxl: Delay entering atomic context during cursor update
  drm/radeon: Fix oops upon driver load on PowerXpress laptops

7 years agoPCI/msi: fix the pci_alloc_irq_vectors_affinity stub
Christoph Hellwig [Sat, 20 May 2017 16:59:54 +0000 (18:59 +0200)]
PCI/msi: fix the pci_alloc_irq_vectors_affinity stub

We need to return an error for any call that asks for MSI / MSI-X
vectors only, so that non-trivial fallback logic can work properly.

Also valid dev->irq and use the "correct" errno value based on feedback
from Linus.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reported-by: Steven Rostedt <rostedt@goodmis.org>
Fixes: aff17164 ("PCI: Provide sensible IRQ vector alloc/free routines")
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agoMerge branch 'nvme-4.12' of git://git.infradead.org/nvme into for-linus
Jens Axboe [Fri, 26 May 2017 15:11:19 +0000 (09:11 -0600)]
Merge branch 'nvme-4.12' of git://git.infradead.org/nvme into for-linus

Christoph writes:

"A couple of fixes for the next rc on the nvme front. Various FC fixes
from James, controller removal fixes from Ming (including a block layer
patch), a APST related device quirk from Andy, a RDMA fix for small
queue depth device from Marta, as well as fixes for the lack of
metadata support in non-PCIe drivers and the printk logging format from
me."

7 years agoblk-mq: Only register debugfs attributes for blk-mq queues
Bart Van Assche [Thu, 25 May 2017 23:38:06 +0000 (16:38 -0700)]
blk-mq: Only register debugfs attributes for blk-mq queues

The code in blk-mq-debugfs.c assumes that it is working on a blk-mq
queue and is not intended to work on a blk-sq queue. Hence only
register blk-mq debugfs attributes for blk-mq queues.

Fixes: commit 9c1051aacde8 ("blk-mq: untangle debugfs and sysfs")
Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
7 years agox86/timers: Move simple_udelay_calibration past init_hypervisor_platform
Jan Kiszka [Wed, 24 May 2017 18:04:41 +0000 (20:04 +0200)]
x86/timers: Move simple_udelay_calibration past init_hypervisor_platform

This ensures that adjustments to x86_platform done by the hypervisor
setup is already respected by this simple calibration.

The current user of this, introduced by 1b5aeebf3a92 ("x86/earlyprintk:
Add support for earlyprintk via USB3 debug port"), comes much later
into play.

Fixes: dd759d93f4dd ("x86/timers: Add simple udelay calibration")
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: http://lkml.kernel.org/r/5e89fe60-aab3-2c1c-aba8-32f8ad376189@siemens.com
7 years agonvme: Quirk APST on Intel 600P/P3100 devices
Andy Lutomirski [Wed, 24 May 2017 22:06:31 +0000 (15:06 -0700)]
nvme: Quirk APST on Intel 600P/P3100 devices

They have known firmware bugs.  A fix is apparently in the works --
once fixed firmware is available, someone from Intel (Hi, Keith!)
can adjust the quirk accordingly.

Cc: stable@vger.kernel.org # v4.11
Cc: Kai-Heng Feng <kai.heng.feng@canonical.com>
Cc: Mario Limonciello <mario_limonciello@dell.com>
Signed-off-by: Andy Lutomirski <luto@kernel.org>
Signed-off-by: Christoph Hellwig <hch@lst.de>
7 years agonvme: only setup block integrity if supported by the driver
Christoph Hellwig [Sat, 20 May 2017 13:14:45 +0000 (15:14 +0200)]
nvme: only setup block integrity if supported by the driver

Currently only the PCIe driver supports metadata, so we should not claim
integrity support for the other drivers.  This prevents nasty crashes
with targets that advertise metadata support on fabrics.

Also use the opportunity to factor out some code into a separate helper
that isn't even compiled if CONFIG_BLK_DEV_INTEGRITY is disabled.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <keith.busch@intel.com>
7 years agonvme: replace is_flags field in nvme_ctrl_ops with a flags field
Christoph Hellwig [Sat, 20 May 2017 13:14:44 +0000 (15:14 +0200)]
nvme: replace is_flags field in nvme_ctrl_ops with a flags field

So that we can have more flags for transport-specific behavior.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <keith.busch@intel.com>
7 years agonvme-pci: consistencly use ctrl->device for logging
Christoph Hellwig [Sat, 20 May 2017 13:14:43 +0000 (15:14 +0200)]
nvme-pci: consistencly use ctrl->device for logging

This is what most of the code already does and gives much more useful
prefixes than the device embedded in the pci_dev.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <keith.busch@intel.com>
7 years agoMerge branch 'drm-fixes-4.12' of git://people.freedesktop.org/~agd5f/linux into drm...
Dave Airlie [Fri, 26 May 2017 01:51:55 +0000 (11:51 +1000)]
Merge branch 'drm-fixes-4.12' of git://people.freedesktop.org/~agd5f/linux into drm-fixes

A bunch of bug fixes:
- Fix display flickering on some chips at high refresh rates
- suspend/resume fix
- hotplug fix
- a couple of segfault fixes for certain cases

* 'drm-fixes-4.12' of git://people.freedesktop.org/~agd5f/linux:
  drm/amdgpu: fix null point error when rmmod amdgpu.
  drm/amd/powerplay: fix a signedness bugs
  drm/amdgpu: fix NULL pointer panic of emit_gds_switch
  drm/radeon: Unbreak HPD handling for r600+
  drm/amd/powerplay/smu7: disable mclk switching for high refresh rates
  drm/amd/powerplay/smu7: add vblank check for mclk switching (v2)
  drm/radeon/ci: disable mclk switching for high refresh rates (v2)
  drm/amdgpu/ci: disable mclk switching for high refresh rates (v2)
  drm/amdgpu: fix fundamental suspend/resume issue

7 years agoMerge tag 'drm-misc-fixes-2017-05-25' of git://anongit.freedesktop.org/git/drm-misc...
Dave Airlie [Fri, 26 May 2017 01:51:28 +0000 (11:51 +1000)]
Merge tag 'drm-misc-fixes-2017-05-25' of git://anongit.freedesktop.org/git/drm-misc into drm-fixes

Core Changes:
- Don't drop vblank reference more than once in cases of ww retry (Daniel)

Driver Changes:
- radeon: Fix oops during radeon probe trying to reference wrong device (Lukas)
- qxl: Avoid sleeping while in atomic context on cursor update (Gabriel)
- gma500: Use VBT mode instead of pre-programmed mode for LVDS (Patrik)

Cc: Lukas Wunner <lukas@wunner.de>
Cc: Gabriel Krisman Bertazi <krisman@collabora.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
* tag 'drm-misc-fixes-2017-05-25' of git://anongit.freedesktop.org/git/drm-misc:
  drm/gma500/psb: Actually use VBT mode when it is found
  drm: Fix deadlock retry loop in page_flip_ioctl
  drm: qxl: Delay entering atomic context during cursor update
  drm/radeon: Fix oops upon driver load on PowerXpress laptops

7 years agobonding: Don't update slave->link until ready to commit
Nithin Sujir [Thu, 25 May 2017 02:45:17 +0000 (19:45 -0700)]
bonding: Don't update slave->link until ready to commit

In the loadbalance arp monitoring scheme, when a slave link change is
detected, the slave->link is immediately updated and slave_state_changed
is set. Later down the function, the rtnl_lock is acquired and the
changes are committed, updating the bond link state.

However, the acquisition of the rtnl_lock can fail. The next time the
monitor runs, since slave->link is already updated, it determines that
link is unchanged. This results in the bond link state permanently out
of sync with the slave link.

This patch modifies bond_loadbalance_arp_mon() to handle link changes
identical to bond_ab_arp_{inspect/commit}(). The new link state is
maintained in slave->new_link until we're ready to commit at which point
it's copied into slave->link.

NOTE: miimon_{inspect/commit}() has a more complex state machine
requiring the use of the bond_{propose,commit}_link_state() functions
which maintains the intermediate state in slave->link_new_state. The arp
monitors don't require that.

Testing: This bug is very easy to reproduce with the following steps.
1. In a loop, toggle a slave link of a bond slave interface.
2. In a separate loop, do ifconfig up/down of an unrelated interface to
create contention for rtnl_lock.
Within a few iterations, the bond link goes out of sync with the slave
link.

Signed-off-by: Nithin Nayak Sujir <nsujir@tintri.com>
Cc: Mahesh Bandewar <maheshb@google.com>
Cc: Jay Vosburgh <jay.vosburgh@canonical.com>
Acked-by: Mahesh Bandewar <maheshb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agotest_bpf: Add a couple of tests for BPF_JSGE.
David Daney [Wed, 24 May 2017 23:35:49 +0000 (16:35 -0700)]
test_bpf: Add a couple of tests for BPF_JSGE.

Some JITs can optimize comparisons with zero.  Add a couple of
BPF_JSGE tests against immediate zero.

Signed-off-by: David Daney <david.daney@cavium.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoMerge branch 'bpf-fixes'
David S. Miller [Thu, 25 May 2017 17:44:29 +0000 (13:44 -0400)]
Merge branch 'bpf-fixes'

Daniel Borkmann says:

====================
Various BPF fixes

Follow-up to fix incorrect pruning when alignment tracking is
in use and to properly clear regs after call to not leave stale
data behind, also a fix that adds bpf_clone_redirect to the
bpf_helper_changes_pkt_data helper and exposes correct map_flags
for lpm map into fdinfo. For details, please see individual
patches.

v1 -> v2:
  - Reworked first patch so that env->strict_alignment is the
    final indicator on whether we have to deal with strict
    alignment rather than having CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
    checks on various locations, so only checking env->strict_alignment
    is sufficient after that. Thanks for spotting, Dave!
  - Added patch 3 and 4.
  - Rest as is.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agobpf: add various verifier test cases
Daniel Borkmann [Wed, 24 May 2017 23:05:09 +0000 (01:05 +0200)]
bpf: add various verifier test cases

This patch adds various verifier test cases:

1) A test case for the pruning issue when tracking alignment
   is used.
2) Various PTR_TO_MAP_VALUE_OR_NULL tests to make sure pointer
   arithmetic turns such register into UNKNOWN_VALUE type.
3) Test cases for the special treatment of LD_ABS/LD_IND to
   make sure verifier doesn't break calling convention here.
   Latter is needed, since f.e. arm64 JIT uses r1 - r5 for
   storing temporary data, so they really must be marked as
   NOT_INIT.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agobpf: fix wrong exposure of map_flags into fdinfo for lpm
Daniel Borkmann [Wed, 24 May 2017 23:05:08 +0000 (01:05 +0200)]
bpf: fix wrong exposure of map_flags into fdinfo for lpm

trie_alloc() always needs to have BPF_F_NO_PREALLOC passed in via
attr->map_flags, since it does not support preallocation yet. We
check the flag, but we never copy the flag into trie->map.map_flags,
which is later on exposed into fdinfo and used by loaders such as
iproute2. Latter uses this in bpf_map_selfcheck_pinned() to test
whether a pinned map has the same spec as the one from the BPF obj
file and if not, bails out, which is currently the case for lpm
since it exposes always 0 as flags.

Also copy over flags in array_map_alloc() and stack_map_alloc().
They always have to be 0 right now, but we should make sure to not
miss to copy them over at a later point in time when we add actual
flags for them to use.

Fixes: b95a5c4db09b ("bpf: add a longest prefix match trie map implementation")
Reported-by: Jarno Rajahalme <jarno@covalent.io>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agobpf: add bpf_clone_redirect to bpf_helper_changes_pkt_data
Daniel Borkmann [Wed, 24 May 2017 23:05:07 +0000 (01:05 +0200)]
bpf: add bpf_clone_redirect to bpf_helper_changes_pkt_data

The bpf_clone_redirect() still needs to be listed in
bpf_helper_changes_pkt_data() since we call into
bpf_try_make_head_writable() from there, thus we need
to invalidate prior pkt regs as well.

Fixes: 36bbef52c7eb ("bpf: direct packet write and access for helpers for clsact progs")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agobpf: properly reset caller saved regs after helper call and ld_abs/ind
Daniel Borkmann [Wed, 24 May 2017 23:05:06 +0000 (01:05 +0200)]
bpf: properly reset caller saved regs after helper call and ld_abs/ind

Currently, after performing helper calls, we clear all caller saved
registers, that is r0 - r5 and fill r0 depending on struct bpf_func_proto
specification. The way we reset these regs can affect pruning decisions
in later paths, since we only reset register's imm to 0 and type to
NOT_INIT. However, we leave out clearing of other variables such as id,
min_value, max_value, etc, which can later on lead to pruning mismatches
due to stale data.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agobpf: fix incorrect pruning decision when alignment must be tracked
Daniel Borkmann [Wed, 24 May 2017 23:05:05 +0000 (01:05 +0200)]
bpf: fix incorrect pruning decision when alignment must be tracked

Currently, when we enforce alignment tracking on direct packet access,
the verifier lets the following program pass despite doing a packet
write with unaligned access:

  0: (61) r2 = *(u32 *)(r1 +76)
  1: (61) r3 = *(u32 *)(r1 +80)
  2: (61) r7 = *(u32 *)(r1 +8)
  3: (bf) r0 = r2
  4: (07) r0 += 14
  5: (25) if r7 > 0x1 goto pc+4
   R0=pkt(id=0,off=14,r=0) R1=ctx R2=pkt(id=0,off=0,r=0)
   R3=pkt_end R7=inv,min_value=0,max_value=1 R10=fp
  6: (2d) if r0 > r3 goto pc+1
   R0=pkt(id=0,off=14,r=14) R1=ctx R2=pkt(id=0,off=0,r=14)
   R3=pkt_end R7=inv,min_value=0,max_value=1 R10=fp
  7: (63) *(u32 *)(r0 -4) = r0
  8: (b7) r0 = 0
  9: (95) exit

  from 6 to 8:
   R0=pkt(id=0,off=14,r=0) R1=ctx R2=pkt(id=0,off=0,r=0)
   R3=pkt_end R7=inv,min_value=0,max_value=1 R10=fp
  8: (b7) r0 = 0
  9: (95) exit

  from 5 to 10:
   R0=pkt(id=0,off=14,r=0) R1=ctx R2=pkt(id=0,off=0,r=0)
   R3=pkt_end R7=inv,min_value=2 R10=fp
  10: (07) r0 += 1
  11: (05) goto pc-6
  6: safe                           <----- here, wrongly found safe
  processed 15 insns

However, if we enforce a pruning mismatch by adding state into r8
which is then being mismatched in states_equal(), we find that for
the otherwise same program, the verifier detects a misaligned packet
access when actually walking that path:

  0: (61) r2 = *(u32 *)(r1 +76)
  1: (61) r3 = *(u32 *)(r1 +80)
  2: (61) r7 = *(u32 *)(r1 +8)
  3: (b7) r8 = 1
  4: (bf) r0 = r2
  5: (07) r0 += 14
  6: (25) if r7 > 0x1 goto pc+4
   R0=pkt(id=0,off=14,r=0) R1=ctx R2=pkt(id=0,off=0,r=0)
   R3=pkt_end R7=inv,min_value=0,max_value=1
   R8=imm1,min_value=1,max_value=1,min_align=1 R10=fp
  7: (2d) if r0 > r3 goto pc+1
   R0=pkt(id=0,off=14,r=14) R1=ctx R2=pkt(id=0,off=0,r=14)
   R3=pkt_end R7=inv,min_value=0,max_value=1
   R8=imm1,min_value=1,max_value=1,min_align=1 R10=fp
  8: (63) *(u32 *)(r0 -4) = r0
  9: (b7) r0 = 0
  10: (95) exit

  from 7 to 9:
   R0=pkt(id=0,off=14,r=0) R1=ctx R2=pkt(id=0,off=0,r=0)
   R3=pkt_end R7=inv,min_value=0,max_value=1
   R8=imm1,min_value=1,max_value=1,min_align=1 R10=fp
  9: (b7) r0 = 0
  10: (95) exit

  from 6 to 11:
   R0=pkt(id=0,off=14,r=0) R1=ctx R2=pkt(id=0,off=0,r=0)
   R3=pkt_end R7=inv,min_value=2
   R8=imm1,min_value=1,max_value=1,min_align=1 R10=fp
  11: (07) r0 += 1
  12: (b7) r8 = 0
  13: (05) goto pc-7                <----- mismatch due to r8
  7: (2d) if r0 > r3 goto pc+1
   R0=pkt(id=0,off=15,r=15) R1=ctx R2=pkt(id=0,off=0,r=15)
   R3=pkt_end R7=inv,min_value=2
   R8=imm0,min_value=0,max_value=0,min_align=2147483648 R10=fp
  8: (63) *(u32 *)(r0 -4) = r0
  misaligned packet access off 2+15+-4 size 4

The reason why we fail to see it in states_equal() is that the
third test in compare_ptrs_to_packet() ...

  if (old->off <= cur->off &&
      old->off >= old->range && cur->off >= cur->range)
          return true;

... will let the above pass. The situation we run into is that
old->off <= cur->off (14 <= 15), meaning that prior walked paths
went with smaller offset, which was later used in the packet
access after successful packet range check and found to be safe
already.

For example: Given is R0=pkt(id=0,off=0,r=0). Adding offset 14
as in above program to it, results in R0=pkt(id=0,off=14,r=0)
before the packet range test. Now, testing this against R3=pkt_end
with 'if r0 > r3 goto out' will transform R0 into R0=pkt(id=0,off=14,r=14)
for the case when we're within bounds. A write into the packet
at offset *(u32 *)(r0 -4), that is, 2 + 14 -4, is valid and
aligned (2 is for NET_IP_ALIGN). After processing this with
all fall-through paths, we later on check paths from branches.
When the above skb->mark test is true, then we jump near the
end of the program, perform r0 += 1, and jump back to the
'if r0 > r3 goto out' test we've visited earlier already. This
time, R0 is of type R0=pkt(id=0,off=15,r=0), and we'll prune
that part because this time we'll have a larger safe packet
range, and we already found that with off=14 all further insn
were already safe, so it's safe as well with a larger off.
However, the problem is that the subsequent write into the packet
with 2 + 15 -4 is then unaligned, and not caught by the alignment
tracking. Note that min_align, aux_off, and aux_off_align were
all 0 in this example.

Since we cannot tell at this time what kind of packet access was
performed in the prior walk and what minimal requirements it has
(we might do so in the future, but that requires more complexity),
fix it to disable this pruning case for strict alignment for now,
and let the verifier do check such paths instead. With that applied,
the test cases pass and reject the program due to misalignment.

Fixes: d1174416747d ("bpf: Track alignment of register values in the verifier.")
Reference: http://patchwork.ozlabs.org/patch/761909/
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoarp: fixed -Wuninitialized compiler warning
Ihar Hrachyshka [Wed, 24 May 2017 22:19:35 +0000 (15:19 -0700)]
arp: fixed -Wuninitialized compiler warning

Commit 7d472a59c0e5ec117220a05de6b370447fb6cb66 ("arp: always override
existing neigh entries with gratuitous ARP") introduced a compiler
warning:

net/ipv4/arp.c:880:35: warning: 'addr_type' may be used uninitialized in
this function [-Wmaybe-uninitialized]

While the code logic seems to be correct and doesn't allow the variable
to be used uninitialized, and the warning is not consistently
reproducible, it's still worth fixing it for other people not to waste
time looking at the warning in case it pops up in the build environment.
Yes, compiler is probably at fault, but we will need to accommodate.

Fixes: 7d472a59c0e5 ("arp: always override existing neigh entries with gratuitous ARP")
Signed-off-by: Ihar Hrachyshka <ihrachys@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agotcp: avoid fastopen API to be used on AF_UNSPEC
Wei Wang [Wed, 24 May 2017 16:59:31 +0000 (09:59 -0700)]
tcp: avoid fastopen API to be used on AF_UNSPEC

Fastopen API should be used to perform fastopen operations on the TCP
socket. It does not make sense to use fastopen API to perform disconnect
by calling it with AF_UNSPEC. The fastopen data path is also prone to
race conditions and bugs when using with AF_UNSPEC.

One issue reported and analyzed by Vegard Nossum is as follows:
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Thread A:                            Thread B:
------------------------------------------------------------------------
sendto()
 - tcp_sendmsg()
     - sk_stream_memory_free() = 0
         - goto wait_for_sndbuf
     - sk_stream_wait_memory()
        - sk_wait_event() // sleep
          |                          sendto(flags=MSG_FASTOPEN, dest_addr=AF_UNSPEC)
  |                           - tcp_sendmsg()
  |                              - tcp_sendmsg_fastopen()
  |                                 - __inet_stream_connect()
  |                                    - tcp_disconnect() //because of AF_UNSPEC
  |                                       - tcp_transmit_skb()// send RST
  |                                    - return 0; // no reconnect!
  |                           - sk_stream_wait_connect()
  |                                 - sock_error()
  |                                    - xchg(&sk->sk_err, 0)
  |                                    - return -ECONNRESET
- ... // wake up, see sk->sk_err == 0
    - skb_entail() on TCP_CLOSE socket

If the connection is reopened then we will send a brand new SYN packet
after thread A has already queued a buffer. At this point I think the
socket internal state (sequence numbers etc.) becomes messed up.

When the new connection is closed, the FIN-ACK is rejected because the
sequence number is outside the window. The other side tries to
retransmit,
but __tcp_retransmit_skb() calls tcp_trim_head() on an empty skb which
corrupts the skb data length and hits a BUG() in copy_and_csum_bits().
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++

Hence, this patch adds a check for AF_UNSPEC in the fastopen data path
and return EOPNOTSUPP to user if such case happens.

Fixes: cf60af03ca4e7 ("tcp: Fast Open client - sendmsg(MSG_FASTOPEN)")
Reported-by: Vegard Nossum <vegard.nossum@oracle.com>
Signed-off-by: Wei Wang <weiwan@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: move somaxconn init from sysctl code
Roman Kapl [Wed, 24 May 2017 08:22:22 +0000 (10:22 +0200)]
net: move somaxconn init from sysctl code

The default value for somaxconn is set in sysctl_core_net_init(), but this
function is not called when kernel is configured without CONFIG_SYSCTL.

This results in the kernel not being able to accept TCP connections,
because the backlog has zero size. Usually, the user ends up with:
"TCP: request_sock_TCP: Possible SYN flooding on port 7. Dropping request.  Check SNMP counters."
If SYN cookies are not enabled the connection is rejected.

Before ef547f2ac16 (tcp: remove max_qlen_log), the effects were less
severe, because the backlog was always at least eight slots long.

Signed-off-by: Roman Kapl <roman.kapl@sysgo.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoInput: elan_i2c - ignore signals when finishing updating firmware
KT Liao [Tue, 23 May 2017 20:41:47 +0000 (13:41 -0700)]
Input: elan_i2c - ignore signals when finishing updating firmware

Use wait_for_completion_timeout() instead of
wait_for_completion_interruptible_timeout() to avoid stray signals ruining
firmware update. Our timeout is only 300 msec so we are fine simply letting
it expire in case device misbehaves.

Signed-off-by: KT Liao <kt.liao@emc.com.tw>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
7 years agoInput: elan_i2c - clear INT before resetting controller
KT Liao [Thu, 25 May 2017 17:06:21 +0000 (10:06 -0700)]
Input: elan_i2c - clear INT before resetting controller

Some old touchpad FWs need to have interrupt cleared before issuing reset
command after updating firmware. We clear interrupt by attempting to read
full report from the controller, and discarding any data read.

Signed-off-by: KT Liao <kt.liao@emc.com.tw>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>