dmaengine: mv_xor_v2: kill the tasklets upon exit

[ Upstream commit 8bbafed8dd5cfa81071b50ead5cb60367fdef3a9 ]

The mv_xor_v2 driver uses a tasklet, initialized during the probe()
routine. However, it forgets to cleanup the tasklet using
tasklet_kill() function during the remove() routine, which this patch
fixes. This prevents the tasklet from potentially running after the
module has been removed.

Fixes: 19a340b1a820 ("dmaengine: mv_xor_v2: new driver")

Signed-off-by: Hanna Hawa <hannah@marvell.com>
Reviewed-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

i2c: aspeed: Fix initial values of master and slave state

[ Upstream commit 517fde0eb5a8f46c54ba6e2c36e32563b23cb14f ]

This patch changes the order of enum aspeed_i2c_master_state and
enum aspeed_i2c_slave_state defines to make their initial value to
ASPEED_I2C_MASTER_INACTIVE and ASPEED_I2C_SLAVE_STOP respectively.
In case of multi-master use, if a slave data comes ahead of the
first master xfer, master_state starts from an invalid state so
this change fixes the issue.

Signed-off-by: Jae Hyun Yoo <jae.hyun.yoo@linux.intel.com>
Reviewed-by: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drivers/base: stop new probing during shutdown

[ Upstream commit 3297c8fc65af5d40501ea7cddff1b195cae57e4e ]

There is a race window in device_shutdown(), which may cause
-1. parent device shut down before child or
-2. no shutdown on a new probing device.

For 1st, taking the following scenario:
         device_shutdown                        new plugin device
  list_del_init(parent_dev);
  spin_unlock(list_lock);
                                                  device_add(child)
                                                  probe child
  shutdown parent_dev
       --> now child is on the tail of devices_kset

For 2nd, taking the following scenario:
         device_shutdown                        new plugin device
                                                  device_add(dev)
  device_lock(dev);
  ...
  device_unlock(dev);
                                                  probe dev
       --> now, the new occurred dev has no opportunity to shutdown

To fix this race issue, just prevent the new probing request. With this
logic, device_shutdown() is more similar to dpm_prepare().

Signed-off-by: Pingfan Liu <kernelfans@gmail.com>
Reviewed-by: Rafael J. Wysocki <rafael@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

KVM: arm/arm64: Fix vgic init race

[ Upstream commit 1d47191de7e15900f8fbfe7cccd7c6e1c2d7c31a ]

The vgic_init function can race with kvm_arch_vcpu_create() which does
not hold kvm_lock() and we therefore have no synchronization primitives
to ensure we're doing the right thing.

As the user is trying to initialize or run the VM while at the same time
creating more VCPUs, we just have to refuse to initialize the VGIC in
this case rather than silently failing with a broken VCPU.

Reviewed-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Christoffer Dall <christoffer.dall@arm.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

platform/x86: toshiba_acpi: Fix defined but not used build warnings

[ Upstream commit c2e2a618eb7104e18fdcf739d4d911563812a81c ]

Fix a build warning in toshiba_acpi.c when CONFIG_PROC_FS is not enabled
by marking the unused function as __maybe_unused.

../drivers/platform/x86/toshiba_acpi.c:1685:12: warning: 'version_proc_show' defined but not used [-Wunused-function]

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Azael Avalos <coproscefalo@gmail.com>
Cc: platform-driver-x86@vger.kernel.org
Cc: Andy Shevchenko <andy@infradead.org>
Signed-off-by: Darren Hart (VMware) <dvhart@infradead.org>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

s390/qeth: reset layer2 attribute on layer switch

[ Upstream commit 70551dc46ffa3555a0b5f3545b0cd87ab67fd002 ]

After the subdriver's remove() routine has completed, the card's layer
mode is undetermined again. Reflect this in the layer2 field.

If qeth_dev_layer2_store() hits an error after remove() was called, the
card _always_ requires a setup(), even if the previous layer mode is
requested again.
But qeth_dev_layer2_store() bails out early if the requested layer mode
still matches the current one. So unless we reset the layer2 field,
re-probing the card back to its previous mode is currently not possible.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

s390/qeth: fix race in used-buffer accounting

[ Upstream commit a702349a4099cd5a7bab0904689d8e0bf8dcd622 ]

By updating q->used_buffers only _after_ do_QDIO() has completed, there
is a potential race against the buffer's TX completion. In the unlikely
case that the TX completion path wins, qeth_qdio_output_handler() would
decrement the counter before qeth_flush_buffers() even incremented it.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ARM: dts: qcom: msm8974-hammerhead: increase load on l20 for sdhci

[ Upstream commit 03864e57770a9541e7ff3990bacf2d9a2fffcd5d ]

The kernel would not boot on the hammerhead hardware due to the
following error:

mmc0: Timeout waiting for hardware interrupt.
mmc0: sdhci: ============ SDHCI REGISTER DUMP ===========
mmc0: sdhci: Sys addr:  0x00000200 | Version:  0x00003802
mmc0: sdhci: Blk size:  0x00000200 | Blk cnt:  0x00000200
mmc0: sdhci: Argument:  0x00000000 | Trn mode: 0x00000023
mmc0: sdhci: Present:   0x03e80000 | Host ctl: 0x00000034
mmc0: sdhci: Power:     0x00000001 | Blk gap:  0x00000000
mmc0: sdhci: Wake-up:   0x00000000 | Clock:    0x00000007
mmc0: sdhci: Timeout:   0x0000000e | Int stat: 0x00000000
mmc0: sdhci: Int enab:  0x02ff900b | Sig enab: 0x02ff100b
mmc0: sdhci: AC12 err:  0x00000000 | Slot int: 0x00000000
mmc0: sdhci: Caps:      0x642dc8b2 | Caps_1:   0x00008007
mmc0: sdhci: Cmd:       0x00000c1b | Max curr: 0x00000000
mmc0: sdhci: Resp[0]:   0x00000c00 | Resp[1]:  0x00000000
mmc0: sdhci: Resp[2]:   0x00000000 | Resp[3]:  0x00000000
mmc0: sdhci: Host ctl2: 0x00000008
mmc0: sdhci: ADMA Err:  0x00000000 | ADMA Ptr: 0x70040220
mmc0: sdhci: ============================================
mmc0: Card stuck in wrong state! mmcblk0 card_busy_detect status: 0xe00
mmc0: cache flush error -110
mmc0: Reset 0x1 never completed.

This patch increases the load on l20 to 0.2 amps for the sdhci
and allows the device to boot normally.

Signed-off-by: Bhushan Shah <bshah@kde.org>
Signed-off-by: Brian Masney <masneyb@onstation.org>
Suggested-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Tested-by: Brian Masney <masneyb@onstation.org>
Signed-off-by: Andy Gross <andy.gross@linaro.org>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

arm64: dts: qcom: db410c: Fix Bluetooth LED trigger

[ Upstream commit e53db018315b7660bb7000a29e79faff2496c2c2 ]

Current LED trigger, 'bt', is not known/used by any existing driver.
Fix this by renaming it to 'bluetooth-power' trigger which is
controlled by the Bluetooth subsystem.

Fixes: 9943230c8860 ("arm64: dts: qcom: Add apq8016-sbc board LED's related device nodes")
Signed-off-by: Loic Poulain <loic.poulain@linaro.org>
Signed-off-by: Andy Gross <andy.gross@linaro.org>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

xen-netfront: fix queue name setting

[ Upstream commit 2d408c0d4574b01b9ed45e02516888bf925e11a9 ]

Commit f599c64fdf7d ("xen-netfront: Fix race between device setup and
open") changed the initialization order: xennet_create_queues() now
happens before we do register_netdev() so using netdev->name in
xennet_init_queue() is incorrect, we end up with the following in
/proc/interrupts:

 60:        139          0   xen-dyn    -event     eth%d-q0-tx
 61:        265          0   xen-dyn    -event     eth%d-q0-rx
 62:        234          0   xen-dyn    -event     eth%d-q1-tx
 63:          1          0   xen-dyn    -event     eth%d-q1-rx

and this looks ugly. Actually, using early netdev name (even when it's
already set) is also not ideal: nowadays we tend to rename eth devices
and queue name may end up not corresponding to the netdev name.

Use nodename from xenbus device for queue naming: this can't change in VM's
lifetime. Now /proc/interrupts looks like

 62:        202          0   xen-dyn    -event     device/vif/0-q0-tx
 63:        317          0   xen-dyn    -event     device/vif/0-q0-rx
 64:        262          0   xen-dyn    -event     device/vif/0-q1-tx
 65:         17          0   xen-dyn    -event     device/vif/0-q1-rx

Fixes: f599c64fdf7d ("xen-netfront: Fix race between device setup and open")
Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Reviewed-by: Ross Lagerwall <ross.lagerwall@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

nfp: avoid buffer leak when FW communication fails

[ Upstream commit 07300f774fec9519663a597987a4083225588be4 ]

After device is stopped we reset the rings by moving all free buffers
to positions [0, cnt - 2], and clear the position cnt - 1 in the ring.
We then proceed to clear the read/write pointers.  This means that if
we try to reset the ring again the code will assume that the next to
fill buffer is at position 0 and swap it with cnt - 1.  Since we
previously cleared position cnt - 1 it will lead to leaking the first
buffer and leaving ring in a bad state.

This scenario can only happen if FW communication fails, in which case
the ring will never be used again, so the fact it's in a bad state will
not be noticed.  Buffer leak is the only problem.  Don't try to move
buffers in the ring if the read/write pointers indicate the ring was
never used or have already been reset.

nfp_net_clear_config_and_disable() is now fully idempotent.

Found by code inspection, FW communication failures are very rare,
and reconfiguring a live device is not common either, so it's unlikely
anyone has ever noticed the leak.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

efi/arm: preserve early mapping of UEFI memory map longer for BGRT

[ Upstream commit 3ea86495aef2f6de26b7cb1599ba350dd6a0c521 ]

The BGRT code validates the contents of the table against the UEFI
memory map, and so it expects it to be mapped when the code runs.

On ARM, this is currently not the case, since we tear down the early
mapping after efi_init() completes, and only create the permanent
mapping in arm_enable_runtime_services(), which executes as an early
initcall, but still leaves a window where the UEFI memory map is not
mapped.

So move the call to efi_memmap_unmap() from efi_init() to
arm_enable_runtime_services().

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
[will: fold in EFI_MEMMAP attribute check from Ard]
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

reset: imx7: Fix always writing bits as 0

[ Upstream commit 26fce0557fa639fb7bbc33e31a57cff7df25c3a0 ]

Right now the only user of reset-imx7 is pci-imx6 and the
reset_control_assert and deassert calls on pciephy_reset don't toggle
the PCIEPHY_BTN and PCIEPHY_G_RST bits as expected. Fix this by writing
1 or 0 respectively.

The reference manual is not very clear regarding SRC_PCIEPHY_RCR but for
other registers like MIPIPHY and HSICPHY the bits are explicitly
documented as "1 means assert, 0 means deassert".

The values are still reversed for IMX7_RESET_PCIE_CTRL_APPS_EN.

Signed-off-by: Leonard Crestez <leonard.crestez@nxp.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

arm64: fix possible spectre-v1 write in ptrace_hbp_set_event()

[ Upstream commit 14d6e289a89780377f8bb09de8926d3c62d763cd ]

It's possible for userspace to control idx. Sanitize idx when using it
as an array index, to inhibit the potential spectre-v1 write gadget.

Found by smatch.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

wan/fsl_ucc_hdlc: use IS_ERR_VALUE() to check return value of qe_muram_alloc

[ Upstream commit fd800f646402c0f85547166b59ca065175928b7b ]

qe_muram_alloc return a unsigned long integer,which should not
compared with zero. check it using IS_ERR_VALUE() to fix this.

Fixes: c19b6d246a35 ("drivers/net: support hdlc function for QE-UCC")
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Smack: Fix handling of IPv4 traffic received by PF_INET6 sockets

[ Upstream commit 129a99890936766f4b69b9da7ed88366313a9210 ]

A socket which has sk_family set to PF_INET6 is able to receive not
only IPv6 but also IPv4 traffic (IPv4-mapped IPv6 addresses).

Prior to this patch, the smk_skb_to_addr_ipv6() could have been
called for socket buffers containing IPv4 packets, in result such
traffic was allowed.

Signed-off-by: Piotr Sawicki <p.sawicki2@partner.samsung.com>
Signed-off-by: Casey Schaufler <casey@schaufler-ca.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mac80211: restrict delayed tailroom needed decrement

[ Upstream commit 133bf90dbb8b873286f8ec2e81ba26e863114b8c ]

As explained in ieee80211_delayed_tailroom_dec(), during roam,
keys of the old AP will be destroyed and new keys will be
installed. Deletion of the old key causes
crypto_tx_tailroom_needed_cnt to go from 1 to 0 and the new key
installation causes a transition from 0 to 1.

Whenever crypto_tx_tailroom_needed_cnt transitions from 0 to 1,
we invoke synchronize_net(); the reason for doing this is to avoid
a race in the TX path as explained in increment_tailroom_need_count().
This synchronize_net() operation can be slow and can affect the station
roam time. To avoid this, decrementing the crypto_tx_tailroom_needed_cnt
is delayed for a while so that upon installation of new key the
transition would be from 1 to 2 instead of 0 to 1 and thereby
improving the roam time.

This is all correct for a STA iftype, but deferring the tailroom_needed
decrement for other iftypes may be unnecessary.

For example, let's consider the case of a 4-addr client connecting to
an AP for which AP_VLAN interface is also created, let the initial
value for tailroom_needed on the AP be 1.

* 4-addr client connects to the AP (AP: tailroom_needed = 1)
* AP will clear old keys, delay decrement of tailroom_needed count
* AP_VLAN is created, it takes the tailroom count from master
  (AP_VLAN: tailroom_needed = 1, AP: tailroom_needed = 1)
* Install new key for the station, assume key is plumbed in the HW,
  there won't be any change in tailroom_needed count on AP iface
* Delayed decrement of tailroom_needed count on AP
  (AP: tailroom_needed = 0, AP_VLAN: tailroom_needed = 1)

Because of the delayed decrement on AP iface, tailroom_needed count goes
out of sync between AP(master iface) and AP_VLAN(slave iface) and
there would be unnecessary tailroom created for the packets going
through AP_VLAN iface.

Also, WARN_ONs were observed while trying to bring down the AP_VLAN
interface:
(warn_slowpath_common) (warn_slowpath_null+0x18/0x20)
(warn_slowpath_null) (ieee80211_free_keys+0x114/0x1e4)
(ieee80211_free_keys) (ieee80211_del_virtual_monitor+0x51c/0x850)
(ieee80211_del_virtual_monitor) (ieee80211_stop+0x30/0x3c)
(ieee80211_stop) (__dev_close_many+0x94/0xb8)
(__dev_close_many) (dev_close_many+0x5c/0xc8)

Restricting delayed decrement to station interface alone fixes the problem
and it makes sense to do so because delayed decrement is done to improve
roam time which is applicable only for client devices.

Signed-off-by: Manikanta Pubbisetty <mpubbise@codeaurora.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

MIPS: jz4740: Bump zload address

[ Upstream commit c6ea7e9747318e5a6774995f4f8e3e0f7c0fa8ba ]

Having the zload address at 0x8060.0000 means the size of the
uncompressed kernel cannot be bigger than around 6 MiB, as it is
deflated at address 0x8001.0000.

This limit is too small; a kernel with some built-in drivers and things
like debugfs enabled will already be over 6 MiB in size, and so will
fail to extract properly.

To fix this, we bump the zload address from 0x8060.0000 to 0x8100.0000.

This is fine, as all the boards featuring Ingenic JZ SoCs have at least
32 MiB of RAM, and use u-boot or compatible bootloaders which won't
hardcode the load address but read it from the uImage's header.

Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Paul Burton <paul.burton@mips.com>
Patchwork: https://patchwork.linux-mips.org/patch/19787/
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: James Hogan <jhogan@kernel.org>
Cc: linux-mips@linux-mips.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ASoC: rt5514: Fix the issue of the delay volume applied

[ Upstream commit d96f8bd28cd0bae3e6702ae90df593628ef6906f ]

The patch fixes the issue of the delay volume applied.

Signed-off-by: Oder Chiou <oder_chiou@realtek.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

staging: bcm2835-camera: handle wait_for_completion_timeout return properly

[ Upstream commit 5b70084f6cbcd53f615433f9d216e01bd71de0bb ]

wait_for_completion_timeout returns unsigned long not int so a variable of
proper type is introduced. Further the check for <= 0 is ambiguous and
should be == 0 here indicating timeout.

Signed-off-by: Nicholas Mc Guire <hofrat@osadl.org>
Fixes: 7b3ad5abf027 ("staging: Import the BCM2835 MMAL-based V4L2 camera driver.")
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

staging: bcm2835-camera: fix timeout handling in wait_for_completion_timeout

[ Upstream commit b7afce51d95726a619743aaad8870db66dfa1479 ]

wait_for_completion_timeout returns unsigned long not int so a variable of
proper type is introduced. Further the check for <= 0 is ambiguous and should
be == 0 here indicating timeout which is the only error case so no additional
check needed here.

Signed-off-by: Nicholas Mc Guire <hofrat@osadl.org>
Fixes: 7b3ad5abf027 ("staging: Import the BCM2835 MMAL-based V4L2 camera driver.")
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

powerpc/powernv: opal_put_chars partial write fix

[ Upstream commit bd90284cc6c1c9e8e48c8eadd0c79574fcce0b81 ]

The intention here is to consume and discard the remaining buffer
upon error. This works if there has not been a previous partial write.
If there has been, then total_len is no longer total number of bytes
to copy. total_len is always "bytes left to copy", so it should be
added to written bytes.

This code may not be exercised any more if partial writes will not be
hit, but this is a small bugfix before a larger change.

Reviewed-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

KVM: arm/arm64: vgic: Fix possible spectre-v1 write in vgic_mmio_write_apr()

[ Upstream commit 6b8b9a48545e08345b8ff77c9fd51b1aebdbefb3 ]

It's possible for userspace to control n. Sanitize n when using it as an
array index, to inhibit the potential spectre-v1 write gadget.

Note that while it appears that n must be bound to the interval [0,3]
due to the way it is extracted from addr, we cannot guarantee that
compiler transformations (and/or future refactoring) will ensure this is
the case, and given this is a slow path it's better to always perform
the masking.

Found by smatch.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Christoffer Dall <christoffer.dall@arm.com>
Cc: Marc Zyngier <marc.zyngier@arm.com>
Cc: kvmarm@lists.cs.columbia.edu
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

nvme-rdma: unquiesce queues when deleting the controller

[ Upstream commit 90140624e8face94207003ac9a9d2a329b309d68 ]

If the controller is going away, we need to unquiesce the IO queues so
that all pending request can fail gracefully before moving forward with
controller deletion. Do that before we destroy the IO queues so
blk_cleanup_queue won't block in freeze.

Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

perf powerpc: Fix callchain ip filtering

[ Upstream commit c715fcfda5a08edabaa15508742be926b7ee51db ]

For powerpc64, redundant entries in the callchain are filtered out by
determining the state of the return address and the stack frame using
DWARF debug information.

For making these filtering decisions we must analyze the debug
information for the location corresponding to the program counter value,
i.e. the first entry in the callchain, and not the LR value; otherwise,
perf may filter out either the second or the third entry in the
callchain incorrectly.

This can be observed on a powerpc64le system running Fedora 27 as shown
below.

Case 1 - Attaching a probe at inet_pton+0x8 (binary offset 0x15af28).
         Return address is still in LR and a new stack frame is not yet
         allocated. The LR value, i.e. the second entry, should not be
 filtered out.

  # objdump -d /usr/lib64/libc-2.26.so | less
  ...
  000000000010eb10 <gaih_inet.constprop.7>:
  ...
    10fa48:       78 bb e4 7e     mr      r4,r23
    10fa4c:       0a 00 60 38     li      r3,10
    10fa50:       d9 b4 04 48     bl      15af28 <inet_pton+0x8>
    10fa54:       00 00 00 60     nop
    10fa58:       ac f4 ff 4b     b       10ef04 <gaih_inet.constprop.7+0x3f4>
  ...
  0000000000110450 <getaddrinfo>:
  ...
    1105a8:       54 00 ff 38     addi    r7,r31,84
    1105ac:       58 00 df 38     addi    r6,r31,88
    1105b0:       69 e5 ff 4b     bl      10eb18 <gaih_inet.constprop.7+0x8>
    1105b4:       78 1b 71 7c     mr      r17,r3
    1105b8:       50 01 7f e8     ld      r3,336(r31)
  ...
  000000000015af20 <inet_pton>:
    15af20:       0b 00 4c 3c     addis   r2,r12,11
    15af24:       e0 c1 42 38     addi    r2,r2,-15904
    15af28:       a6 02 08 7c     mflr    r0
    15af2c:       f0 ff c1 fb     std     r30,-16(r1)
    15af30:       f8 ff e1 fb     std     r31,-8(r1)
  ...

  # perf probe -x /usr/lib64/libc-2.26.so -a inet_pton+0x8
  # perf record -e probe_libc:inet_pton -g ping -6 -c 1 ::1
  # perf script

Before:

  ping  4507 [002] 514985.546540: probe_libc:inet_pton: (7fffa7dbaf28)
              7fffa7dbaf28 __GI___inet_pton+0x8 (/usr/lib64/libc-2.26.so)
              7fffa7d705b4 getaddrinfo+0x164 (/usr/lib64/libc-2.26.so)
                 13fb52d70 _init+0xbfc (/usr/bin/ping)
              7fffa7c836a0 generic_start_main.isra.0+0x140 (/usr/lib64/libc-2.26.so)
              7fffa7c83898 __libc_start_main+0xb8 (/usr/lib64/libc-2.26.so)
                         0 [unknown] ([unknown])

After:

  ping  4507 [002] 514985.546540: probe_libc:inet_pton: (7fffa7dbaf28)
              7fffa7dbaf28 __GI___inet_pton+0x8 (/usr/lib64/libc-2.26.so)
              7fffa7d6fa54 gaih_inet.constprop.7+0xf44 (/usr/lib64/libc-2.26.so)
              7fffa7d705b4 getaddrinfo+0x164 (/usr/lib64/libc-2.26.so)
                 13fb52d70 _init+0xbfc (/usr/bin/ping)
              7fffa7c836a0 generic_start_main.isra.0+0x140 (/usr/lib64/libc-2.26.so)
              7fffa7c83898 __libc_start_main+0xb8 (/usr/lib64/libc-2.26.so)
                         0 [unknown] ([unknown])

Case 2 - Attaching a probe at _int_malloc+0x180 (binary offset 0x9cf10).
         Return address in still in LR and a new stack frame has already
         been allocated but not used. The caller's caller, i.e. the third
 entry, is invalid and should be filtered out and not the second
 one.

  # objdump -d /usr/lib64/libc-2.26.so | less
  ...
  000000000009cd90 <_int_malloc>:
     9cd90:       17 00 4c 3c     addis   r2,r12,23
     9cd94:       70 a3 42 38     addi    r2,r2,-23696
     9cd98:       26 00 80 7d     mfcr    r12
     9cd9c:       f8 ff e1 fb     std     r31,-8(r1)
     9cda0:       17 00 e4 3b     addi    r31,r4,23
     9cda4:       d8 ff 61 fb     std     r27,-40(r1)
     9cda8:       78 23 9b 7c     mr      r27,r4
     9cdac:       1f 00 bf 2b     cmpldi  cr7,r31,31
     9cdb0:       f0 ff c1 fb     std     r30,-16(r1)
     9cdb4:       b0 ff c1 fa     std     r22,-80(r1)
     9cdb8:       78 1b 7e 7c     mr      r30,r3
     9cdbc:       08 00 81 91     stw     r12,8(r1)
     9cdc0:       11 ff 21 f8     stdu    r1,-240(r1)
     9cdc4:       4c 01 9d 41     bgt     cr7,9cf10 <_int_malloc+0x180>
     9cdc8:       20 00 a4 2b     cmpldi  cr7,r4,32
  ...
     9cf08:       00 00 00 60     nop
     9cf0c:       00 00 42 60     ori     r2,r2,0
     9cf10:       e4 06 ff 7b     rldicr  r31,r31,0,59
     9cf14:       40 f8 a4 7f     cmpld   cr7,r4,r31
     9cf18:       68 05 9d 41     bgt     cr7,9d480 <_int_malloc+0x6f0>
  ...
  000000000009e3c0 <tcache_init.part.4>:
  ...
     9e420:       40 02 80 38     li      r4,576
     9e424:       78 fb e3 7f     mr      r3,r31
     9e428:       71 e9 ff 4b     bl      9cd98 <_int_malloc+0x8>
     9e42c:       00 00 a3 2f     cmpdi   cr7,r3,0
     9e430:       78 1b 7e 7c     mr      r30,r3
  ...
  000000000009f7a0 <__libc_malloc>:
  ...
     9f8f8:       00 00 89 2f     cmpwi   cr7,r9,0
     9f8fc:       1c ff 9e 40     bne     cr7,9f818 <__libc_malloc+0x78>
     9f900:       c9 ea ff 4b     bl      9e3c8 <tcache_init.part.4+0x8>
     9f904:       00 00 00 60     nop
     9f908:       e8 90 22 e9     ld      r9,-28440(r2)
  ...

  # perf probe -x /usr/lib64/libc-2.26.so -a _int_malloc+0x180
  # perf record -e probe_libc:_int_malloc -g ./test-malloc
  # perf script

Before:

  test-malloc  6554 [009] 515975.797403: probe_libc:_int_malloc: (7fffa6e6cf10)
              7fffa6e6cf10 _int_malloc+0x180 (/usr/lib64/libc-2.26.so)
              7fffa6dd0000 [unknown] (/usr/lib64/libc-2.26.so)
              7fffa6e6f904 malloc+0x164 (/usr/lib64/libc-2.26.so)
              7fffa6e6f9fc malloc+0x25c (/usr/lib64/libc-2.26.so)
                  100006b4 main+0x38 (/home/testuser/test-malloc)
              7fffa6df36a0 generic_start_main.isra.0+0x140 (/usr/lib64/libc-2.26.so)
              7fffa6df3898 __libc_start_main+0xb8 (/usr/lib64/libc-2.26.so)
                         0 [unknown] ([unknown])

After:

  test-malloc  6554 [009] 515975.797403: probe_libc:_int_malloc: (7fffa6e6cf10)
              7fffa6e6cf10 _int_malloc+0x180 (/usr/lib64/libc-2.26.so)
              7fffa6e6e42c tcache_init.part.4+0x6c (/usr/lib64/libc-2.26.so)
              7fffa6e6f904 malloc+0x164 (/usr/lib64/libc-2.26.so)
              7fffa6e6f9fc malloc+0x25c (/usr/lib64/libc-2.26.so)
                  100006b4 main+0x38 (/home/sandipan/test-malloc)
              7fffa6df36a0 generic_start_main.isra.0+0x140 (/usr/lib64/libc-2.26.so)
              7fffa6df3898 __libc_start_main+0xb8 (/usr/lib64/libc-2.26.so)
                         0 [unknown] ([unknown])

Signed-off-by: Sandipan Das <sandipan@linux.ibm.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Maynard Johnson <maynard@us.ibm.com>
Cc: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Cc: Ravi Bangoria <ravi.bangoria@linux.vnet.ibm.com>
Cc: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Fixes: a60335ba3298 ("perf tools powerpc: Adjust callchain based on DWARF debug info")
Link: http://lkml.kernel.org/r/24bb726d91ed173aebc972ec3f41a2ef2249434e.1530724939.git.sandipan@linux.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ARM: exynos: Clear global variable on init error path

[ Upstream commit cd4806911cee3901bc2b5eb95603cf1958720b57 ]

For most of Exynos SoCs, Power Management Unit (PMU) address space is
mapped into global variable 'pmu_base_addr' very early when initializing
PMU interrupt controller.  A lot of other machine code depends on it so
when doing iounmap() on this address, clear the global as well to avoid
usage of invalid value (pointing to unmapped memory region).

Properly mapped PMU address space is a requirement for all other machine
code so this fix is purely theoretical.  Boot will fail immediately in
many other places after following this error path.

Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

fbdev: Distinguish between interlaced and progressive modes

[ Upstream commit 1ba0a59cea41ea05fda92daaf2a2958a2246b9cf ]

I discovered the problem when developing a frame buffer driver for the
PlayStation 2 (not yet merged), using the following video modes for the
PlayStation 3 in drivers/video/fbdev/ps3fb.c:

    }, {
        /* 1080if */
        "1080if", 50, 1920, 1080, 13468, 148, 484, 36, 4, 88, 5,
        FB_SYNC_BROADCAST, FB_VMODE_INTERLACED
    }, {
        /* 1080pf */
        "1080pf", 50, 1920, 1080, 6734, 148, 484, 36, 4, 88, 5,
        FB_SYNC_BROADCAST, FB_VMODE_NONINTERLACED
    },

In ps3fb_probe, the mode_option module parameter is used with fb_find_mode
but it can only select the interlaced variant of 1920x1080 since the loop
matching the modes does not take the difference between interlaced and
progressive modes into account.

In short, without the patch, progressive 1920x1080 cannot be chosen as a
mode_option parameter since fb_find_mode (falsely) thinks interlace is a
perfect match.

Signed-off-by: Fredrik Noring <noring@nocrew.org>
Cc: "Maciej W. Rozycki" <macro@linux-mips.org>
[b.zolnierkie: updated patch description]
Signed-off-by: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

video: fbdev: pxafb: clear allocated memory for video modes

[ Upstream commit b951d80aaf224b1f774e10def672f5e37488e4ee ]

When parsing the video modes from DT properties, make sure to zero out
memory before using it. This is important because not all fields in the mode
struct are explicitly initialized, even though they are used later on.

Fixes: 420a488278e86 ("video: fbdev: pxafb: initial devicetree conversion")
Reviewed-by: Robert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: Daniel Mack <daniel@zonque.org>
Signed-off-by: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

perf powerpc: Fix callchain ip filtering when return address is in a register

[ Upstream commit 9068533e4f470daf2b0f29c71d865990acd8826e ]

For powerpc64, perf will filter out the second entry in the callchain,
i.e. the LR value, if the return address of the function corresponding
to the probed location has already been saved on its caller's stack.

The state of the return address is determined using debug information.
At any point within a function, if the return address is already saved
somewhere, a DWARF expression can tell us about its location. If the
return address in still in LR only, no DWARF expression would exist.

Typically, the instructions in a function's prologue first copy the LR
value to R0 and then pushes R0 on to the stack. If LR has already been
copied to R0 but R0 is yet to be pushed to the stack, we can still get a
DWARF expression that says that the return address is in R0. This is
indicating that getting a DWARF expression for the return address does
not guarantee the fact that it has already been saved on the stack.

This can be observed on a powerpc64le system running Fedora 27 as shown
below.

  # objdump -d /usr/lib64/libc-2.26.so | less
  ...
  000000000015af20 <inet_pton>:
    15af20:       0b 00 4c 3c     addis   r2,r12,11
    15af24:       e0 c1 42 38     addi    r2,r2,-15904
    15af28:       a6 02 08 7c     mflr    r0
    15af2c:       f0 ff c1 fb     std     r30,-16(r1)
    15af30:       f8 ff e1 fb     std     r31,-8(r1)
    15af34:       78 1b 7f 7c     mr      r31,r3
    15af38:       78 23 83 7c     mr      r3,r4
    15af3c:       78 2b be 7c     mr      r30,r5
    15af40:       10 00 01 f8     std     r0,16(r1)
    15af44:       c1 ff 21 f8     stdu    r1,-64(r1)
    15af48:       28 00 81 f8     std     r4,40(r1)
  ...

  # readelf --debug-dump=frames-interp /usr/lib64/libc-2.26.so | less
  ...
  00027024 0000000000000024 00027028 FDE cie=00000000 pc=000000000015af20..000000000015af88
     LOC           CFA      r30   r31   ra
  000000000015af20 r1+0     u     u     u
  000000000015af34 r1+0     c-16  c-8   r0
  000000000015af48 r1+64    c-16  c-8   c+16
  000000000015af5c r1+0     c-16  c-8   c+16
  000000000015af78 r1+0     u     u
  ...

  # perf probe -x /usr/lib64/libc-2.26.so -a inet_pton+0x18
  # perf record -e probe_libc:inet_pton -g ping -6 -c 1 ::1
  # perf script

Before:

  ping  2829 [005] 512917.460174: probe_libc:inet_pton: (7fff7e2baf38)
              7fff7e2baf38 __GI___inet_pton+0x18 (/usr/lib64/libc-2.26.so)
              7fff7e2705b4 getaddrinfo+0x164 (/usr/lib64/libc-2.26.so)
                 12f152d70 _init+0xbfc (/usr/bin/ping)
              7fff7e1836a0 generic_start_main.isra.0+0x140 (/usr/lib64/libc-2.26.so)
              7fff7e183898 __libc_start_main+0xb8 (/usr/lib64/libc-2.26.so)
                         0 [unknown] ([unknown])

After:

  ping  2829 [005] 512917.460174: probe_libc:inet_pton: (7fff7e2baf38)
              7fff7e2baf38 __GI___inet_pton+0x18 (/usr/lib64/libc-2.26.so)
              7fff7e26fa54 gaih_inet.constprop.7+0xf44 (/usr/lib64/libc-2.26.so)
              7fff7e2705b4 getaddrinfo+0x164 (/usr/lib64/libc-2.26.so)
                 12f152d70 _init+0xbfc (/usr/bin/ping)
              7fff7e1836a0 generic_start_main.isra.0+0x140 (/usr/lib64/libc-2.26.so)
              7fff7e183898 __libc_start_main+0xb8 (/usr/lib64/libc-2.26.so)
                         0 [unknown] ([unknown])

Reported-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com>
Signed-off-by: Sandipan Das <sandipan@linux.ibm.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Maynard Johnson <maynard@us.ibm.com>
Cc: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Cc: Ravi Bangoria <ravi.bangoria@linux.vnet.ibm.com>
Cc: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/r/66e848a7bdf2d43b39210a705ff6d828a0865661.1530724939.git.sandipan@linux.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

fbdev/via: fix defined but not used warning

[ Upstream commit b6566b47a67e07fdca44cf51abb14e2fbe17d3eb ]

Fix a build warning in viafbdev.c when CONFIG_PROC_FS is not enabled
by marking the unused function as __maybe_unused.

../drivers/video/fbdev/via/viafbdev.c:1471:12: warning: 'viafb_sup_odev_proc_show' defined but not used [-Wunused-function]

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Signed-off-by: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

video: goldfishfb: fix memory leak on driver remove

[ Upstream commit 5958fde72d04e7b8c6de3669d1f794a90997e3eb ]

goldfish_fb_probe() allocates memory for fb, but goldfish_fb_remove() does
not have deallocation of fb, which leads to memory leak on probe/remove.

The patch adds deallocation into goldfish_fb_remove().

Found by Linux Driver Verification project (linuxtesting.org).

Signed-off-by: Anton Vasilyev <vasilyev@ispras.ru>
Cc: Aleksandar Markovic <aleksandar.markovic@mips.com>
Cc: Miodrag Dinic <miodrag.dinic@mips.com>
Cc: Goran Ferenc <goran.ferenc@mips.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

perf tools: Fix struct comm_str removal crash

[ Upstream commit 46b3722cc7765582354488da633aafffcb138458 ]

We occasionaly hit following assert failure in 'perf top', when processing the
/proc info in multiple threads.

  perf: ...include/linux/refcount.h:109: refcount_inc:
        Assertion `!(!refcount_inc_not_zero(r))' failed.

The gdb backtrace looks like this:

  [Switching to Thread 0x7ffff11ba700 (LWP 13749)]
  0x00007ffff50839fb in raise () from /lib64/libc.so.6
  (gdb)
  #0  0x00007ffff50839fb in raise () from /lib64/libc.so.6
  #1  0x00007ffff5085800 in abort () from /lib64/libc.so.6
  #2  0x00007ffff507c0da in __assert_fail_base () from /lib64/libc.so.6
  #3  0x00007ffff507c152 in __assert_fail () from /lib64/libc.so.6
  #4  0x0000000000535373 in refcount_inc (r=0x7fffdc009be0)
      at ...include/linux/refcount.h:109
  #5  0x00000000005354f1 in comm_str__get (cs=0x7fffdc009bc0)
      at util/comm.c:24
  #6  0x00000000005356bd in __comm_str__findnew (str=0x7fffd000b260 ":2",
      root=0xbed5c0 <comm_str_root>) at util/comm.c:72
  #7  0x000000000053579e in comm_str__findnew (str=0x7fffd000b260 ":2",
      root=0xbed5c0 <comm_str_root>) at util/comm.c:95
  #8  0x000000000053582e in comm__new (str=0x7fffd000b260 ":2",
      timestamp=0, exec=false) at util/comm.c:111
  #9  0x00000000005363bc in thread__new (pid=2, tid=2) at util/thread.c:57
  #10 0x0000000000523da0 in ____machine__findnew_thread (machine=0xbfde38,
      threads=0xbfdf28, pid=2, tid=2, create=true) at util/machine.c:457
  #11 0x0000000000523eb4 in __machine__findnew_thread (machine=0xbfde38,
  ...

The failing assertion is this one:

  REFCOUNT_WARN(!refcount_inc_not_zero(r), ...

The problem is that we keep global comm_str_root list, which
is accessed by multiple threads during the 'perf top' startup
and following 2 paths can race:

  thread 1:
    ...
    thread__new
      comm__new
        comm_str__findnew
          down_write(&comm_str_lock);
          __comm_str__findnew
            comm_str__get

  thread 2:
    ...
    comm__override or comm__free
      comm_str__put
        refcount_dec_and_test
          down_write(&comm_str_lock);
          rb_erase(&cs->rb_node, &comm_str_root);

Because thread 2 first decrements the refcnt and only after then it removes the
struct comm_str from the list, the thread 1 can find this object on the list
with refcnt equls to 0 and hit the assert.

This patch fixes the thread 1 __comm_str__findnew path, by ignoring objects
that already dropped the refcnt to 0. For the rest of the objects we take the
refcnt before comparing its name and release it afterwards with comm_str__put,
which can also release the object completely.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Lukasz Odzioba <lukasz.odzioba@intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Wang Nan <wangnan0@huawei.com>
Cc: kernel-team@lge.com
Link: http://lkml.kernel.org/r/20180720101740.GA27176@krava
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

fbdev: omapfb: off by one in omapfb_register_client()

[ Upstream commit 5ec1ec35b2979b59d0b33381e7c9aac17e159d16 ]

The omapfb_register_client[] array has OMAPFB_PLANE_NUM elements so the
> should be >= or we are one element beyond the end of the array.

Fixes: 8b08cf2b64f5 ("OMAP: add TI OMAP framebuffer driver")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Imre Deak <imre.deak@solidboot.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

perf tools: Synthesize GROUP_DESC feature in pipe mode

[ Upstream commit e8fedff1cc729fd227924305152ccc6f580e8c83 ]

Stephan reported, that pipe mode does not carry the group information
and thus the piped report won't display the grouped output for following
command:

  # perf record -e '{cycles,instructions,branches}' -a sleep 4 | perf report

It has no idea about the group setup, so it will display events
separately:

  # Overhead  Command          Shared Object             ...
  # ........  ...............  .......................
  #
       6.71%  swapper          [kernel.kallsyms]
       2.28%  offlineimap      libpython2.7.so.1.0
       0.78%  perf             [kernel.kallsyms]
  ...

Fix GROUP_DESC feature record to be synthesized in pipe mode, so the
report output is grouped if there are groups defined in record:

  #                 Overhead  Command          Shared    ...
  # ........................  ...............  .......
  #
       7.57%   0.16%   0.30%  swapper          [kernel
       1.87%   3.15%   2.46%  offlineimap      libpyth
       1.33%   0.00%   0.00%  perf             [kernel
  ...

Reported-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Tested-by: Stephane Eranian <eranian@google.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: David Carrillo-Cisneros <davidcc@google.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20180712135202.14774-1-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

gfs2: Don't reject a supposedly full bitmap if we have blocks reserved

[ Upstream commit e79e0e1428188b24c3b57309ffa54a33c4ae40c4 ]

Before this patch, you could get into situations like this:

1. Process 1 searches for X free blocks, finds them, makes a reservation
2. Process 2 searches for free blocks in the same rgrp, but now the
   bitmap is full because process 1's reservation is skipped over.
   So it marks the bitmap as GBF_FULL.
3. Process 1 tries to allocate blocks from its own reservation, but
   since the GBF_FULL bit is set, it skips over the rgrp and searches
   elsewhere, thus not using its own reservation.

This patch adds an additional check to allow processes to use their
own reservations.

Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

perf test: Fix subtest number when showing results

[ Upstream commit 9ef0112442bdddef5fb55adf20b3a5464b33de75 ]

Perf test 40 for example has several subtests numbered 1-4 when
displaying the start of the subtest. When the subtest results
are displayed the subtests are numbered 0-3.

Use this command to generate trace output:

  [root@s35lp76 perf]# ./perf test -Fv 40 2>/tmp/bpf1

Fix this by adjusting the subtest number when show the
subtest result.

Output before:

  [root@s35lp76 perf]# egrep '(^40\.[0-4]| subtest [0-4]:)' /tmp/bpf1
  40.1: Basic BPF filtering                                 :
  BPF filter subtest 0: Ok
  40.2: BPF pinning                                         :
  BPF filter subtest 1: Ok
  40.3: BPF prologue generation                             :
  BPF filter subtest 2: Ok
  40.4: BPF relocation checker                              :
  BPF filter subtest 3: Ok
  [root@s35lp76 perf]#

Output after:

  root@s35lp76 ~]# egrep '(^40\.[0-4]| subtest [0-4]:)' /tmp/bpf1
  40.1: Basic BPF filtering                                 :
  BPF filter subtest 1: Ok
  40.2: BPF pinning                                         :
  BPF filter subtest 2: Ok
  40.3: BPF prologue generation                             :
  BPF filter subtest 3: Ok
  40.4: BPF relocation checker                              :
  BPF filter subtest 4: Ok
  [root@s35lp76 ~]#

Signed-off-by: Thomas Richter <tmricht@linux.ibm.com>
Reviewed-by: Hendrik Brueckner <brueckner@linux.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Link: http://lkml.kernel.org/r/20180724134858.100644-1-tmricht@linux.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

media: ov5645: Supported external clock is 24MHz

[ Upstream commit 4adb0a0432f489c5eb802b33dae7737f69e6fd7a ]

The external clock frequency was set to 23.88MHz by mistake
because of a platform which cannot get closer to 24MHz.
The supported by the driver external clock is 24MHz so
set it correctly and also fix the values of the pixel
clock and link clock.
However allow 1% tolerance to the external clock as this
difference is small enough to be insignificant.

Signed-off-by: Todor Tomov <todor.tomov@linaro.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mtd/maps: fix solutionengine.c printk format warnings

[ Upstream commit 1d25e3eeed1d987404e2d2e451eebac8c15cecc1 ]

Fix 2 printk format warnings (this driver is currently only used by
arch/sh/) by using "%pap" instead of "%lx".

Fixes these build warnings:

../drivers/mtd/maps/solutionengine.c: In function 'init_soleng_maps':
../include/linux/kern_levels.h:5:18: warning: format '%lx' expects argument of type 'long unsigned int', but argument 2 has type 'resource_size_t' {aka 'unsigned int'} [-Wformat=]
../drivers/mtd/maps/solutionengine.c:62:54: note: format string is defined here
  printk(KERN_NOTICE "Solution Engine: Flash at 0x%08lx, EPROM at 0x%08lx\n",
                                                  ~~~~^
                                                  %08x
../include/linux/kern_levels.h:5:18: warning: format '%lx' expects argument of type 'long unsigned int', but argument 3 has type 'resource_size_t' {aka 'unsigned int'} [-Wformat=]
../drivers/mtd/maps/solutionengine.c:62:72: note: format string is defined here
  printk(KERN_NOTICE "Solution Engine: Flash at 0x%08lx, EPROM at 0x%08lx\n",
                                                                    ~~~~^
                                                                    %08x

Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Brian Norris <computersforpeace@gmail.com>
Cc: Boris Brezillon <boris.brezillon@bootlin.com>
Cc: Marek Vasut <marek.vasut@gmail.com>
Cc: Richard Weinberger <richard@nod.at>
Cc: linux-mtd@lists.infradead.org
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: Rich Felker <dalias@libc.org>
Cc: linux-sh@vger.kernel.org
Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

IB/ipoib: Fix error return code in ipoib_dev_init()

[ Upstream commit 99a7e2bf704d64c966dfacede1ba2d9b47cb676e ]

Fix to return a negative error code from the ipoib_neigh_hash_init()
error handling case instead of 0, as done elsewhere in this function.

Fixes: 515ed4f3aab4 ("IB/IPoIB: Separate control and data related initializations")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

block: allow max_discard_segments to be stacked

[ Upstream commit 42c9cdfe1e11e083dceb0f0c4977b758cf7403b9 ]

Set max_discard_segments to USHRT_MAX in blk_set_stacking_limits() so
that blk_stack_limits() can stack up this limit for stacked devices.

before:

$ cat /sys/block/nvme0n1/queue/max_discard_segments
256
$ cat /sys/block/dm-0/queue/max_discard_segments
1

after:

$ cat /sys/block/nvme0n1/queue/max_discard_segments
256
$ cat /sys/block/dm-0/queue/max_discard_segments
256

Fixes: 1e739730c5b9e ("block: optionally merge discontiguous discard bios into a single request")
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

IB/rxe: Drop QP0 silently

[ Upstream commit 536ca245c512aedfd84cde072d7b3ca14b6e1792 ]

According to "Annex A16: RDMA over Converged Ethernet (RoCE)":

A16.4.3 MANAGEMENT INTERFACES

As defined in the base specification, a special Queue Pair, QP0 is defined
solely for communication between subnet manager(s) and subnet management
agents. Since such an IB-defined subnet management architecture is outside
the scope of this annex, it follows that there is also no requirement that
a port which conforms to this annex be associated with a QP0. Thus, for
end nodes designed to conform to this annex, the concept of QP0 is
undefined and unused for any port connected to an Ethernet network.

CA16-8: A packet arriving at a RoCE port containing a BTH with the
destination QP field set to QP0 shall be silently dropped.

Signed-off-by: Zhu Yanjun <yanjun.zhu@oracle.com>
Acked-by: Moni Shoua <monis@mellanox.com>
Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

media: videobuf2-core: check for q->error in vb2_core_qbuf()

[ Upstream commit b509d733d337417bcb7fa4a35be3b9a49332b724 ]

The vb2_core_qbuf() function didn't check if q->error was set. It is
checked in __buf_prepare(), but that function isn't called if the buffer
was already prepared before with VIDIOC_PREPARE_BUF.

So check it at the start of vb2_core_qbuf() as well.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

MIPS: ath79: fix system restart

[ Upstream commit f8a7bfe1cb2c1ebfa07775c9c8ac0ad3ba8e5ff5 ]

This patch disables irq on reboot to fix hang issues that were observed
due to pending interrupts.

Signed-off-by: Felix Fietkau <nbd@nbd.name>
Signed-off-by: John Crispin <john@phrozen.org>
Signed-off-by: Paul Burton <paul.burton@mips.com>
Patchwork: https://patchwork.linux-mips.org/patch/19913/
Cc: James Hogan <jhogan@kernel.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

dmaengine: pl330: fix irq race with terminate_all

[ Upstream commit e49756544a21f5625b379b3871d27d8500764670 ]

In pl330_update() when checking if a channel has been aborted, the
channel's lock is not taken, only the overall pl330_dmac lock.  But in
pl330_terminate_all() the aborted flag (req_running==-1) is set under
the channel lock and not the pl330_dmac lock.

With threaded interrupts, this leads to a potential race:

    pl330_terminate_all         pl330_update
    -------------------         ------------
    lock channel
                                entry
    lock pl330
    _stop channel
    unlock pl330
                                lock pl330
                                check req_running != -1
    req_running = -1
                                _start channel

Signed-off-by: John Keeping <john@metanate.com>
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

media: tw686x: Fix oops on buffer alloc failure

[ Upstream commit 5a1a2f63d840dc2631505b607e11ff65ac1b7d3c ]

The error path currently calls tw686x_video_free() which requires
vc->dev to be initialized, causing a NULL dereference on uninitizalized
channels.

Fix this by setting the vc->dev fields for all the channels first.

Fixes: f8afaa8dbc0d ("[media] tw686x: Introduce an interface to support multiple DMA modes")

Signed-off-by: Krzysztof Ha?asa <khalasa@piap.pl>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

kbuild: add .DELETE_ON_ERROR special target

[ Upstream commit 9c2af1c7377a8a6ef86e5cabf80978f3dbbb25c0 ]

If Make gets a fatal signal while a shell is executing, it may delete
the target file that the recipe was supposed to update.  This is needed
to make sure that it is remade from scratch when Make is next run; if
Make is interrupted after the recipe has begun to write the target file,
it results in an incomplete file whose time stamp is newer than that
of the prerequisites files.  Make automatically deletes the incomplete
file on interrupt unless the target is marked .PRECIOUS.

The situation is just the same as when the shell fails for some reasons.
Usually when a recipe line fails, if it has changed the target file at
all, the file is corrupted, or at least it is not completely updated.
Yet the file’s time stamp says that it is now up to date, so the next
time Make runs, it will not try to update that file.

However, Make does not cater to delete the incomplete target file in
this case.  We need to add .DELETE_ON_ERROR somewhere in the Makefile
to request it.

scripts/Kbuild.include seems a suitable place to add it because it is
included from almost all sub-makes.

Please note .DELETE_ON_ERROR is not effective for phony targets.

The external module building should never ever touch the kernel tree.
The following recipe fails if include/generated/autoconf.h is missing.
However, include/config/auto.conf is not deleted since it is a phony
target.

 PHONY += include/config/auto.conf

 include/config/auto.conf:
         $(Q)test -e include/generated/autoconf.h -a -e $@ || (          \
         echo >&2;                                                       \
         echo >&2 "  ERROR: Kernel configuration is invalid.";           \
         echo >&2 "         include/generated/autoconf.h or $@ are missing.";\
         echo >&2 "         Run 'make oldconfig && make prepare' on kernel src to fix it."; \
         echo >&2 ;                                                      \
         /bin/false)

Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

clk: clk-fixed-factor: Clear OF_POPULATED flag in case of failure

[ Upstream commit f6dab4233d6b64d719109040503b567f71fbfa01 ]

Fixed factor clock has two initializations at of_clk_init() time
and during platform driver probe. Before of_clk_init() call,
node is marked as populated and so its probe never gets called.

During of_clk_init() fixed factor clock registration may fail if
any of its parent clock is not registered. In this case, it doesn't
get chance to retry registration from probe. Clear OF_POPULATED
flag if fixed factor clock registration fails so that clock
registration is attempted again from probe.

Signed-off-by: Rajan Vaja <rajan.vaja@xilinx.com>
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

clk: core: Potentially free connection id

[ Upstream commit 365f7a89c881e84f1ebc925f65f899d5d7ce547e ]

Patch "clk: core: Copy connection id" made it so that the connector id
'con_id' is kstrdup_const()ed to cater to drivers that pass non-constant
connection ids. The patch added the corresponding kfree_const to
__clk_free_clk(), but struct clk's can be freed also via __clk_put().
Add the kfree_const call to __clk_put() and add comments to both
functions to remind that the logic in them should be kept in sync.

Fixes: 253160a8ad06 ("clk: core: Copy connection id")
Signed-off-by: Mikko Perttunen <mperttunen@nvidia.com>
Reviewed-by: Leonard Crestez <leonard.crestez@nxp.com>
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

clk: imx6ul: fix missing of_node_put()

[ Upstream commit 11177e7a7aaef95935592072985526ebf0a3df43 ]

of_find_compatible_node() is returning a device node with refcount
incremented and must be explicitly decremented after the last use
which is right after the us in of_iomap() here.

Signed-off-by: Nicholas Mc Guire <hofrat@osadl.org>
Fixes: 787b4271a6a0 ("clk: imx: add imx6ul clk tree support")
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

gfs2: Special-case rindex for gfs2_grow

[ Upstream commit 776125785a87ff05d49938bd5b9f336f2a05bff6 ]

To speed up the common case of appending to a file,
gfs2_write_alloc_required presumes that writing beyond the end of a file
will always require additional blocks to be allocated.  This assumption
is incorrect for preallocates files, but there are no negative
consequences as long as *some* space is still left on the filesystem.

One special file that always has some space preallocated beyond the end
of the file is the rindex: when growing a filesystem, gfs2_grow adds one
or more new resource groups and appends records describing those
resource groups to the rindex; the preallocated space ensures that this
is always possible.

However, when a filesystem is completely full, gfs2_write_alloc_required
will indicate that an additional allocation is required, and appending
the next record to the rindex will fail even though space for that
record has already been preallocated.  To fix that, skip the incorrect
optimization in gfs2_write_alloc_required, but for the rindex only.
Other writes to preallocated space beyond the end of the file are still
allowed to fail on completely full filesystems.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Reviewed-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

amd-xgbe: use dma_mapping_error to check map errors

[ Upstream commit b24dbfe9ce03d9f83306616f22fb0e04e8960abe ]

The dma_mapping_error() returns true or false, but we want
to return -ENOMEM if there was an error.

Fixes: 174fd2597b0b ("amd-xgbe: Implement split header receive support")
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

xfrm: fix 'passing zero to ERR_PTR()' warning

[ Upstream commit 934ffce1343f22ed5e2d0bd6da4440f4848074de ]

Fix a static code checker warning:

  net/xfrm/xfrm_policy.c:1836 xfrm_resolve_and_create_bundle() warn: passing zero to 'ERR_PTR'

xfrm_tmpl_resolve return 0 just means no xdst found, return NULL
instead of passing zero to ERR_PTR.

Fixes: d809ec895505 ("xfrm: do not assume that template resolving always returns xfrms")
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ALSA: usb-audio: Fix multiple definitions in AU0828_DEVICE() macro

[ Upstream commit bd1cd0eb2ce9141100628d476ead4de485501b29 ]

AU0828_DEVICE() macro in quirks-table.h uses USB_DEVICE_VENDOR_SPEC()
for expanding idVendor and idProduct fields.  However, the latter
macro adds also match_flags and bInterfaceClass, which are different
from the values AU0828_DEVICE() macro sets after that.

For fixing them, just expand idVendor and idProduct fields manually in
AU0828_DEVICE().

This fixes sparse warnings like:
  sound/usb/quirks-table.h:2892:1: warning: Initializer entry defined twice

Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ALSA: msnd: Fix the default sample sizes

[ Upstream commit 7c500f9ea139d0c9b80fdea5a9c911db3166ea54 ]

The default sample sizes set by msnd driver are bogus; it sets ALSA
PCM format, not the actual bit width.

Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

iommu/io-pgtable-arm-v7s: Abort allocation when table address overflows the PTE

[ Upstream commit 29859aeb8a6ea17ba207933a81b6b77b4d4df81a ]

When run on a 64-bit system in selftest, the v7s driver may obtain page
table with physical addresses larger than 32-bit. Level-2 tables are 1KB
and are are allocated with slab, which doesn't accept the GFP_DMA32
flag. Currently map() truncates the address written in the PTE, causing
iova_to_phys() or unmap() to access invalid memory. Kasan reports it as
a use-after-free. To avoid any nasty surprise, test if the physical
address fits in a PTE before returning a new table. 32-bit systems,
which are the main users of this page table format, shouldn't see any
difference.

Signed-off-by: Jean-Philippe Brucker <jean-philippe.brucker@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

iommu/arm-smmu-v3: sync the OVACKFLG to PRIQ consumer register

[ Upstream commit 0d535967ac658966c6ade8f82b5799092f7d5441 ]

When PRI queue occurs overflow, driver should update the OVACKFLG to
the PRIQ consumer register, otherwise subsequent PRI requests will not
be processed.

Cc: Will Deacon <will.deacon@arm.com>
Cc: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Miao Zhong <zhongmiao@hisilicon.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usb: dwc3: change stream event enable bit back to 13

[ Upstream commit 9a7faac3650216112e034b157289bf1a48a99e2d ]

Commit ff3f0789b3dc ("usb: dwc3: use BIT() macro where possible")
changed DWC3_DEPCFG_STREAM_EVENT_EN from bit 13 to bit 12.

Spotted this cleanup typo while looking at diffs between 4.9.35 and
4.14.16 for a separate issue.

Fixes: ff3f0789b3dc ("usb: dwc3: use BIT() macro where possible")
Signed-off-by: Erich E. Hoover <ehoover@sweptlaser.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

hv/netvsc: Fix NULL dereference at single queue mode fallback

commit b19b46346f483ae055fa027cb2d5c2ca91484b91 upstream.

The recent commit 916c5e1413be ("hv/netvsc: fix handling of fallback
to single queue mode") tried to fix the fallback behavior to a single
queue mode, but it changed the function to return zero incorrectly,
while the function should return an object pointer.  Eventually this
leads to a NULL dereference at the callers that expect non-NULL
value.

Fix it by returning the proper net_device object.

Fixes: 916c5e1413be ("hv/netvsc: fix handling of fallback to single queue mode")
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Cc: Alakesh Haloi <alakeshh@amazon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tcp: really ignore MSG_ZEROCOPY if no SO_ZEROCOPY

[ Upstream commit 5cf4a8532c992bb22a9ecd5f6d93f873f4eaccc2 ]

According to the documentation in msg_zerocopy.rst, the SO_ZEROCOPY
flag was introduced because send(2) ignores unknown message flags and
any legacy application which was accidentally passing the equivalent of
MSG_ZEROCOPY earlier should not see any new behaviour.

Before commit f214f915e7db ("tcp: enable MSG_ZEROCOPY"), a send(2) call
which passed the equivalent of MSG_ZEROCOPY without setting SO_ZEROCOPY
would succeed.  However, after that commit, it fails with -ENOBUFS.  So
it appears that the SO_ZEROCOPY flag fails to fulfill its intended
purpose.  Fix it.

Fixes: f214f915e7db ("tcp: enable MSG_ZEROCOPY")
Signed-off-by: Vincent Whitchurch <vincent.whitchurch@axis.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

erspan: return PACKET_REJECT when the appropriate tunnel is not found

[ Upstream commit 5a64506b5c2c3cdb29d817723205330378075448 ]

If erspan tunnel hasn't been established, we'd better send icmp port
unreachable message after receive erspan packets.

Fixes: 84e54fe0a5ea ("gre: introduce native tunnel support for ERSPAN")
Cc: William Tu <u9012063@gmail.com>
Signed-off-by: Haishuang Yan <yanhaishuang@cmss.chinamobile.com>
Acked-by: William Tu <u9012063@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

erspan: fix error handling for erspan tunnel

[ Upstream commit 51dc63e3911fbb1f0a7a32da2fe56253e2040ea4 ]

When processing icmp unreachable message for erspan tunnel, tunnel id
should be erspan_net_id instead of ipgre_net_id.

Fixes: 84e54fe0a5ea ("gre: introduce native tunnel support for ERSPAN")
Cc: William Tu <u9012063@gmail.com>
Signed-off-by: Haishuang Yan <yanhaishuang@cmss.chinamobile.com>
Acked-by: William Tu <u9012063@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net/tls: Set count of SG entries if sk_alloc_sg returns -ENOSPC

[ Upstream commit 52ea992cfac357b73180d5c051dca43bc8d20c2a ]

tls_sw_sendmsg() allocates plaintext and encrypted SG entries using
function sk_alloc_sg(). In case the number of SG entries hit
MAX_SKB_FRAGS, sk_alloc_sg() returns -ENOSPC and sets the variable for
current SG index to '0'. This leads to calling of function
tls_push_record() with 'sg_encrypted_num_elem = 0' and later causes
kernel crash. To fix this, set the number of SG elements to the number
of elements in plaintext/encrypted SG arrays in case sk_alloc_sg()
returns -ENOSPC.

Fixes: 3c4d7559159b ("tls: kernel TLS support")
Signed-off-by: Vakul Garg <vakul.garg@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net/mlx5: E-Switch, Fix memory leak when creating switchdev mode FDB tables

[ Upstream commit c88a026e01219488e745f4f0267fd76c2bb68421 ]

The memory allocated for the slow path table flow group input structure
was not freed upon successful return, fix that.

Fixes: 1967ce6ea5c8 ("net/mlx5: E-Switch, Refactor fast path FDB table creation in switchdev mode")
Signed-off-by: Raed Salem <raeds@mellanox.com>
Reviewed-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net/mlx5: Fix debugfs cleanup in the device init/remove flow

[ Upstream commit 5df816e7f43f1297c40021ef17ec6e722b45c82f ]

When initializing the device (procedure init_one), the driver
calls mlx5_pci_init to perform pci initialization. As part of this
initialization, mlx5_pci_init creates a debugfs directory.
If this creation fails, init_one aborts, returning failure to
the caller (which is the probe method caller).

The main reason for such a failure to occur is if the debugfs
directory already exists. This can happen if the last time
mlx5_pci_close was called, debugfs_remove (silently) failed due
to the debugfs directory not being empty.

Guarantee that such a debugfs_remove failure will not occur by
instead calling debugfs_remove_recursive in procedure mlx5_pci_close.

Fixes: 59211bd3b632 ("net/mlx5: Split the load/unload flow into hardware and software flows")
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Reviewed-by: Daniel Jurgens <danielj@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net/mlx5: Check for error in mlx5_attach_interface

[ Upstream commit 47bc94b82291e007da61ee1b3d18c77871f3e158 ]

Currently, mlx5_attach_interface does not check for error
after calling intf->attach or intf->add. When these two calls
fails, the client is not initialized and will cause issues such as
kernel panic on invalid address in the teardown path (mlx5_detach_interface)

Fixes: 737a234bb638 ("net/mlx5: Introduce attach/detach to interface API")
Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

rds: fix two RCU related problems

[ Upstream commit cc4dfb7f70a344f24c1c71e298deea0771dadcb2 ]

When a rds sock is bound, it is inserted into the bind_hash_table
which is protected by RCU. But when releasing rds sock, after it
is removed from this hash table, it is freed immediately without
respecting RCU grace period. This could cause some use-after-free
as reported by syzbot.

Mark the rds sock with SOCK_RCU_FREE before inserting it into the
bind_hash_table, so that it would be always freed after a RCU grace
period.

The other problem is in rds_find_bound(), the rds sock could be
freed in between rhashtable_lookup_fast() and rds_sock_addref(),
so we need to extend RCU read lock protection in rds_find_bound()
to close this race condition.

Reported-and-tested-by: syzbot+8967084bcac563795dc6@syzkaller.appspotmail.com
Reported-by: syzbot+93a5839deb355537440f@syzkaller.appspotmail.com
Cc: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Cc: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Cc: rds-devel@oss.oracle.com
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oarcle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: qca_spi: Fix race condition in spi transfers

[ Upstream commit e65a9e480e91ddf9e15155454d370cead64689c8 ]

With performance optimization the spi transfer and messages of basic
register operations like qcaspi_read_register moved into the private
driver structure. But they weren't protected against mutual access
(e.g. between driver kthread and ethtool). So dumping the QCA7000
registers via ethtool during network traffic could make spi_sync
hang forever, because the completion in spi_message is overwritten.

So revert the optimization completely.

Fixes: 291ab06ecf676 ("net: qualcomm: new Ethernet over SPI driver for QCA700")
Signed-off-by: Stefan Wahren <stefan.wahren@i2se.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net/mlx5: Fix use-after-free in self-healing flow

[ Upstream commit 76d5581c870454be5f1f1a106c57985902e7ea20 ]

When the mlx5 health mechanism detects a problem while the driver
is in the middle of init_one or remove_one, the driver needs to prevent
the health mechanism from scheduling future work; if future work
is scheduled, there is a problem with use-after-free: the system WQ
tries to run the work item (which has been freed) at the scheduled
future time.

Prevent this by disabling work item scheduling in the health mechanism
when the driver is in the middle of init_one() or remove_one().

Fixes: e126ba97dba9 ("mlx5: Add driver for Mellanox Connect-IB adapters")
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Reviewed-by: Feras Daoud <ferasda@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

be2net: Fix memory leak in be_cmd_get_profile_config()

[ Upstream commit 9d7f19dc4673fbafebfcbf30eb90e09fa7d1c037 ]

DMA allocated memory is lost in be_cmd_get_profile_config() when we
call it with non-NULL port_res parameter.

Signed-off-by: Petr Oros <poros@redhat.com>
Reviewed-by: Ivan Vecera <ivecera@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Linux 4.14.71

mm: get rid of vmacache_flush_all() entirely

commit 7a9cdebdcc17e426fb5287e4a82db1dfe86339b2 upstream.

Jann Horn points out that the vmacache_flush_all() function is not only
potentially expensive, it's buggy too.  It also happens to be entirely
unnecessary, because the sequence number overflow case can be avoided by
simply making the sequence number be 64-bit.  That doesn't even grow the
data structures in question, because the other adjacent fields are
already 64-bit.

So simplify the whole thing by just making the sequence number overflow
case go away entirely, which gets rid of all the complications and makes
the code faster too.  Win-win.

[ Oleg Nesterov points out that the VMACACHE_FULL_FLUSHES statistics
  also just goes away entirely with this ]

Reported-by: Jann Horn <jannh@google.com>
Suggested-by: Will Deacon <will.deacon@arm.com>
Acked-by: Davidlohr Bueso <dave@stgolabs.net>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

autofs: fix autofs_sbi() does not check super block type

commit 0633da48f0793aeba27f82d30605624416723a91 upstream.

autofs_sbi() does not check the superblock magic number to verify it has
been given an autofs super block.

Backport Note: autofs4 has been renamed to autofs upstream. As a result
the upstream patch does not apply cleanly onto 4.14.y.

Link: http://lkml.kernel.org/r/153475422934.17131.7563724552005298277.stgit@pluto.themaw.net
Reported-by: <syzbot+87c3c541582e56943277@syzkaller.appspotmail.com>
Signed-off-by: Ian Kent <raven@themaw.net>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Zubin Mithra <zsm@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tuntap: fix use after free during release

commit 7063efd33bb15abc0160347f89eb5aba6b7d000e upstream.

After commit b196d88aba8a ("tun: fix use after free for ptr_ring") we
need clean up tx ring during release(). But unfortunately, it tries to
do the cleanup blindly after socket were destroyed which will lead
another use-after-free. Fix this by doing the cleanup before dropping
the last reference of the socket in __tun_detach().

Backport Note :-
Upstream commit moves the ptr_ring_cleanup call from tun_chr_close to
__tun_detach. Upstream applied that patch after replacing skb_array with
ptr_ring. This patch moves the skb_array_cleanup call from
tun_chr_close to __tun_detach.

Reported-by: Andrei Vagin <avagin@virtuozzo.com>
Acked-by: Andrei Vagin <avagin@virtuozzo.com>
Fixes: b196d88aba8a ("tun: fix use after free for ptr_ring")
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Zubin Mithra <zsm@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tun: fix use after free for ptr_ring

commit b196d88aba8ac72b775137854121097f4c4c6862 upstream.

We used to initialize ptr_ring during TUNSETIFF, this is because its
size depends on the tx_queue_len of netdevice. And we try to clean it
up when socket were detached from netdevice. A race were spotted when
trying to do uninit during a read which will lead a use after free for
pointer ring. Solving this by always initialize a zero size ptr_ring
in open() and do resizing during TUNSETIFF, and then we can safely do
cleanup during close(). With this, there's no need for the workaround
that was introduced by commit 4df0bfc79904 ("tun: fix a memory leak
for tfile->tx_array").

Backport Note :-
Comparison with the upstream patch:
[1] A "semantic revert" of the changes made in
    4df0bfc799("tun: fix a memory leak for tfile->tx_array").
        4df0bfc799 was applied upstream, and then skb array was changed
to use ptr_ring. The upstream patch then removes the changes introduced
by 4df0bfc799. This backport does the same; "revert" the changes
made by 4df0bfc799.
[2] xdp_rxq_info_unreg() being called in relevant locations
        As xdp_rxq_info related patches are not present in 4.14, these
changes are not needed in the backport.
[3] An instance of ptr_ring_init needs to be replaced by skb_array_init
Inside tun_attach()
[4] ptr_ring_cleanup needs to be replaced by skb_array_cleanup
Inside tun_chr_close()

Note that the backport for 7063efd33b ("tuntap: fix use after free during release")
needs to be applied on top of this patch.

Reported-by: syzbot+e8b902c3c3fadf0a9dba@syzkaller.appspotmail.com
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Michael S. Tsirkin <mst@redhat.com>
Fixes: 1576d9860599 ("tun: switch to use skb array for tx")
Signed-off-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Zubin Mithra <zsm@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mtd: ubi: wl: Fix error return code in ubi_wl_init()

commit 7233982ade15eeac05c6f351e8d347406e6bcd2f upstream.

Fix to return error code -ENOMEM from the kmem_cache_alloc() error
handling case instead of 0, as done elsewhere in this function.

Fixes: f78e5623f45b ("ubi: fastmap: Erase outdated anchor PEBs during
attach")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Reviewed-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
Cc: Ben Hutchings <ben.hutchings@codethink.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ip: frags: fix crash in ip_do_fragment()

commit 5d407b071dc369c26a38398326ee2be53651cfe4 upstream

A kernel crash occurrs when defragmented packet is fragmented
in ip_do_fragment().
In defragment routine, skb_orphan() is called and
skb->ip_defrag_offset is set. but skb->sk and
skb->ip_defrag_offset are same union member. so that
frag->sk is not NULL.
Hence crash occurrs in skb->sk check routine in ip_do_fragment() when
defragmented packet is fragmented.

test commands:
   %iptables -t nat -I POSTROUTING -j MASQUERADE
   %hping3 192.168.4.2 -s 1000 -p 2000 -d 60000

splat looks like:
[  261.069429] kernel BUG at net/ipv4/ip_output.c:636!
[  261.075753] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI
[  261.083854] CPU: 1 PID: 1349 Comm: hping3 Not tainted 4.19.0-rc2+ #3
[  261.100977] RIP: 0010:ip_do_fragment+0x1613/0x2600
[  261.106945] Code: e8 e2 38 e3 fe 4c 8b 44 24 18 48 8b 74 24 08 e9 92 f6 ff ff 80 3c 02 00 0f 85 da 07 00 00 48 8b b5 d0 00 00 00 e9 25 f6 ff ff <0f> 0b 0f 0b 44 8b 54 24 58 4c 8b 4c 24 18 4c 8b 5c 24 60 4c 8b 6c
[  261.127015] RSP: 0018:ffff8801031cf2c0 EFLAGS: 00010202
[  261.134156] RAX: 1ffff1002297537b RBX: ffffed0020639e6e RCX: 0000000000000004
[  261.142156] RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffff880114ba9bd8
[  261.150157] RBP: ffff880114ba8a40 R08: ffffed0022975395 R09: ffffed0022975395
[  261.158157] R10: 0000000000000001 R11: ffffed0022975394 R12: ffff880114ba9ca4
[  261.166159] R13: 0000000000000010 R14: ffff880114ba9bc0 R15: dffffc0000000000
[  261.174169] FS:  00007fbae2199700(0000) GS:ffff88011b400000(0000) knlGS:0000000000000000
[  261.183012] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  261.189013] CR2: 00005579244fe000 CR3: 0000000119bf4000 CR4: 00000000001006e0
[  261.198158] Call Trace:
[  261.199018]  ? dst_output+0x180/0x180
[  261.205011]  ? save_trace+0x300/0x300
[  261.209018]  ? ip_copy_metadata+0xb00/0xb00
[  261.213034]  ? sched_clock_local+0xd4/0x140
[  261.218158]  ? kill_l4proto+0x120/0x120 [nf_conntrack]
[  261.223014]  ? rt_cpu_seq_stop+0x10/0x10
[  261.227014]  ? find_held_lock+0x39/0x1c0
[  261.233008]  ip_finish_output+0x51d/0xb50
[  261.237006]  ? ip_fragment.constprop.56+0x220/0x220
[  261.243011]  ? nf_ct_l4proto_register_one+0x5b0/0x5b0 [nf_conntrack]
[  261.250152]  ? rcu_is_watching+0x77/0x120
[  261.255010]  ? nf_nat_ipv4_out+0x1e/0x2b0 [nf_nat_ipv4]
[  261.261033]  ? nf_hook_slow+0xb1/0x160
[  261.265007]  ip_output+0x1c7/0x710
[  261.269005]  ? ip_mc_output+0x13f0/0x13f0
[  261.273002]  ? __local_bh_enable_ip+0xe9/0x1b0
[  261.278152]  ? ip_fragment.constprop.56+0x220/0x220
[  261.282996]  ? nf_hook_slow+0xb1/0x160
[  261.287007]  raw_sendmsg+0x21f9/0x4420
[  261.291008]  ? dst_output+0x180/0x180
[  261.297003]  ? sched_clock_cpu+0x126/0x170
[  261.301003]  ? find_held_lock+0x39/0x1c0
[  261.306155]  ? stop_critical_timings+0x420/0x420
[  261.311004]  ? check_flags.part.36+0x450/0x450
[  261.315005]  ? _raw_spin_unlock_irq+0x29/0x40
[  261.320995]  ? _raw_spin_unlock_irq+0x29/0x40
[  261.326142]  ? cyc2ns_read_end+0x10/0x10
[  261.330139]  ? raw_bind+0x280/0x280
[  261.334138]  ? sched_clock_cpu+0x126/0x170
[  261.338995]  ? check_flags.part.36+0x450/0x450
[  261.342991]  ? __lock_acquire+0x4500/0x4500
[  261.348994]  ? inet_sendmsg+0x11c/0x500
[  261.352989]  ? dst_output+0x180/0x180
[  261.357012]  inet_sendmsg+0x11c/0x500
[ ... ]

v2:
 - clear skb->sk at reassembly routine.(Eric Dumarzet)

Fixes: fa0f527358bd ("ip: use rb trees for IP frag queue.")
Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ip: process in-order fragments efficiently

This patch changes the runtime behavior of IP defrag queue:
incoming in-order fragments are added to the end of the current
list/"run" of in-order fragments at the tail.

On some workloads, UDP stream performance is substantially improved:

RX: ./udp_stream -F 10 -T 2 -l 60
TX: ./udp_stream -c -H <host> -F 10 -T 5 -l 60

with this patchset applied on a 10Gbps receiver:

  throughput=9524.18
  throughput_units=Mbit/s

upstream (net-next):

  throughput=4608.93
  throughput_units=Mbit/s

Reported-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Peter Oskolkov <posk@google.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit a4fd284a1f8fd4b6c59aa59db2185b1e17c5c11c)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ip: add helpers to process in-order fragments faster.

This patch introduces several helper functions/macros that will be
used in the follow-up patch. No runtime changes yet.

The new logic (fully implemented in the second patch) is as follows:

* Nodes in the rb-tree will now contain not single fragments, but lists
  of consecutive fragments ("runs").

* At each point in time, the current "active" run at the tail is
  maintained/tracked. Fragments that arrive in-order, adjacent
  to the previous tail fragment, are added to this tail run without
  triggering the re-balancing of the rb-tree.

* If a fragment arrives out of order with the offset _before_ the tail run,
  it is inserted into the rb-tree as a single fragment.

* If a fragment arrives after the current tail fragment (with a gap),
  it starts a new "tail" run, as is inserted into the rb-tree
  at the end as the head of the new run.

skb->cb is used to store additional information
needed here (suggested by Eric Dumazet).

Reported-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Peter Oskolkov <posk@google.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 353c9cb360874e737fb000545f783df756c06f9a)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipv4: frags: precedence bug in ip_expire()

We accidentally removed the parentheses here, but they are required
because '!' has higher precedence than '&'.

Fixes: fa0f527358bd ("ip: use rb trees for IP frag queue.")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 70837ffe3085c9a91488b52ca13ac84424da1042)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: sk_buff rbnode reorg

commit bffa72cf7f9df842f0016ba03586039296b4caaf upstream

skb->rbnode shares space with skb->next, skb->prev and skb->tstamp

Current uses (TCP receive ofo queue and netem) need to save/restore
tstamp, while skb->dev is either NULL (TCP) or a constant for a given
queue (netem).

Since we plan using an RB tree for TCP retransmit queue to speedup SACK
processing with large BDP, this patch exchanges skb->dev and
skb->tstamp.

This saves some overhead in both TCP and netem.

v2: removes the swtstamp field from struct tcp_skb_cb

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Soheil Hassas Yeganeh <soheil@google.com>
Cc: Wei Wang <weiwan@google.com>
Cc: Willem de Bruijn <willemb@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: add rb_to_skb() and other rb tree helpers

Geeralize private netem_rb_to_skb()

TCP rtx queue will soon be converted to rb-tree,
so we will need skb_rbtree_walk() helpers.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 18a4c0eab2623cc95be98a1e6af1ad18e7695977)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: pskb_trim_rcsum() and CHECKSUM_COMPLETE are friends

After working on IP defragmentation lately, I found that some large
packets defeat CHECKSUM_COMPLETE optimization because of NIC adding
zero paddings on the last (small) fragment.

While removing the padding with pskb_trim_rcsum(), we set skb->ip_summed
to CHECKSUM_NONE, forcing a full csum validation, even if all prior
fragments had CHECKSUM_COMPLETE set.

We can instead compute the checksum of the part we are trimming,
usually smaller than the part we keep.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 88078d98d1bb085d72af8437707279e203524fa5)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipv6: defrag: drop non-last frags smaller than min mtu

don't bother with pathological cases, they only waste cycles.
IPv6 requires a minimum MTU of 1280 so we should never see fragments
smaller than this (except last frag).

v3: don't use awkward "-offset + len"
v2: drop IPv4 part, which added same check w. IPV4_MIN_MTU (68).
    There were concerns that there could be even smaller frags
    generated by intermediate nodes, e.g. on radio networks.

Cc: Peter Oskolkov <posk@google.com>
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 0ed4229b08c13c84a3c301a08defdc9e7f4467e6)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: modify skb_rbtree_purge to return the truesize of all purged skbs.

Tested: see the next patch is the series.

Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Peter Oskolkov <posk@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 385114dec8a49b5e5945e77ba7de6356106713f4)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: speed up skb_rbtree_purge()

As measured in my prior patch ("sch_netem: faster rb tree removal"),
rbtree_postorder_for_each_entry_safe() is nice looking but much slower
than using rb_next() directly, except when tree is small enough
to fit in CPU caches (then the cost is the same)

Also note that there is not even an increase of text size :
$ size net/core/skbuff.o.before net/core/skbuff.o
   text    data     bss     dec     hex filename
  40711    1298       0   42009    a419 net/core/skbuff.o.before
  40711    1298       0   42009    a419 net/core/skbuff.o

From: Eric Dumazet <edumazet@google.com>

Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 7c90584c66cc4b033a3b684b0e0950f79e7b7166)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ip: discard IPv4 datagrams with overlapping segments.

This behavior is required in IPv6, and there is little need
to tolerate overlapping fragments in IPv4. This change
simplifies the code and eliminates potential DDoS attack vectors.

Tested: ran ip_defrag selftest (not yet available uptream).

Suggested-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Peter Oskolkov <posk@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Florian Westphal <fw@strlen.de>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 7969e5c40dfd04799d4341f1b7cd266b6e47f227)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

inet: frags: fix ip6frag_low_thresh boundary

Giving an integer to proc_doulongvec_minmax() is dangerous on 64bit arches,
since linker might place next to it a non zero value preventing a change
to ip6frag_low_thresh.

ip6frag_low_thresh is not used anymore in the kernel, but we do not
want to prematuraly break user scripts wanting to change it.

Since specifying a minimal value of 0 for proc_doulongvec_minmax()
is moot, let's remove these zero values in all defrag units.

Fixes: 6e00f7dd5e4e ("ipv6: frags: fix /proc/sys/net/ipv6/ip6frag_low_thresh")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Maciej Żenczykowski <maze@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 3d23401283e80ceb03f765842787e0e79ff598b7)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

inet: frags: get rid of ipfrag_skb_cb/FRAG_CB

ip_defrag uses skb->cb[] to store the fragment offset, and unfortunately
this integer is currently in a different cache line than skb->next,
meaning that we use two cache lines per skb when finding the insertion point.

By aliasing skb->ip_defrag_offset and skb->dev, we pack all the fields
in a single cache line and save precious memory bandwidth.

Note that after the fast path added by Changli Gao in commit
d6bebca92c66 ("fragment: add fast path for in-order fragments")
this change wont help the fast path, since we still need
to access prev->len (2nd cache line), but will show great
benefits when slow path is entered, since we perform
a linear scan of a potentially long list.

Also, note that this potential long list is an attack vector,
we might consider also using an rb-tree there eventually.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit bf66337140c64c27fa37222b7abca7e49d63fb57)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

inet: frags: reorganize struct netns_frags

Put the read-mostly fields in a separate cache line
at the beginning of struct netns_frags, to reduce
false sharing noticed in inet_frag_kill()

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit c2615cf5a761b32bf74e85bddc223dfff3d9b9f0)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

rhashtable: reorganize struct rhashtable layout

While under frags DDOS I noticed unfortunate false sharing between
@nelems and @params.automatic_shrinking

Move @nelems at the end of struct rhashtable so that first cache line
is shared between all cpus, because almost never dirtied.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit e5d672a0780d9e7118caad4c171ec88b8299398d)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipv6: frags: rewrite ip6_expire_frag_queue()

Make it similar to IPv4 ip_expire(), and release the lock
before calling icmp functions.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 05c0b86b9696802fd0ce5676a92a63f1b455bdf3)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

inet: frags: do not clone skb in ip_expire()

An skb_clone() was added in commit ec4fbd64751d ("inet: frag: release
spinlock before calling icmp_send()")

While fixing the bug at that time, it also added a very high cost
for DDOS frags, as the ICMP rate limit is applied after this
expensive operation (skb_clone() + consume_skb(), implying memory
allocations, copy, and freeing)

We can use skb_get(head) here, all we want is to make sure skb wont
be freed by another cpu.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 1eec5d5670084ee644597bd26c25e22c69b9f748)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

inet: frags: break the 2GB limit for frags storage

Some users are willing to provision huge amounts of memory to be able
to perform reassembly reasonnably well under pressure.

Current memory tracking is using one atomic_t and integers.

Switch to atomic_long_t so that 64bit arches can use more than 2GB,
without any cost for 32bit arches.

Note that this patch avoids an overflow error, if high_thresh was set
to ~2GB, since this test in inet_frag_alloc() was never true :

if (... || frag_mem_limit(nf) > nf->high_thresh)

Tested:

$ echo 16000000000 >/proc/sys/net/ipv4/ipfrag_high_thresh

<frag DDOS>

$ grep FRAG /proc/net/sockstat
FRAG: inuse 14705885 memory 16000002880

$ nstat -n ; sleep 1 ; nstat | grep Reas
IpReasmReqds                    3317150            0.0
IpReasmFails                    3317112            0.0

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 3e67f106f619dcfaf6f4e2039599bdb69848c714)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

inet: frags: remove inet_frag_maybe_warn_overflow()

This function is obsolete, after rhashtable addition to inet defrag.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 2d44ed22e607f9a285b049de2263e3840673a260)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

inet: frags: get rif of inet_frag_evicting()

This refactors ip_expire() since one indentation level is removed.

Note: in the future, we should try hard to avoid the skb_clone()
since this is a serious performance cost.
Under DDOS, the ICMP message wont be sent because of rate limits.

Fact that ip6_expire_frag_queue() does not use skb_clone() is
disturbing too. Presumably IPv6 should have the same
issue than the one we fixed in commit ec4fbd64751d
("inet: frag: release spinlock before calling icmp_send()")

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 399d1404be660d355192ff4df5ccc3f4159ec1e4)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

inet: frags: remove some helpers

Remove sum_frag_mem_limit(), ip_frag_mem() & ip6_frag_mem()

Also since we use rhashtable we can bring back the number of fragments
in "grep FRAG /proc/net/sockstat /proc/net/sockstat6" that was
removed in commit 434d305405ab ("inet: frag: don't account number
of fragment queues")

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 6befe4a78b1553edb6eed3a78b4bcd9748526672)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

inet: frags: use rhashtables for reassembly units

Some applications still rely on IP fragmentation, and to be fair linux
reassembly unit is not working under any serious load.

It uses static hash tables of 1024 buckets, and up to 128 items per bucket (!!!)

A work queue is supposed to garbage collect items when host is under memory
pressure, and doing a hash rebuild, changing seed used in hash computations.

This work queue blocks softirqs for up to 25 ms when doing a hash rebuild,
occurring every 5 seconds if host is under fire.

Then there is the problem of sharing this hash table for all netns.

It is time to switch to rhashtables, and allocate one of them per netns
to speedup netns dismantle, since this is a critical metric these days.

Lookup is now using RCU. A followup patch will even remove
the refcount hold/release left from prior implementation and save
a couple of atomic operations.

Before this patch, 16 cpus (16 RX queue NIC) could not handle more
than 1 Mpps frags DDOS.

After the patch, I reach 9 Mpps without any tuning, and can use up to 2GB
of storage for the fragments (exact number depends on frags being evicted
after timeout)

$ grep FRAG /proc/net/sockstat
FRAG: inuse 1966916 memory 2140004608

A followup patch will change the limits for 64bit arches.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Kirill Tkhai <ktkhai@virtuozzo.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Florian Westphal <fw@strlen.de>
Cc: Jesper Dangaard Brouer <brouer@redhat.com>
Cc: Alexander Aring <alex.aring@gmail.com>
Cc: Stefan Schmidt <stefan@osg.samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 648700f76b03b7e8149d13cc2bdb3355035258a9)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

rhashtable: add schedule points

Rehashing and destroying large hash table takes a lot of time,
and happens in process context. It is safe to add cond_resched()
in rhashtable_rehash_table() and rhashtable_free_and_destroy()

Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit ae6da1f503abb5a5081f9f6c4a6881de97830f3e)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipv6: export ip6 fragments sysctl to unprivileged users

IPv4 was changed in commit 52a773d645e9 ("net: Export ip fragment
sysctl to unprivileged users")

The only sysctl that is not per-netns is not used :
ip6frag_secret_interval

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Nikolay Borisov <kernel@kyup.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 18dcbe12fe9fca0ab825f7eff993060525ac2503)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

inet: frags: refactor lowpan_net_frag_init()

We want to call lowpan_net_frag_init() earlier.
Similar to commit "inet: frags: refactor ipv6_frag_init()"

This is a prereq to "inet: frags: use rhashtables for reassembly units"

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 807f1844df4ac23594268fa9f41902d0549e92aa)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>