openwrt/staging/blogic.git
6 years agoqed: Fix memory/entry leak in qed_init_sp_request()
Denis Bolotin [Thu, 8 Nov 2018 14:46:08 +0000 (16:46 +0200)]
qed: Fix memory/entry leak in qed_init_sp_request()

Free the allocated SPQ entry or return the acquired SPQ entry to the free
list in error flows.

Signed-off-by: Denis Bolotin <denis.bolotin@cavium.com>
Signed-off-by: Michal Kalderon <michal.kalderon@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoinet: frags: better deal with smp races
Eric Dumazet [Fri, 9 Nov 2018 01:34:27 +0000 (17:34 -0800)]
inet: frags: better deal with smp races

Multiple cpus might attempt to insert a new fragment in rhashtable,
if for example RPS is buggy, as reported by 배석진 in
https://patchwork.ozlabs.org/patch/994601/

We use rhashtable_lookup_get_insert_key() instead of
rhashtable_insert_fast() to let cpus losing the race
free their own inet_frag_queue and use the one that
was inserted by another cpu.

Fixes: 648700f76b03 ("inet: frags: use rhashtables for reassembly units")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: 배석진 <soukjin.bae@samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: hns3: bugfix for not checking return value
Huazhong Tan [Thu, 8 Nov 2018 02:13:19 +0000 (10:13 +0800)]
net: hns3: bugfix for not checking return value

hns3_reset_notify_init_enet() only return error early if the return
value of hns3_restore_vlan() is not 0.

This patch adds checking for the return value of hns3_restore_vlan.

Fixes: 7fa6be4fd2f6 ("net: hns3: fix incorrect return value/type of some functions")
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoqlcnic: remove assumption that vlan_tci != 0
Michał Mirosław [Wed, 7 Nov 2018 16:50:53 +0000 (17:50 +0100)]
qlcnic: remove assumption that vlan_tci != 0

VLAN.TCI == 0 is perfectly valid (802.1p), so allow it to be accelerated.

Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoibmvnic: fix accelerated VLAN handling
Michał Mirosław [Wed, 7 Nov 2018 16:50:52 +0000 (17:50 +0100)]
ibmvnic: fix accelerated VLAN handling

Don't request tag insertion when it isn't present in outgoing skb.

Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'FDDI-defza-Fix-a-bunch-of-small-issues'
David S. Miller [Thu, 8 Nov 2018 05:53:31 +0000 (21:53 -0800)]
Merge branch 'FDDI-defza-Fix-a-bunch-of-small-issues'

Maciej W. Rozycki says:

====================
FDDI: defza: Fix a bunch of small issues

 Here is a bunch of small fixes addressing issues that I missed in my
final round of testing.  None of these affect run-time behaviour.  One was
actually found by the kbuild bot, which turned out to be more pedantic
than my compiler.  See individual change descriptions for details.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoFDDI: defza: Make the driver version string constant
Maciej W. Rozycki [Wed, 7 Nov 2018 12:07:10 +0000 (12:07 +0000)]
FDDI: defza: Make the driver version string constant

The driver version string is obviously not meant to be changed at run
time, so mark it `const'.

Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoFDDI: defza: Move SMT Tx data buffer declaration next to its skb
Maciej W. Rozycki [Wed, 7 Nov 2018 12:07:05 +0000 (12:07 +0000)]
FDDI: defza: Move SMT Tx data buffer declaration next to its skb

Move the temporary data buffer used when tapping into the SMT Tx queue
from the outer function level into the conditional block it's actually
used in and its containing skb is also declared, making the structure of
code better.

Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoFDDI: defza: Add missing comment closing
Maciej W. Rozycki [Wed, 7 Nov 2018 12:06:59 +0000 (12:06 +0000)]
FDDI: defza: Add missing comment closing

Fix:

drivers/net/fddi/defza.h:238:1: warning: "/*" within comment [-Wcomment]

by adding a missing comment closing.

Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoFDDI: defza: Fix SPDX annotation
Maciej W. Rozycki [Wed, 7 Nov 2018 12:06:52 +0000 (12:06 +0000)]
FDDI: defza: Fix SPDX annotation

The SPDX annotation for this driver does not match the license text,
which specifies GNU GPL 2 or later.  Make the two match by correcting
the SPDX tag.

Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/net...
David S. Miller [Thu, 8 Nov 2018 01:08:51 +0000 (17:08 -0800)]
Merge branch '100GbE' of git://git./linux/kernel/git/jkirsher/net-queue

Jeff Kirsher says:

====================
Intel Wired LAN Driver Updates 2018-11-07

This series contains fixes to igb, i40e and ice drivers.

Anirudh fixes an issue during rebuild of the ice driver, where we need
to set the carrier state, as well as start or stop the queues all based
on the link status.  Removed functions that were duplicating current
functionality in the VSI rebuild/replay framework.

Dave fixes a potential resource collision during the remove path, so add
a check to see if we are in the middle of a reset.  Fixed the remove
path to ensure we call netif_napi_del() to free vectors before we set
vsi->netdev to NULL.

Akeem fixes an issue when the receive or transmit pause parameter is
set, results in link loss on the interface.  Fixed the spelling of
"Enabling" in error message.

Victor fixes potential memory leak by also freeing the related VSI
contexts in the unload path.

Md Fahad fixes a flag during port VLAN insertion, which was not being
set properly.

Brett fixes a transmit timeout during stress due to the hardware tail
and software tail were incorrectly out of sync.

Miroslav Lichvar fixes the igb PHC timecounter update interval to be
sure the timecounter is updated in time.

Chinh fixes the req_speeds variable to be u16 instead of u8 so that it
can handle all the link speeds.

Jake fixes i40e to add back the missing feature flags, which was causing
IP-in-IP offloads to be reported as not supported.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoi40e: enable NETIF_F_NTUPLE and NETIF_F_HW_TC at driver load
Jacob Keller [Mon, 29 Oct 2018 17:52:43 +0000 (10:52 -0700)]
i40e: enable NETIF_F_NTUPLE and NETIF_F_HW_TC at driver load

The assignment of the feature flag NETIF_F_NTUPLE and NETIF_F_HW_TC
occurs prior to the initial setup of the local hw_features variable.

This means the features are set as user-changeable, but are not set in
the currently active feature list. This results in the features being
disabled at the driver's initial load.

Move the assignment after the initial assignment of hw_features, and
assign to the local variable. This ensures that NETIF_F_NTUPLE and
NETIF_F_HW_TC are marked as user-changeable, and also enables them by
default when the driver loads.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoi40e: restore NETIF_F_GSO_IPXIP[46] to netdev features
Jacob Keller [Mon, 29 Oct 2018 17:52:42 +0000 (10:52 -0700)]
i40e: restore NETIF_F_GSO_IPXIP[46] to netdev features

Since commit bacd75cfac8a ("i40e/i40evf: Add capability exchange for
outer checksum", 2017-04-06) the i40e driver has not reported support
for IP-in-IP offloads. This likely occurred due to a bad rebase, as the
commit extracts hw_enc_features into its own variable. As part of this
change, it dropped the NETIF_F_FSO_IPXIP flags from the
netdev->hw_enc_features. This was unfortunately not caught during code
review.

Fix this by adding back the missing feature flags.

For reference, NETIF_F_GSO_IPXIP4 was added in commit 7e13318daa4a
("net: define gso types for IPx over IPv4 and IPv6", 2016-05-20),
replacing NETIF_F_GSO_IPIP and NETIF_F_GSO_SIT.

NETIF_F_GSO_IPXIP6 was added in commit bf2d1df39502 ("intel: Add support
for IPv6 IP-in-IP offload", 2016-05-20).

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoice: Change req_speeds to be u16
Chinh T Cao [Mon, 5 Nov 2018 20:18:35 +0000 (12:18 -0800)]
ice: Change req_speeds to be u16

Since the req_speeds field in struct ice_link_status is a u8,
req_speeds & ICE_AQ_LINK_SPEED_40GB always returns 0. This was caught
by a coverity scan.

Fix this by changing req_speeds to be u16.

Reported-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Chinh T Cao <chinh.t.cao@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoigb: shorten maximum PHC timecounter update interval
Miroslav Lichvar [Fri, 26 Oct 2018 17:13:00 +0000 (19:13 +0200)]
igb: shorten maximum PHC timecounter update interval

The timecounter needs to be updated at least once per ~550 seconds in
order to avoid a 40-bit SYSTIM timestamp to be misinterpreted as an old
timestamp.

Since commit 500462a9de65 ("timers: Switch to a non-cascading wheel"),
scheduling of delayed work seems to be less accurate and a requested
delay of 540 seconds may actually be longer than 550 seconds. Also, the
PHC may be adjusted to run up to 6% faster than real time and the system
clock up to 10% slower. Shorten the delay to 360 seconds to be sure the
timecounter is updated in time.

This fixes an issue with HW timestamps on 82580/I350/I354 being off by
~1100 seconds for few seconds every ~9 minutes.

Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Miroslav Lichvar <mlichvar@redhat.com>
Acked-by: Jacob Keller <jacob.e.keller@intel.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoice: Fix the bytecount sent to netdev_tx_sent_queue
Brett Creeley [Fri, 26 Oct 2018 17:40:59 +0000 (10:40 -0700)]
ice: Fix the bytecount sent to netdev_tx_sent_queue

Currently if the driver does a TSO offload the bytecount sent to
netdev_tx_sent_queue will be incorrect. This is because in ice_tso we
overwrite the initial value that we set in ice_tx_map. This creates a
mismatch between the Tx and Tx clean flow. In the Tx clean flow we
calculate the bytecount (called total_bytes) as we clean the
descriptors so the value used in the Tx clean path is correct. Fix this
by using += in ice_tso instead of =. This fixes the mismatch in
bytecount mentioned above.

Signed-off-by: Brett Creeley <brett.creeley@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoice: Fix tx_timeout in PF driver
Brett Creeley [Fri, 26 Oct 2018 17:40:58 +0000 (10:40 -0700)]
ice: Fix tx_timeout in PF driver

Prior to this commit the driver was running into tx_timeouts when a
queue was stressed enough. This was happening because the HW tail
and SW tail (NTU) were incorrectly out of sync. Consequently this was
causing the HW head to collide with the HW tail, which to the hardware
means that all descriptors posted for Tx have been processed.

Due to the Tx logic used in the driver SW tail and HW tail are allowed
to be out of sync. This is done as an optimization because it allows the
driver to write HW tail as infrequently as possible, while still
updating the SW tail index to keep track. However, there are situations
where this results in the tail never getting updated, resulting in Tx
timeouts.

Tx HW tail write condition:
if (netif_xmit_stopped(txring_txq(tx_ring) || !skb->xmit_more)
writel(sw_tail, tx_ring->tail);

An issue was found in the Tx logic that was causing the afore mentioned
condition for updating HW tail to never happen, causing tx_timeouts.

In ice_xmit_frame_ring we calculate how many descriptors we need for the
Tx transaction based on the skb the kernel hands us. This is then passed
into ice_maybe_stop_tx along with some extra padding to determine if we
have enough descriptors available for this transaction. If we don't then
we return -EBUSY to the stack, otherwise we move on and eventually
prepare the Tx descriptors accordingly in ice_tx_map and set
next_to_watch. In ice_tx_map we make another call to ice_maybe_stop_tx
with a value of MAX_SKB_FRAGS + 4. The key here is that this value is
possibly less than the value we sent in the first call to
ice_maybe_stop_tx in ice_xmit_frame_ring. Now, if the number of unused
descriptors is between MAX_SKB_FRAGS + 4 and the value used in the first
call to ice_maybe_stop_tx in ice_xmit_frame_ring then we do not update
the HW tail because of the "Tx HW tail write condition" above. This is
because in ice_maybe_stop_tx we return success from ice_maybe_stop_tx
instead of calling __ice_maybe_stop_tx and subsequently calling
netif_stop_subqueue, which sets the __QUEUE_STATE_DEV_XOFF bit. This
bit is then checked in the "Tx HW tail write condition" by calling
netif_xmit_stopped and subsequently updating HW tail if the
afore mentioned bit is set.

In ice_clean_tx_irq, if next_to_watch is not NULL, we end up cleaning
the descriptors that HW sets the DD bit on and we have the budget. The
HW head will eventually run into the HW tail in response to the
description in the paragraph above.

The next time through ice_xmit_frame_ring we make the initial call to
ice_maybe_stop_tx with another skb from the stack. This time we do not
have enough descriptors available and we return NETDEV_TX_BUSY to the
stack and end up setting next_to_watch to NULL.

This is where we are stuck. In ice_clean_tx_irq we never clean anything
because next_to_watch is always NULL and in ice_xmit_frame_ring we never
update HW tail because we already return NETDEV_TX_BUSY to the stack and
eventually we hit a tx_timeout.

This issue was fixed by making sure that the second call to
ice_maybe_stop_tx in ice_tx_map is passed a value that is >= the value
that was used on the initial call to ice_maybe_stop_tx in
ice_xmit_frame_ring. This was done by adding the following defines to
make the logic more clear and to reduce the chance of mucking this up
again:

ICE_CACHE_LINE_BYTES 64
ICE_DESCS_PER_CACHE_LINE (ICE_CACHE_LINE_BYTES / \
 sizeof(struct ice_tx_desc))
ICE_DESCS_FOR_CTX_DESC 1
ICE_DESCS_FOR_SKB_DATA_PTR 1

The ICE_CACHE_LINE_BYTES being 64 is an assumption being made so we
don't have to figure this out on every pass through the Tx path. Instead
I added a sanity check in ice_probe to verify cache line size and print
a message if it's not 64 Bytes. This will make it easier to file issues
if they are seen when the cache line size is not 64 Bytes when reading
from the GLPCI_CNF2 register.

Signed-off-by: Brett Creeley <brett.creeley@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoice: Fix napi delete calls for remove
Dave Ertman [Fri, 26 Oct 2018 17:40:57 +0000 (10:40 -0700)]
ice: Fix napi delete calls for remove

In the remove path, the vsi->netdev is being set to NULL before the call
to free vectors. This is causing the netif_napi_del call to never be made.

Add a call to ice_napi_del to the same location as the calls to
unregister_netdev and just prior to them. This will use the reverse flow
as the register and netif_napi_add calls.

Signed-off-by: Dave Ertman <david.m.ertman@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoice: Fix typo in error message
Anirudh Venkataramanan [Fri, 26 Oct 2018 17:40:56 +0000 (10:40 -0700)]
ice: Fix typo in error message

Print should say "Enabling" instead of "Enaabling"

Signed-off-by: Akeem G Abodunrin <akeem.g.abodunrin@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoice: Fix flags for port VLAN
Md Fahad Iqbal Polash [Fri, 26 Oct 2018 17:40:55 +0000 (10:40 -0700)]
ice: Fix flags for port VLAN

According to the spec, whenever insert PVID field is set, the VLAN
driver insertion mode should be set to 01b which isn't done currently.
Fix it.

Signed-off-by: Md Fahad Iqbal Polash <md.fahad.iqbal.polash@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoice: Remove duplicate addition of VLANs in replay path
Anirudh Venkataramanan [Fri, 26 Oct 2018 17:40:54 +0000 (10:40 -0700)]
ice: Remove duplicate addition of VLANs in replay path

ice_restore_vlan and active_vlans were originally put in place to
reprogram VLAN filters in the replay path. This is now done as part
of the much broader VSI rebuild/replay framework. So remove both
ice_restore_vlan and active_vlans

Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoice: Free VSI contexts during for unload
Victor Raj [Fri, 26 Oct 2018 17:40:53 +0000 (10:40 -0700)]
ice: Free VSI contexts during for unload

In the unload path, all VSIs are freed. Also free the related VSI
contexts to prevent memory leaks.

Signed-off-by: Victor Raj <victor.raj@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoice: Fix dead device link issue with flow control
Akeem G Abodunrin [Fri, 26 Oct 2018 17:40:52 +0000 (10:40 -0700)]
ice: Fix dead device link issue with flow control

Setting Rx or Tx pause parameter currently results in link loss on the
interface, requiring the platform/host to be cold power cycled. Fix it.

Signed-off-by: Akeem G Abodunrin <akeem.g.abodunrin@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoice: Check for reset in progress during remove
Anirudh Venkataramanan [Fri, 26 Oct 2018 17:40:51 +0000 (10:40 -0700)]
ice: Check for reset in progress during remove

The remove path does not currently check to see if a
reset is in progress before proceeding.  This can cause
a resource collision resulting in various types of errors.

Check for reset in progress and wait for a reasonable
amount of time before allowing the remove to progress.

Signed-off-by: Dave Ertman <david.m.ertman@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoice: Set carrier state and start/stop queues in rebuild
Anirudh Venkataramanan [Fri, 26 Oct 2018 17:40:50 +0000 (10:40 -0700)]
ice: Set carrier state and start/stop queues in rebuild

Set the carrier state post rebuild by querying the link status. Also
start/stop queues based on link status.

Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agonet: phy: Allow BCM54616S PHY to setup internal TX/RX clock delay
Tao Ren [Mon, 5 Nov 2018 22:35:40 +0000 (14:35 -0800)]
net: phy: Allow BCM54616S PHY to setup internal TX/RX clock delay

This patch allows users to enable/disable internal TX and/or RX clock
delay for BCM54616S PHYs so as to satisfy RGMII timing specifications.

On a particular platform, whether TX and/or RX clock delay is required
depends on how PHY connected to the MAC IP. This requirement can be
specified through "phy-mode" property in the platform device tree.

The patch is inspired by commit 733336262b28 ("net: phy: Allow BCM5481x
PHYs to setup internal TX/RX clock delay").

Signed-off-by: Tao Ren <taoren@fb.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge tag 'trace-v4.20-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt...
Linus Torvalds [Tue, 6 Nov 2018 16:12:10 +0000 (08:12 -0800)]
Merge tag 'trace-v4.20-rc1' of git://git./linux/kernel/git/rostedt/linux-trace

Pull tracing fix from Steven Rostedt:
 "Masami found a slight bug in his code where he transposed the
  arguments of a call to strpbrk.

  The reason this wasn't detected in our tests is that the only way this
  would transpire is when a kprobe event with a symbol offset is
  attached to a function that belongs to a module that isn't loaded yet.
  When the kprobe trace event is added, the offset would be truncated
  after it was parsed, and when the module is loaded, it would use the
  symbol without the offset (as the nul character added by the parsing
  would not be replaced with the original character)"

* tag 'trace-v4.20-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  tracing/kprobes: Fix strpbrk() argument order

6 years agoMerge branch 'spectre' of git://git.armlinux.org.uk/~rmk/linux-arm
Linus Torvalds [Tue, 6 Nov 2018 16:10:01 +0000 (08:10 -0800)]
Merge branch 'spectre' of git://git.armlinux.org.uk/~rmk/linux-arm

Pull ARM fix from Russell King:
 "Ard spotted a typo in one of the assembly files which leads to a
  kernel oops when that code path is executed. Fix this"

* 'spectre' of git://git.armlinux.org.uk/~rmk/linux-arm:
  ARM: 8809/1: proc-v7: fix Thumb annotation of cpu_v7_hvc_switch_mm

6 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Tue, 6 Nov 2018 15:44:04 +0000 (07:44 -0800)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) Handle errors mid-stream of an all dump, from Alexey Kodanev.

 2) Fix build of openvswitch with certain combinations of netfilter
    options, from Arnd Bergmann.

 3) Fix interactions between GSO and BQL, from Eric Dumazet.

 4) Don't put a '/' in RTL8201F's sysfs file name, from Holger
    Hoffstätte.

 5) S390 qeth driver fixes from Julian Wiedmann.

 6) Allow ipv6 link local addresses for netconsole when both source and
    destination are link local, from Matwey V. Kornilov.

 7) Fix the BPF program address seen in /proc/kallsyms, from Song Liu.

 8) Initialize mutex before use in dsa microchip driver, from Tristram
    Ha.

 9) Out-of-bounds access in hns3, from Yunsheng Lin.

10) Various netfilter fixes from Stefano Brivio, Jozsef Kadlecsik, Jiri
    Slaby, Florian Westphal, Eric Westbrook, Andrey Ryabinin, and Pablo
    Neira Ayuso.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (50 commits)
  net: alx: make alx_drv_name static
  net: bpfilter: fix iptables failure if bpfilter_umh is disabled
  sock_diag: fix autoloading of the raw_diag module
  net: core: netpoll: Enable netconsole IPv6 link local address
  ipv6: properly check return value in inet6_dump_all()
  rtnetlink: restore handling of dumpit return value in rtnl_dump_all()
  net/ipv6: Move anycast init/cleanup functions out of CONFIG_PROC_FS
  bonding/802.3ad: fix link_failure_count tracking
  net: phy: realtek: fix RTL8201F sysfs name
  sctp: define SCTP_SS_DEFAULT for Stream schedulers
  sctp: fix strchange_flags name for Stream Change Event
  mlxsw: spectrum: Fix IP2ME CPU policer configuration
  openvswitch: fix linking without CONFIG_NF_CONNTRACK_LABELS
  qed: fix link config error handling
  net: hns3: Fix for out-of-bounds access when setting pfc back pressure
  net/mlx4_en: use __netdev_tx_sent_queue()
  net: do not abort bulk send on BQL status
  net: bql: add __netdev_tx_sent_queue()
  s390/qeth: report 25Gbit link speed
  s390/qeth: sanitize ARP requests
  ...

6 years agoARM: 8809/1: proc-v7: fix Thumb annotation of cpu_v7_hvc_switch_mm
Ard Biesheuvel [Mon, 5 Nov 2018 13:54:56 +0000 (14:54 +0100)]
ARM: 8809/1: proc-v7: fix Thumb annotation of cpu_v7_hvc_switch_mm

Due to what appears to be a copy/paste error, the opening ENTRY()
of cpu_v7_hvc_switch_mm() lacks a matching ENDPROC(), and instead,
the one for cpu_v7_smc_switch_mm() is duplicated.

Given that it is ENDPROC() that emits the Thumb annotation, the
cpu_v7_hvc_switch_mm() routine will be called in ARM mode on a
Thumb2 kernel, resulting in the following splat:

  Internal error: Oops - undefined instruction: 0 [#1] SMP THUMB2
  Modules linked in:
  CPU: 0 PID: 1 Comm: swapper/0 Not tainted 4.18.0-rc1-00030-g4d28ad89189d-dirty #488
  Hardware name: QEMU KVM Virtual Machine, BIOS 0.0.0 02/06/2015
  PC is at cpu_v7_hvc_switch_mm+0x12/0x18
  LR is at flush_old_exec+0x31b/0x570
  pc : [<c0316efe>]    lr : [<c04117c7>]    psr: 00000013
  sp : ee899e50  ip : 00000000  fp : 00000001
  r10: eda28f34  r9 : eda31800  r8 : c12470e0
  r7 : eda1fc00  r6 : eda53000  r5 : 00000000  r4 : ee88c000
  r3 : c0316eec  r2 : 00000001  r1 : eda53000  r0 : 6da6c000
  Flags: nzcv  IRQs on  FIQs on  Mode SVC_32  ISA ARM  Segment none

Note the 'ISA ARM' in the last line.

Fix this by using the correct name in ENDPROC().

Cc: <stable@vger.kernel.org>
Fixes: 10115105cb3a ("ARM: spectre-v2: add firmware based hardening")
Reviewed-by: Dave Martin <Dave.Martin@arm.com>
Acked-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
6 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf
David S. Miller [Tue, 6 Nov 2018 01:19:25 +0000 (17:19 -0800)]
Merge git://git./pub/scm/linux/kernel/git/pablo/nf

Pablo Neira Ayuso says:

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

The following patchset contains the first batch of Netfilter fixes for
your net tree:

1) Fix splat with IPv6 defragmenting locally generated fragments,
   from Florian Westphal.

2) Fix Incorrect check for missing attribute in nft_osf.

3) Missing INT_MIN & INT_MAX definition for netfilter bridge uapi
   header, from Jiri Slaby.

4) Revert map lookup in nft_numgen, this is already possible with
   the existing infrastructure without this extension.

5) Fix wrong listing of set reference counter, make counter
   synchronous again, from Stefano Brivio.

6) Fix CIDR 0 in hash:net,port,net, from Eric Westbrook.

7) Fix allocation failure with large set, use kvcalloc().
   From Andrey Ryabinin.

8) No need to disable BH when fetch ip set comment, patch from
   Jozsef Kadlecsik.

9) Sanity check for valid sysfs entry in xt_IDLETIMER, from
   Taehee Yoo.

10) Fix suspicious rcu usage via ip_set() macro at netlink dump,
    from Jozsef Kadlecsik.

11) Fix setting default timeout via nfnetlink_cttimeout, this
    comes with preparation patch to add nf_{tcp,udp,...}_pernet()
    helper.

12) Allow ebtables table nat to be of filter type via nft_compat.
    From Florian Westphal.

13) Incorrect calculation of next bucket in early_drop, do no bump
    hash value, update bucket counter instead. From Vasily Khoruzhick.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: alx: make alx_drv_name static
Rasmus Villemoes [Mon, 5 Nov 2018 17:52:21 +0000 (18:52 +0100)]
net: alx: make alx_drv_name static

alx_drv_name is not used outside main.c, so there's no reason for it to
have external linkage.

Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: bpfilter: fix iptables failure if bpfilter_umh is disabled
Taehee Yoo [Mon, 5 Nov 2018 13:31:41 +0000 (22:31 +0900)]
net: bpfilter: fix iptables failure if bpfilter_umh is disabled

When iptables command is executed, ip_{set/get}sockopt() try to upload
bpfilter.ko if bpfilter is enabled. if it couldn't find bpfilter.ko,
command is failed.
bpfilter.ko is generated if CONFIG_BPFILTER_UMH is enabled.
ip_{set/get}sockopt() only checks CONFIG_BPFILTER.
So that if CONFIG_BPFILTER is enabled and CONFIG_BPFILTER_UMH is disabled,
iptables command is always failed.

test config:
   CONFIG_BPFILTER=y
   # CONFIG_BPFILTER_UMH is not set

test command:
   %iptables -L
   iptables: No chain/target/match by that name.

Fixes: d2ba09c17a06 ("net: add skeleton of bpfilter kernel module")
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agosock_diag: fix autoloading of the raw_diag module
Andrei Vagin [Mon, 5 Nov 2018 06:37:15 +0000 (22:37 -0800)]
sock_diag: fix autoloading of the raw_diag module

IPPROTO_RAW isn't registred as an inet protocol, so
inet_protos[protocol] is always NULL for it.

Cc: Cyrill Gorcunov <gorcunov@gmail.com>
Cc: Xin Long <lucien.xin@gmail.com>
Fixes: bf2ae2e4bf93 ("sock_diag: request _diag module only when the family or proto has been registered")
Signed-off-by: Andrei Vagin <avagin@gmail.com>
Reviewed-by: Cyrill Gorcunov <gorcunov@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: core: netpoll: Enable netconsole IPv6 link local address
Matwey V. Kornilov [Fri, 2 Nov 2018 18:19:36 +0000 (21:19 +0300)]
net: core: netpoll: Enable netconsole IPv6 link local address

There is no reason to discard using source link local address when
remote netconsole IPv6 address is set to be link local one.

The patch allows administrators to use IPv6 netconsole without
explicitly configuring source address:

    netconsole=@/,@fe80::5054:ff:fe2f:6012/

Signed-off-by: Matwey V. Kornilov <matwey@sai.msu.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoipv6: properly check return value in inet6_dump_all()
Alexey Kodanev [Fri, 2 Nov 2018 16:11:05 +0000 (19:11 +0300)]
ipv6: properly check return value in inet6_dump_all()

Make sure we call fib6_dump_end() if it happens that skb->len
is zero. rtnl_dump_all() can reset cb->args on the next loop
iteration there.

Fixes: 08e814c9e8eb ("net/ipv6: Bail early if user only wants cloned entries")
Fixes: ae677bbb4441 ("net: Don't return invalid table id error when dumping all families")
Signed-off-by: Alexey Kodanev <alexey.kodanev@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agortnetlink: restore handling of dumpit return value in rtnl_dump_all()
Alexey Kodanev [Fri, 2 Nov 2018 16:11:04 +0000 (19:11 +0300)]
rtnetlink: restore handling of dumpit return value in rtnl_dump_all()

For non-zero return from dumpit() we should break the loop
in rtnl_dump_all() and return the result. Otherwise, e.g.,
we could get the memory leak in inet6_dump_fib() [1]. The
pointer to the allocated struct fib6_walker there (saved
in cb->args) can be lost, reset on the next iteration.

Fix it by partially restoring the previous behavior before
commit c63586dc9b3e ("net: rtnl_dump_all needs to propagate
error from dumpit function"). The returned error from
dumpit() is still passed further.

[1]:
unreferenced object 0xffff88001322a200 (size 96):
  comm "sshd", pid 1484, jiffies 4296032768 (age 1432.542s)
  hex dump (first 32 bytes):
    00 01 00 00 00 00 ad de 00 02 00 00 00 00 ad de  ................
    18 09 41 36 00 88 ff ff 18 09 41 36 00 88 ff ff  ..A6......A6....
  backtrace:
    [<0000000095846b39>] kmem_cache_alloc_trace+0x151/0x220
    [<000000007d12709f>] inet6_dump_fib+0x68d/0x940
    [<000000002775a316>] rtnl_dump_all+0x1d9/0x2d0
    [<00000000d7cd302b>] netlink_dump+0x945/0x11a0
    [<000000002f43485f>] __netlink_dump_start+0x55d/0x800
    [<00000000f76bbeec>] rtnetlink_rcv_msg+0x4fa/0xa00
    [<000000009b5761f3>] netlink_rcv_skb+0x29c/0x420
    [<0000000087a1dae1>] rtnetlink_rcv+0x15/0x20
    [<00000000691b703b>] netlink_unicast+0x4e3/0x6c0
    [<00000000b5be0204>] netlink_sendmsg+0x7f2/0xba0
    [<0000000096d2aa60>] sock_sendmsg+0xba/0xf0
    [<000000008c1b786f>] __sys_sendto+0x1e4/0x330
    [<0000000019587b3f>] __x64_sys_sendto+0xe1/0x1a0
    [<00000000071f4d56>] do_syscall_64+0x9f/0x300
    [<000000002737577f>] entry_SYSCALL_64_after_hwframe+0x44/0xa9
    [<0000000057587684>] 0xffffffffffffffff

Fixes: c63586dc9b3e ("net: rtnl_dump_all needs to propagate error from dumpit function")
Signed-off-by: Alexey Kodanev <alexey.kodanev@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/ipv6: Move anycast init/cleanup functions out of CONFIG_PROC_FS
Jeff Barnhill [Mon, 5 Nov 2018 20:36:45 +0000 (20:36 +0000)]
net/ipv6: Move anycast init/cleanup functions out of CONFIG_PROC_FS

Move the anycast.c init and cleanup functions which were inadvertently
added inside the CONFIG_PROC_FS definition.

Fixes: 2384d02520ff ("net/ipv6: Add anycast addresses to a global hashtable")
Signed-off-by: Jeff Barnhill <0xeffeff@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agocompiler: remove __no_sanitize_address_or_inline again
Martin Schwidefsky [Mon, 5 Nov 2018 06:36:28 +0000 (07:36 +0100)]
compiler: remove __no_sanitize_address_or_inline again

The __no_sanitize_address_or_inline and __no_kasan_or_inline defines
are almost identical. The only difference is that __no_kasan_or_inline
does not have the 'notrace' attribute.

To be able to replace __no_sanitize_address_or_inline with the older
definition, add 'notrace' to __no_kasan_or_inline and change to two
users of __no_sanitize_address_or_inline in the s390 code.

The 'notrace' option is necessary for e.g. the __load_psw_mask function
in arch/s390/include/asm/processor.h. Without the option it is possible
to trace __load_psw_mask which leads to kernel stack overflow.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Pointed-out-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Acked-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agotracing/kprobes: Fix strpbrk() argument order
Masami Hiramatsu [Thu, 1 Nov 2018 14:29:28 +0000 (23:29 +0900)]
tracing/kprobes: Fix strpbrk() argument order

Fix strpbrk()'s argument order, it must pass acceptable string
in 2nd argument. Note that this can cause a kernel panic where
it recovers backup character to code->data.

Link: http://lkml.kernel.org/r/154108256792.2604.1816052586385217811.stgit@devbox
Fixes: a6682814f371 ("tracing/kprobes: Allow kprobe-events to record module symbol")
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
6 years agobonding/802.3ad: fix link_failure_count tracking
Jarod Wilson [Sun, 4 Nov 2018 19:59:46 +0000 (14:59 -0500)]
bonding/802.3ad: fix link_failure_count tracking

Commit 4d2c0cda07448ea6980f00102dc3964eb25e241c set slave->link to
BOND_LINK_DOWN for 802.3ad bonds whenever invalid speed/duplex values
were read, to fix a problem with slaves getting into weird states, but
in the process, broke tracking of link failures, as going straight to
BOND_LINK_DOWN when a link is indeed down (cable pulled, switch rebooted)
means we broke out of bond_miimon_inspect()'s BOND_LINK_DOWN case because
!link_state was already true, we never incremented commit, and never got
a chance to call bond_miimon_commit(), where slave->link_failure_count
would be incremented. I believe the simple fix here is to mark the slave
as BOND_LINK_FAIL, and let bond_miimon_inspect() transition the link from
_FAIL to either _UP or _DOWN, and in the latter case, we now get proper
incrementing of link_failure_count again.

Fixes: 4d2c0cda0744 ("bonding: speed/duplex update at NETDEV_UP event")
CC: Mahesh Bandewar <maheshb@google.com>
CC: David S. Miller <davem@davemloft.net>
CC: netdev@vger.kernel.org
CC: stable@vger.kernel.org
Signed-off-by: Jarod Wilson <jarod@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: phy: realtek: fix RTL8201F sysfs name
Holger Hoffstätte [Sun, 4 Nov 2018 18:02:42 +0000 (19:02 +0100)]
net: phy: realtek: fix RTL8201F sysfs name

Since 4.19 the following error in sysfs has appeared when using the
r8169 NIC driver:

$cd /sys/module/realtek/drivers
$ls -l
ls: cannot access 'mdio_bus:RTL8201F 10/100Mbps Ethernet': No such file or directory
[..garbled dir entries follow..]

Apparently the forward slash in "10/100Mbps Ethernet" is interpreted
as directory separator that leads nowhere, and was introduced in commit
513588dd44b ("net: phy: realtek: add RTL8201F phy-id and functions").

Fix this by removing the offending slash in the driver name.

Other drivers in net/phy seem to have the same problem, but I cannot
test/verify them.

Fixes: 513588dd44b ("net: phy: realtek: add RTL8201F phy-id and functions")
Signed-off-by: Holger Hoffstätte <holger@applied-asynchrony.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoLinux 4.20-rc1
Linus Torvalds [Sun, 4 Nov 2018 23:37:52 +0000 (15:37 -0800)]
Linux 4.20-rc1

6 years agoMerge tag 'tags/upstream-4.20-rc1' of git://git.infradead.org/linux-ubifs
Linus Torvalds [Sun, 4 Nov 2018 22:46:04 +0000 (14:46 -0800)]
Merge tag 'tags/upstream-4.20-rc1' of git://git.infradead.org/linux-ubifs

Pull UBIFS updates from Richard Weinberger:

 - Full filesystem authentication feature, UBIFS is now able to have the
   whole filesystem structure authenticated plus user data encrypted and
   authenticated.

 - Minor cleanups

* tag 'tags/upstream-4.20-rc1' of git://git.infradead.org/linux-ubifs: (26 commits)
  ubifs: Remove unneeded semicolon
  Documentation: ubifs: Add authentication whitepaper
  ubifs: Enable authentication support
  ubifs: Do not update inode size in-place in authenticated mode
  ubifs: Add hashes and HMACs to default filesystem
  ubifs: authentication: Authenticate super block node
  ubifs: Create hash for default LPT
  ubfis: authentication: Authenticate master node
  ubifs: authentication: Authenticate LPT
  ubifs: Authenticate replayed journal
  ubifs: Add auth nodes to garbage collector journal head
  ubifs: Add authentication nodes to journal
  ubifs: authentication: Add hashes to index nodes
  ubifs: Add hashes to the tree node cache
  ubifs: Create functions to embed a HMAC in a node
  ubifs: Add helper functions for authentication support
  ubifs: Add separate functions to init/crc a node
  ubifs: Format changes for authentication support
  ubifs: Store read superblock node
  ubifs: Drop write_node
  ...

6 years agoMerge tag 'nfs-for-4.20-2' of git://git.linux-nfs.org/projects/trondmy/linux-nfs
Linus Torvalds [Sun, 4 Nov 2018 16:20:09 +0000 (08:20 -0800)]
Merge tag 'nfs-for-4.20-2' of git://git.linux-nfs.org/projects/trondmy/linux-nfs

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

  Bugfix:
   - Fix build issues on architectures that don't provide 64-bit cmpxchg

  Cleanups:
   - Fix a spelling mistake"

* tag 'nfs-for-4.20-2' of git://git.linux-nfs.org/projects/trondmy/linux-nfs:
  NFS: fix spelling mistake, EACCESS -> EACCES
  SUNRPC: Use atomic(64)_t for seq_send(64)

6 years agoMerge branch 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 4 Nov 2018 16:15:15 +0000 (08:15 -0800)]
Merge branch 'timers-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull more timer updates from Thomas Gleixner:
 "A set of commits for the new C-SKY architecture timers"

* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  dt-bindings: timer: gx6605s SOC timer
  clocksource/drivers/c-sky: Add gx6605s SOC system timer
  dt-bindings: timer: C-SKY Multi-processor timer
  clocksource/drivers/c-sky: Add C-SKY SMP timer

6 years agoMerge tag 'ntb-4.20' of git://github.com/jonmason/ntb
Linus Torvalds [Sun, 4 Nov 2018 16:12:44 +0000 (08:12 -0800)]
Merge tag 'ntb-4.20' of git://github.com/jonmason/ntb

Pull NTB updates from Jon Mason:
 "Fairly minor changes and bug fixes:

  NTB IDT thermal changes and hook into hwmon, ntb_netdev clean-up of
  private struct, and a few bug fixes"

* tag 'ntb-4.20' of git://github.com/jonmason/ntb:
  ntb: idt: Alter the driver info comments
  ntb: idt: Discard temperature sensor IRQ handler
  ntb: idt: Add basic hwmon sysfs interface
  ntb: idt: Alter temperature read method
  ntb_netdev: Simplify remove with client device drvdata
  NTB: transport: Try harder to alloc an aligned MW buffer
  ntb: ntb_transport: Mark expected switch fall-throughs
  ntb: idt: Set PCIe bus address to BARLIMITx
  NTB: ntb_hw_idt: replace IS_ERR_OR_NULL with regular NULL checks
  ntb: intel: fix return value for ndev_vec_mask()
  ntb_netdev: fix sleep time mismatch

6 years agosctp: define SCTP_SS_DEFAULT for Stream schedulers
Xin Long [Sat, 3 Nov 2018 06:01:31 +0000 (14:01 +0800)]
sctp: define SCTP_SS_DEFAULT for Stream schedulers

According to rfc8260#section-4.3.2, SCTP_SS_DEFAULT is required to
defined as SCTP_SS_FCFS or SCTP_SS_RR.

SCTP_SS_FCFS is used for SCTP_SS_DEFAULT's value in this patch.

Fixes: 5bbbbe32a431 ("sctp: introduce stream scheduler foundations")
Reported-by: Jianwen Ji <jiji@redhat.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agosctp: fix strchange_flags name for Stream Change Event
Xin Long [Sat, 3 Nov 2018 05:59:45 +0000 (13:59 +0800)]
sctp: fix strchange_flags name for Stream Change Event

As defined in rfc6525#section-6.1.3, SCTP_STREAM_CHANGE_DENIED
and SCTP_STREAM_CHANGE_FAILED should be used instead of
SCTP_ASSOC_CHANGE_DENIED and SCTP_ASSOC_CHANGE_FAILED.

To keep the compatibility, fix it by adding two macros.

Fixes: b444153fb5a6 ("sctp: add support for generating add stream change event notification")
Reported-by: Jianwen Ji <jiji@redhat.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf
David S. Miller [Sun, 4 Nov 2018 02:33:35 +0000 (19:33 -0700)]
Merge git://git./pub/scm/linux/kernel/git/bpf/bpf

Daniel Borkmann says:

====================
pull-request: bpf 2018-11-03

The following pull-request contains BPF updates for your *net* tree.

The main changes are:

1) Fix BPF prog kallsyms and bpf_prog_get_info_by_fd() jited address export
   to not use page start but actual start instead to work correctly with
   profiling e.g. finding hot instructions from stack traces. Also fix latter
   with regards to dumping address and jited len for !multi prog, from Song.

2) Fix bpf_prog_get_info_by_fd() for !root to zero info.nr_jited_func_lens
   instead of leaving the user provided length, from Daniel.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agomlxsw: spectrum: Fix IP2ME CPU policer configuration
Shalom Toledo [Fri, 2 Nov 2018 19:49:15 +0000 (19:49 +0000)]
mlxsw: spectrum: Fix IP2ME CPU policer configuration

The CPU policer used to police packets being trapped via a local route
(IP2ME) was incorrectly configured to police based on bytes per second
instead of packets per second.

Change the policer to police based on packets per second and avoid
packet loss under certain circumstances.

Fixes: 9148e7cf73ce ("mlxsw: spectrum: Add policers for trap groups")
Signed-off-by: Shalom Toledo <shalomt@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoopenvswitch: fix linking without CONFIG_NF_CONNTRACK_LABELS
Arnd Bergmann [Fri, 2 Nov 2018 15:36:55 +0000 (16:36 +0100)]
openvswitch: fix linking without CONFIG_NF_CONNTRACK_LABELS

When CONFIG_CC_OPTIMIZE_FOR_DEBUGGING is enabled, the compiler
fails to optimize out a dead code path, which leads to a link failure:

net/openvswitch/conntrack.o: In function `ovs_ct_set_labels':
conntrack.c:(.text+0x2e60): undefined reference to `nf_connlabels_replace'

In this configuration, we can take a shortcut, and completely
remove the contrack label code. This may also help the regular
optimization.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoqed: fix link config error handling
Arnd Bergmann [Fri, 2 Nov 2018 15:36:24 +0000 (16:36 +0100)]
qed: fix link config error handling

gcc-8 notices that qed_mcp_get_transceiver_data() may fail to
return a result to the caller:

drivers/net/ethernet/qlogic/qed/qed_mcp.c: In function 'qed_mcp_trans_speed_mask':
drivers/net/ethernet/qlogic/qed/qed_mcp.c:1955:2: error: 'transceiver_type' may be used uninitialized in this function [-Werror=maybe-uninitialized]

When an error is returned by qed_mcp_get_transceiver_data(), we
should propagate that to the caller of qed_mcp_trans_speed_mask()
rather than continuing with uninitialized data.

Fixes: c56a8be7e7aa ("qed: Add supported link and advertise link to display in ethtool.")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 4 Nov 2018 01:37:09 +0000 (18:37 -0700)]
Merge branch 'sched-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull scheduler fixes from Ingo Molnar:
 "A memory (under-)allocation fix and a comment fix"

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched/topology: Fix off by one bug
  sched/rt: Update comment in pick_next_task_rt()

6 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 4 Nov 2018 01:25:17 +0000 (18:25 -0700)]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Ingo Molnar:
 "A number of fixes and some late updates:

   - make in_compat_syscall() behavior on x86-32 similar to other
     platforms, this touches a number of generic files but is not
     intended to impact non-x86 platforms.

   - objtool fixes

   - PAT preemption fix

   - paravirt fixes/cleanups

   - cpufeatures updates for new instructions

   - earlyprintk quirk

   - make microcode version in sysfs world-readable (it is already
     world-readable in procfs)

   - minor cleanups and fixes"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  compat: Cleanup in_compat_syscall() callers
  x86/compat: Adjust in_compat_syscall() to generic code under !COMPAT
  objtool: Support GCC 9 cold subfunction naming scheme
  x86/numa_emulation: Fix uniform-split numa emulation
  x86/paravirt: Remove unused _paravirt_ident_32
  x86/mm/pat: Disable preemption around __flush_tlb_all()
  x86/paravirt: Remove GPL from pv_ops export
  x86/traps: Use format string with panic() call
  x86: Clean up 'sizeof x' => 'sizeof(x)'
  x86/cpufeatures: Enumerate MOVDIR64B instruction
  x86/cpufeatures: Enumerate MOVDIRI instruction
  x86/earlyprintk: Add a force option for pciserial device
  objtool: Support per-function rodata sections
  x86/microcode: Make revision and processor flags world-readable

6 years agoMerge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 4 Nov 2018 01:13:43 +0000 (18:13 -0700)]
Merge branch 'perf-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull perf updates and fixes from Ingo Molnar:
 "These are almost all tooling updates: 'perf top', 'perf trace' and
  'perf script' fixes and updates, an UAPI header sync with the merge
  window versions, license marker updates, much improved Sparc support
  from David Miller, and a number of fixes"

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (66 commits)
  perf intel-pt/bts: Calculate cpumode for synthesized samples
  perf intel-pt: Insert callchain context into synthesized callchains
  perf tools: Don't clone maps from parent when synthesizing forks
  perf top: Start display thread earlier
  tools headers uapi: Update linux/if_link.h header copy
  tools headers uapi: Update linux/netlink.h header copy
  tools headers: Sync the various kvm.h header copies
  tools include uapi: Update linux/mmap.h copy
  perf trace beauty: Use the mmap flags table generated from headers
  perf beauty: Wire up the mmap flags table generator to the Makefile
  perf beauty: Add a generator for MAP_ mmap's flag constants
  tools include uapi: Update asound.h copy
  tools arch uapi: Update asm-generic/unistd.h and arm64 unistd.h copies
  tools include uapi: Update linux/fs.h copy
  perf callchain: Honour the ordering of PERF_CONTEXT_{USER,KERNEL,etc}
  perf cs-etm: Correct CPU mode for samples
  perf unwind: Take pgoff into account when reporting elf to libdwfl
  perf top: Do not use overwrite mode by default
  perf top: Allow disabling the overwrite mode
  perf trace: Beautify mount's first pathname arg
  ...

6 years agoMerge branch 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 4 Nov 2018 01:12:09 +0000 (18:12 -0700)]
Merge branch 'irq-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull irq fixes from Ingo Molnar:
 "An irqchip driver fix and a memory (over-)allocation fix"

* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  irqchip/irq-mvebu-sei: Fix a NULL vs IS_ERR() bug in probe function
  irq/matrix: Fix memory overallocation

6 years agosched/topology: Fix off by one bug
Peter Zijlstra [Fri, 2 Nov 2018 13:22:25 +0000 (14:22 +0100)]
sched/topology: Fix off by one bug

With the addition of the NUMA identity level, we increased @level by
one and will run off the end of the array in the distance sort loop.

Fixed: 051f3ca02e46 ("sched/topology: Introduce NUMA identity node sched domain")
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
6 years agoMerge branch 'core/urgent' into x86/urgent, to pick up objtool fix
Ingo Molnar [Sat, 3 Nov 2018 22:42:16 +0000 (23:42 +0100)]
Merge branch 'core/urgent' into x86/urgent, to pick up objtool fix

Signed-off-by: Ingo Molnar <mingo@kernel.org>
6 years agonet: hns3: Fix for out-of-bounds access when setting pfc back pressure
Yunsheng Lin [Fri, 2 Nov 2018 09:47:48 +0000 (17:47 +0800)]
net: hns3: Fix for out-of-bounds access when setting pfc back pressure

The vport should be initialized to hdev->vport for each bp group,
otherwise it will cause out-of-bounds access and bp setting not
correct problem.

[   35.254124] BUG: KASAN: slab-out-of-bounds in hclge_pause_setup_hw+0x2a0/0x3f8 [hclge]
[   35.254126] Read of size 2 at addr ffff803b6651581a by task kworker/0:1/14

[   35.254132] CPU: 0 PID: 14 Comm: kworker/0:1 Not tainted 4.19.0-rc7-hulk+ #85
[   35.254133] Hardware name: Huawei D06/D06, BIOS Hisilicon D06 UEFI RC0 - B052 (V0.52) 09/14/2018
[   35.254141] Workqueue: events work_for_cpu_fn
[   35.254144] Call trace:
[   35.254147]  dump_backtrace+0x0/0x2f0
[   35.254149]  show_stack+0x24/0x30
[   35.254154]  dump_stack+0x110/0x184
[   35.254157]  print_address_description+0x168/0x2b0
[   35.254160]  kasan_report+0x184/0x310
[   35.254162]  __asan_load2+0x7c/0xa0
[   35.254170]  hclge_pause_setup_hw+0x2a0/0x3f8 [hclge]
[   35.254177]  hclge_tm_init_hw+0x794/0x9f0 [hclge]
[   35.254184]  hclge_tm_schd_init+0x48/0x58 [hclge]
[   35.254191]  hclge_init_ae_dev+0x778/0x1168 [hclge]
[   35.254196]  hnae3_register_ae_dev+0x14c/0x298 [hnae3]
[   35.254206]  hns3_probe+0x88/0xa8 [hns3]
[   35.254210]  local_pci_probe+0x7c/0xf0
[   35.254212]  work_for_cpu_fn+0x34/0x50
[   35.254214]  process_one_work+0x4d4/0xa38
[   35.254216]  worker_thread+0x55c/0x8d8
[   35.254219]  kthread+0x1b0/0x1b8
[   35.254222]  ret_from_fork+0x10/0x1c

[   35.254224] The buggy address belongs to the page:
[   35.254228] page:ffff7e00ed994400 count:1 mapcount:0 mapping:0000000000000000 index:0x0 compound_mapcount: 0
[   35.273835] flags: 0xfffff8000008000(head)
[   35.282007] raw: 0fffff8000008000 dead000000000100 dead000000000200 0000000000000000
[   35.282010] raw: 0000000000000000 0000000000000000 00000001ffffffff 0000000000000000
[   35.282012] page dumped because: kasan: bad access detected

[   35.282014] Memory state around the buggy address:
[   35.282017]  ffff803b66515700: fe fe fe fe fe fe fe fe fe fe fe fe fe fe fe fe
[   35.282019]  ffff803b66515780: fe fe fe fe fe fe fe fe fe fe fe fe fe fe fe fe
[   35.282021] >ffff803b66515800: fe fe fe fe fe fe fe fe fe fe fe fe fe fe fe fe
[   35.282022]                             ^
[   35.282024]  ffff803b66515880: fe fe fe fe fe fe fe fe fe fe fe fe fe fe fe fe
[   35.282026]  ffff803b66515900: fe fe fe fe fe fe fe fe fe fe fe fe fe fe fe fe
[   35.282028] ==================================================================
[   35.282029] Disabling lock debugging due to kernel taint
[   35.282747] hclge driver initialization finished.

Fixes: 67bf2541f4b9 ("net: hns3: Fixes the back pressure setting when sriov is enabled")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'net-bql-better-deal-with-GSO'
David S. Miller [Sat, 3 Nov 2018 22:40:01 +0000 (15:40 -0700)]
Merge branch 'net-bql-better-deal-with-GSO'

Eric Dumazet says:

====================
net: bql: better deal with GSO

While BQL bulk dequeue works well for TSO packets, it is
not very efficient as soon as GSO is involved.

On a GSO only workload (UDP or TCP), this patch series
can save about 8 % of cpu cycles on a 40Gbit mlx4 NIC,
by keeping optimal batching, and avoiding expensive
doorbells, qdisc requeues and reschedules.

This patch series :

- Add __netdev_tx_sent_queue() so that drivers
  can implement efficient BQL and xmit_more support.

- Implement a work around in dev_hard_start_xmit()
  for drivers not using __netdev_tx_sent_queue()

- changes mlx4 to use __netdev_tx_sent_queue()

v2: Tariq and Willem feedback addressed.
    added __netdev_tx_sent_queue() (Willem suggestion)
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/mlx4_en: use __netdev_tx_sent_queue()
Eric Dumazet [Wed, 31 Oct 2018 15:39:14 +0000 (08:39 -0700)]
net/mlx4_en: use __netdev_tx_sent_queue()

doorbell only depends on xmit_more and netif_tx_queue_stopped()

Using __netdev_tx_sent_queue() avoids messing with BQL stop flag,
and is more generic.

This patch increases performance on GSO workload by keeping
doorbells to the minimum required.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: do not abort bulk send on BQL status
Eric Dumazet [Wed, 31 Oct 2018 15:39:13 +0000 (08:39 -0700)]
net: do not abort bulk send on BQL status

Before calling dev_hard_start_xmit(), upper layers tried
to cook optimal skb list based on BQL budget.

Problem is that GSO packets can end up comsuming more than
the BQL budget.

Breaking the loop is not useful, since requeued packets
are ahead of any packets still in the qdisc.

It is also more expensive, since next TX completion will
push these packets later, while skbs are not in cpu caches.

It is also a behavior difference with TSO packets, that can
break the BQL limit by a large amount.

Note that drivers should use __netdev_tx_sent_queue()
in order to have optimal xmit_more support, and avoid
useless atomic operations as shown in the following patch.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: bql: add __netdev_tx_sent_queue()
Eric Dumazet [Wed, 31 Oct 2018 15:39:12 +0000 (08:39 -0700)]
net: bql: add __netdev_tx_sent_queue()

When qdisc_run() tries to use BQL budget to bulk-dequeue a batch
of packets, GSO can later transform this list in another list
of skbs, and each skb is sent to device ndo_start_xmit(),
one at a time, with skb->xmit_more being set to one but
for last skb.

Problem is that very often, BQL limit is hit in the middle of
the packet train, forcing dev_hard_start_xmit() to stop the
bulk send and requeue the end of the list.

BQL role is to avoid head of line blocking, making sure
a qdisc can deliver high priority packets before low priority ones.

But there is no way requeued packets can be bypassed by fresh
packets in the qdisc.

Aborting the bulk send increases TX softirqs, and hot cache
lines (after skb_segment()) are wasted.

Note that for TSO packets, we never split a packet in the middle
because of BQL limit being hit.

Drivers should be able to update BQL counters without
flipping/caring about BQL status, if the current skb
has xmit_more set.

Upper layers are ultimately responsible to stop sending another
packet train when BQL limit is hit.

Code template in a driver might look like the following :

send_doorbell = __netdev_tx_sent_queue(tx_queue, nr_bytes, skb->xmit_more);

Note that __netdev_tx_sent_queue() use is not mandatory,
since following patch will change dev_hard_start_xmit()
to not care about BQL status.

But it is highly recommended so that xmit_more full benefits
can be reached (less doorbells sent, and less atomic operations as well)

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge tag 'armsoc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc
Linus Torvalds [Sat, 3 Nov 2018 19:13:57 +0000 (12:13 -0700)]
Merge tag 'armsoc-fixes' of git://git./linux/kernel/git/arm/arm-soc

Pull ARM SoC fixes from Olof Johansson:
 "A few fixes who have come in near or during the merge window:

   - Removal of a VLA usage in Marvell mpp platform code

   - Enable some IPMI options for ARM64 servers by default, helps
     testing

   - Enable PREEMPT on 32-bit ARMv7 defconfig

   - Minor fix for stm32 DT (removal of an unused DMA property)

   - Bugfix for TI OMAP1-based ams-delta (-EINVAL -> IRQ_NOTCONNECTED)"

* tag 'armsoc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc:
  ARM: dts: stm32: update HASH1 dmas property on stm32mp157c
  ARM: orion: avoid VLA in orion_mpp_conf
  ARM: defconfig: Update multi_v7 to use PREEMPT
  arm64: defconfig: Enable some IPMI configs
  soc: ti: QMSS: Fix usage of irq_set_affinity_hint
  ARM: OMAP1: ams-delta: Fix impossible .irq < 0

6 years agoMerge tag 'arm64-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64...
Linus Torvalds [Sat, 3 Nov 2018 17:55:23 +0000 (10:55 -0700)]
Merge tag 'arm64-upstream' of git://git./linux/kernel/git/arm64/linux

Pull more arm64 updates from Catalin Marinas:

 - fix W+X page (mark RO) allocated by the arm64 kprobes code

 - Makefile fix for .i files in out of tree modules

* tag 'arm64-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm64: kprobe: make page to RO mode when allocate it
  arm64: kdump: fix small typo
  arm64: makefile fix build of .i file in external module case

6 years agoMerge tag 'dma-mapping-4.20-2' of git://git.infradead.org/users/hch/dma-mapping
Linus Torvalds [Sat, 3 Nov 2018 17:53:33 +0000 (10:53 -0700)]
Merge tag 'dma-mapping-4.20-2' of git://git.infradead.org/users/hch/dma-mapping

Pull dma-mapping fix from Christoph Hellwig:
 "Avoid compile warnings on non-default arm64 configs"

* tag 'dma-mapping-4.20-2' of git://git.infradead.org/users/hch/dma-mapping:
  arm64: fix warnings without CONFIG_IOMMU_DMA

6 years agoMerge tag 'kbuild-v4.20-2' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy...
Linus Torvalds [Sat, 3 Nov 2018 17:47:33 +0000 (10:47 -0700)]
Merge tag 'kbuild-v4.20-2' of git://git./linux/kernel/git/masahiroy/linux-kbuild

Pull Kbuild updates from Masahiro Yamada:

 - clean-up leftovers in Kconfig files

 - remove stale oldnoconfig and silentoldconfig targets

 - remove unneeded cc-fullversion and cc-name variables

 - improve merge_config script to allow overriding option prefix

* tag 'kbuild-v4.20-2' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild:
  kbuild: remove cc-name variable
  kbuild: replace cc-name test with CONFIG_CC_IS_CLANG
  merge_config.sh: Allow to define config prefix
  kbuild: remove unused cc-fullversion variable
  kconfig: remove silentoldconfig target
  kconfig: remove oldnoconfig target
  powerpc: PCI_MSI needs PCI
  powerpc: remove CONFIG_MCA leftovers
  powerpc: remove CONFIG_PCI_QSPAN
  scsi: aha152x: rename the PCMCIA define

6 years agoMerge tag '4.20-rc1-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Sat, 3 Nov 2018 17:45:55 +0000 (10:45 -0700)]
Merge tag '4.20-rc1-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6

Pull cifs fixes and updates from Steve French:
 "Three small fixes (one Kerberos related, one for stable, and another
  fixes an oops in xfstest 377), two helpful debugging improvements,
  three patches for cifs directio and some minor cleanup"

* tag '4.20-rc1-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6:
  cifs: fix signed/unsigned mismatch on aio_read patch
  cifs: don't dereference smb_file_target before null check
  CIFS: Add direct I/O functions to file_operations
  CIFS: Add support for direct I/O write
  CIFS: Add support for direct I/O read
  smb3: missing defines and structs for reparse point handling
  smb3: allow more detailed protocol info on open files for debugging
  smb3: on kerberos mount if server doesn't specify auth type use krb5
  smb3: add trace point for tree connection
  cifs: fix spelling mistake, EACCESS -> EACCES
  cifs: fix return value for cifs_listxattr

6 years agoMerge branch 's390-qeth-fixes'
David S. Miller [Sat, 3 Nov 2018 17:44:06 +0000 (10:44 -0700)]
Merge branch 's390-qeth-fixes'

Julian Wiedmann says:

====================
s390/qeth: fixes 2018-11-02

please apply one round of qeth fixes for -net.

Patch 1 is rather large and removes a use-after-free hazard from many of our
debug trace entries.
Patch 2 is yet another fix-up for the L3 subdriver's new IP address management
code.
Patch 3 and 4 resolve some fallout from the recent changes wrt how/when qeth
allocates its net_device.
Patch 5 makes sure we don't set reserved bits when building HW commands from
user-provided data.
And finally, patch 6 allows ethtool to play nice with new HW.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agos390/qeth: report 25Gbit link speed
Julian Wiedmann [Fri, 2 Nov 2018 18:04:13 +0000 (19:04 +0100)]
s390/qeth: report 25Gbit link speed

This adds the various identifiers for 25Gbit cards, and wires them up
into sysfs and ethtool.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agos390/qeth: sanitize ARP requests
Julian Wiedmann [Fri, 2 Nov 2018 18:04:12 +0000 (19:04 +0100)]
s390/qeth: sanitize ARP requests

The ARP_{ADD,REMOVE}_ENTRY cmd structs contain reserved fields.
Introduce a common helper that doesn't raw-copy the user-provided data
into the cmd, but only sets those fields that are strictly needed for
the command.

This also sets the correct command length for ARP_REMOVE_ENTRY.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agos390/qeth: fix initial operstate
Julian Wiedmann [Fri, 2 Nov 2018 18:04:11 +0000 (19:04 +0100)]
s390/qeth: fix initial operstate

Setting the carrier 'on' for an unregistered netdevice doesn't update
its operstate. Fix this by delaying the update until the netdevice has
been registered.

Fixes: 91cc98f51e3d ("s390/qeth: remove duplicated carrier state tracking")
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agos390/qeth: unregister netdevice only when registered
Julian Wiedmann [Fri, 2 Nov 2018 18:04:10 +0000 (19:04 +0100)]
s390/qeth: unregister netdevice only when registered

qeth only registers its netdevice when the qeth device is first set
online. Thus a device that has never been set online will trigger
a WARN ("network todo 'hsi%d' but state 0") in unregister_netdev() when
removed.

Fix this by protecting the unregister step, just like we already protect
against repeated registering of the netdevice.

Fixes: d3d1b205e89f ("s390/qeth: allocate netdevice early")
Reported-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agos390/qeth: fix HiperSockets sniffer
Julian Wiedmann [Fri, 2 Nov 2018 18:04:09 +0000 (19:04 +0100)]
s390/qeth: fix HiperSockets sniffer

Sniffing mode for L3 HiperSockets requires that no IP addresses are
registered with the HW. The preferred way to achieve this is for
userspace to delete all the IPs on the interface. But qeth is expected
to also tolerate a configuration where that is not the case, by skipping
the IP registration when in sniffer mode.
Since commit 5f78e29ceebf ("qeth: optimize IP handling in rx_mode callback")
reworked the IP registration logic in the L3 subdriver, this no longer
works. When the qeth device is set online, qeth_l3_recover_ip() now
unconditionally registers all unicast addresses from our internal
IP table.

While we could fix this particular problem by skipping
qeth_l3_recover_ip() on a sniffer device, the more future-proof change
is to skip the IP address registration at the lowest level. This way we
a) catch any future code path that attempts to register an IP address
   without considering the sniffer scenario, and
b) continue to build up our internal IP table, so that if sniffer mode
   is switched off later we can operate just like normal.

Fixes: 5f78e29ceebf ("qeth: optimize IP handling in rx_mode callback")
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agos390/qeth: sanitize strings in debug messages
Julian Wiedmann [Fri, 2 Nov 2018 18:04:08 +0000 (19:04 +0100)]
s390/qeth: sanitize strings in debug messages

As Documentation/s390/s390dbf.txt states quite clearly, using any
pointer in sprinf-formatted s390dbf debug entries is dangerous.
The pointers are dereferenced whenever the trace file is read from.
So if the referenced data has a shorter life-time than the trace file,
any read operation can result in a use-after-free.

So rip out all hazardous use of indirect data, and replace any usage of
dev_name() and such by the Bus ID number.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'work.afs' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
Linus Torvalds [Sat, 3 Nov 2018 17:35:52 +0000 (10:35 -0700)]
Merge branch 'work.afs' of git://git./linux/kernel/git/viro/vfs

Pull 9p fix from Al Viro:
 "Regression fix for net/9p handling of iov_iter; broken by braino when
  switching to iov_iter_is_kvec() et.al., spotted and fixed by Marc"

* 'work.afs' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
  iov_iter: Fix 9p virtio breakage

6 years agoMerge tag 'scsi-misc' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Linus Torvalds [Sat, 3 Nov 2018 17:34:03 +0000 (10:34 -0700)]
Merge tag 'scsi-misc' of git://git./linux/kernel/git/jejb/scsi

Pull more SCSI updates from James Bottomley:
 "This is a set of minor small (and safe changes) that didn't make the
  initial pull request plus some bug fixes"

* tag 'scsi-misc' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: mvsas: Remove set but not used variable 'id'
  scsi: qla2xxx: Remove two arguments from qlafx00_error_entry()
  scsi: qla2xxx: Make sure that qlafx00_ioctl_iosb_entry() initializes 'res'
  scsi: qla2xxx: Remove a set-but-not-used variable
  scsi: qla2xxx: Make qla2x00_sysfs_write_nvram() easier to analyze
  scsi: qla2xxx: Declare local functions 'static'
  scsi: qla2xxx: Improve several kernel-doc headers
  scsi: qla2xxx: Modify fall-through annotations
  scsi: 3w-sas: 3w-9xxx: Use unsigned char for cdb
  scsi: mvsas: Use dma_pool_zalloc
  scsi: target: Don't request modules that aren't even built
  scsi: target: Set response length for REPORT TARGET PORT GROUPS

6 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Sat, 3 Nov 2018 17:21:43 +0000 (10:21 -0700)]
Merge branch 'akpm' (patches from Andrew)

Merge more updates from Andrew Morton:

 - more ocfs2 work

 - various leftovers

* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
  memory_hotplug: cond_resched in __remove_pages
  bfs: add sanity check at bfs_fill_super()
  kernel/sysctl.c: remove duplicated include
  kernel/kexec_file.c: remove some duplicated includes
  mm, thp: consolidate THP gfp handling into alloc_hugepage_direct_gfpmask
  ocfs2: fix clusters leak in ocfs2_defrag_extent()
  ocfs2: dlmglue: clean up timestamp handling
  ocfs2: don't put and assigning null to bh allocated outside
  ocfs2: fix a misuse a of brelse after failing ocfs2_check_dir_entry
  ocfs2: don't use iocb when EIOCBQUEUED returns
  ocfs2: without quota support, avoid calling quota recovery
  ocfs2: remove ocfs2_is_o2cb_active()
  mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings
  include/linux/notifier.h: SRCU: fix ctags
  mm: handle no memcg case in memcg_kmem_charge() properly

6 years agomemory_hotplug: cond_resched in __remove_pages
Michal Hocko [Fri, 2 Nov 2018 22:48:46 +0000 (15:48 -0700)]
memory_hotplug: cond_resched in __remove_pages

We have received a bug report that unbinding a large pmem (>1TB) can
result in a soft lockup:

  NMI watchdog: BUG: soft lockup - CPU#9 stuck for 23s! [ndctl:4365]
  [...]
  Supported: Yes
  CPU: 9 PID: 4365 Comm: ndctl Not tainted 4.12.14-94.40-default #1 SLE12-SP4
  Hardware name: Intel Corporation S2600WFD/S2600WFD, BIOS SE5C620.86B.01.00.0833.051120182255 05/11/2018
  task: ffff9cce7d4410c0 task.stack: ffffbe9eb1bc4000
  RIP: 0010:__put_page+0x62/0x80
  Call Trace:
   devm_memremap_pages_release+0x152/0x260
   release_nodes+0x18d/0x1d0
   device_release_driver_internal+0x160/0x210
   unbind_store+0xb3/0xe0
   kernfs_fop_write+0x102/0x180
   __vfs_write+0x26/0x150
   vfs_write+0xad/0x1a0
   SyS_write+0x42/0x90
   do_syscall_64+0x74/0x150
   entry_SYSCALL_64_after_hwframe+0x3d/0xa2
  RIP: 0033:0x7fd13166b3d0

It has been reported on an older (4.12) kernel but the current upstream
code doesn't cond_resched in the hot remove code at all and the given
range to remove might be really large.  Fix the issue by calling
cond_resched once per memory section.

Link: http://lkml.kernel.org/r/20181031125840.23982-1-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Johannes Thumshirn <jthumshirn@suse.de>
Cc: Dan Williams <dan.j.williams@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agobfs: add sanity check at bfs_fill_super()
Tetsuo Handa [Fri, 2 Nov 2018 22:48:42 +0000 (15:48 -0700)]
bfs: add sanity check at bfs_fill_super()

syzbot is reporting too large memory allocation at bfs_fill_super() [1].
Since file system image is corrupted such that bfs_sb->s_start == 0,
bfs_fill_super() is trying to allocate 8MB of continuous memory. Fix
this by adding a sanity check on bfs_sb->s_start, __GFP_NOWARN and
printf().

[1] https://syzkaller.appspot.com/bug?id=16a87c236b951351374a84c8a32f40edbc034e96

Link: http://lkml.kernel.org/r/1525862104-3407-1-git-send-email-penguin-kernel@I-love.SAKURA.ne.jp
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Reported-by: syzbot <syzbot+71c6b5d68e91149fc8a4@syzkaller.appspotmail.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Tigran Aivazian <aivazian.tigran@gmail.com>
Cc: Matthew Wilcox <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agokernel/sysctl.c: remove duplicated include
Michael Schupikov [Fri, 2 Nov 2018 22:48:38 +0000 (15:48 -0700)]
kernel/sysctl.c: remove duplicated include

Remove one include of <linux/pipe_fs_i.h>.
No functional changes.

Link: http://lkml.kernel.org/r/20181004134223.17735-1-michael@schupikov.de
Signed-off-by: Michael Schupikov <michael@schupikov.de>
Reviewed-by: Richard Weinberger <richard@nod.at>
Acked-by: Luis Chamberlain <mcgrof@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agokernel/kexec_file.c: remove some duplicated includes
zhong jiang [Fri, 2 Nov 2018 22:48:35 +0000 (15:48 -0700)]
kernel/kexec_file.c: remove some duplicated includes

We include kexec.h and slab.h twice in kexec_file.c. It's unnecessary.
hence just remove them.

Link: http://lkml.kernel.org/r/1537498098-19171-1-git-send-email-zhongjiang@huawei.com
Signed-off-by: zhong jiang <zhongjiang@huawei.com>
Reviewed-by: Bhupesh Sharma <bhsharma@redhat.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Baoquan He <bhe@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agomm, thp: consolidate THP gfp handling into alloc_hugepage_direct_gfpmask
Michal Hocko [Fri, 2 Nov 2018 22:48:31 +0000 (15:48 -0700)]
mm, thp: consolidate THP gfp handling into alloc_hugepage_direct_gfpmask

THP allocation mode is quite complex and it depends on the defrag mode.
This complexity is hidden in alloc_hugepage_direct_gfpmask from a large
part currently. The NUMA special casing (namely __GFP_THISNODE) is
however independent and placed in alloc_pages_vma currently. This both
adds an unnecessary branch to all vma based page allocation requests and
it makes the code more complex unnecessarily as well. Not to mention
that e.g. shmem THP used to do the node reclaiming unconditionally
regardless of the defrag mode until recently. This was not only
unexpected behavior but it was also hardly a good default behavior and I
strongly suspect it was just a side effect of the code sharing more than
a deliberate decision which suggests that such a layering is wrong.

Get rid of the thp special casing from alloc_pages_vma and move the
logic to alloc_hugepage_direct_gfpmask. __GFP_THISNODE is applied to the
resulting gfp mask only when the direct reclaim is not requested and
when there is no explicit numa binding to preserve the current logic.

Please note that there's also a slight difference wrt MPOL_BIND now. The
previous code would avoid using __GFP_THISNODE if the local node was
outside of policy_nodemask(). After this patch __GFP_THISNODE is avoided
for all MPOL_BIND policies. So there's a difference that if local node
is actually allowed by the bind policy's nodemask, previously
__GFP_THISNODE would be added, but now it won't be. From the behavior
POV this is still correct because the policy nodemask is used.

Link: http://lkml.kernel.org/r/20180925120326.24392-3-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Alex Williamson <alex.williamson@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Stefan Priebe - Profihost AG <s.priebe@profihost.ag>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agoocfs2: fix clusters leak in ocfs2_defrag_extent()
Larry Chen [Fri, 2 Nov 2018 22:48:27 +0000 (15:48 -0700)]
ocfs2: fix clusters leak in ocfs2_defrag_extent()

ocfs2_defrag_extent() might leak allocated clusters.  When the file
system has insufficient space, the number of claimed clusters might be
less than the caller wants.  If that happens, the original code might
directly commit the transaction without returning clusters.

This patch is based on code in ocfs2_add_clusters_in_btree().

[akpm@linux-foundation.org: include localalloc.h, reduce scope of data_ac]
Link: http://lkml.kernel.org/r/20180904041621.16874-3-lchen@suse.com
Signed-off-by: Larry Chen <lchen@suse.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Cc: Changwei Ge <ge.changwei@h3c.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agoocfs2: dlmglue: clean up timestamp handling
Arnd Bergmann [Fri, 2 Nov 2018 22:48:23 +0000 (15:48 -0700)]
ocfs2: dlmglue: clean up timestamp handling

The handling of timestamps outside of the 1970..2038 range in the dlm
glue is rather inconsistent: on 32-bit architectures, this has always
wrapped around to negative timestamps in the 1902..1969 range, while on
64-bit kernels all timestamps are interpreted as positive 34 bit numbers
in the 1970..2514 year range.

Now that the VFS code handles 64-bit timestamps on all architectures, we
can make the behavior more consistent here, and return the same result
that we had on 64-bit already, making the file system y2038 safe in the
process.  Outside of dlmglue, it already uses 64-bit on-disk timestamps
anway, so that part is fine.

For consistency, I'm changing ocfs2_pack_timespec() to clamp anything
outside of the supported range to the minimum and maximum values.  This
avoids a possible ambiguity of values before 1970 in particular, which
used to be interpreted as times at the end of the 2514 range previously.

Link: http://lkml.kernel.org/r/20180619155826.4106487-1-arnd@arndb.de
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Cc: Changwei Ge <ge.changwei@h3c.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agoocfs2: don't put and assigning null to bh allocated outside
Changwei Ge [Fri, 2 Nov 2018 22:48:19 +0000 (15:48 -0700)]
ocfs2: don't put and assigning null to bh allocated outside

ocfs2_read_blocks() and ocfs2_read_blocks_sync() are both used to read
several blocks from disk.  Currently, the input argument *bhs* can be
NULL or NOT.  It depends on the caller's behavior.  If the function
fails in reading blocks from disk, the corresponding bh will be assigned
to NULL and put.

Obviously, above process for non-NULL input bh is not appropriate.
Because the caller doesn't even know its bhs are put and re-assigned.

If buffer head is managed by caller, ocfs2_read_blocks and
ocfs2_read_blocks_sync() should not evaluate it to NULL.  It will cause
caller accessing illegal memory, thus crash.

Link: http://lkml.kernel.org/r/HK2PR06MB045285E0F4FBB561F9F2F9B3D5680@HK2PR06MB0452.apcprd06.prod.outlook.com
Signed-off-by: Changwei Ge <ge.changwei@h3c.com>
Reviewed-by: Guozhonghua <guozhonghua@h3c.com>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Cc: Changwei Ge <ge.changwei@h3c.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agoocfs2: fix a misuse a of brelse after failing ocfs2_check_dir_entry
Changwei Ge [Fri, 2 Nov 2018 22:48:15 +0000 (15:48 -0700)]
ocfs2: fix a misuse a of brelse after failing ocfs2_check_dir_entry

Somehow, file system metadata was corrupted, which causes
ocfs2_check_dir_entry() to fail in function ocfs2_dir_foreach_blk_el().

According to the original design intention, if above happens we should
skip the problematic block and continue to retrieve dir entry.  But
there is obviouse misuse of brelse around related code.

After failure of ocfs2_check_dir_entry(), current code just moves to
next position and uses the problematic buffer head again and again
during which the problematic buffer head is released for multiple times.
I suppose, this a serious issue which is long-lived in ocfs2.  This may
cause other file systems which is also used in a the same host insane.

So we should also consider about bakcporting this patch into linux
-stable.

Link: http://lkml.kernel.org/r/HK2PR06MB045211675B43EED794E597B6D56E0@HK2PR06MB0452.apcprd06.prod.outlook.com
Signed-off-by: Changwei Ge <ge.changwei@h3c.com>
Suggested-by: Changkuo Shi <shi.changkuo@h3c.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agoocfs2: don't use iocb when EIOCBQUEUED returns
Changwei Ge [Fri, 2 Nov 2018 22:48:11 +0000 (15:48 -0700)]
ocfs2: don't use iocb when EIOCBQUEUED returns

When -EIOCBQUEUED returns, it means that aio_complete() will be called
from dio_complete(), which is an asynchronous progress against
write_iter.  Generally, IO is a very slow progress than executing
instruction, but we still can't take the risk to access a freed iocb.

And we do face a BUG crash issue.  Using the crash tool, iocb is
obviously freed already.

  crash> struct -x kiocb ffff881a350f5900
  struct kiocb {
    ki_filp = 0xffff881a350f5a80,
    ki_pos = 0x0,
    ki_complete = 0x0,
    private = 0x0,
    ki_flags = 0x0
  }

And the backtrace shows:
  ocfs2_file_write_iter+0xcaa/0xd00 [ocfs2]
  aio_run_iocb+0x229/0x2f0
  do_io_submit+0x291/0x540
  SyS_io_submit+0x10/0x20
  system_call_fastpath+0x16/0x75

Link: http://lkml.kernel.org/r/1523361653-14439-1-git-send-email-ge.changwei@h3c.com
Signed-off-by: Changwei Ge <ge.changwei@h3c.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agoocfs2: without quota support, avoid calling quota recovery
Guozhonghua [Fri, 2 Nov 2018 22:48:07 +0000 (15:48 -0700)]
ocfs2: without quota support, avoid calling quota recovery

During one dead node's recovery by other node, quota recovery work will
be queued.  We should avoid calling quota when it is not supported, so
check the quota flags.

Link: http://lkml.kernel.org/r/71604351584F6A4EBAE558C676F37CA401071AC9FB@H3CMLB12-EX.srv.huawei-3com.com
Signed-off-by: guozhonghua <guozhonghua@h3c.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Cc: Changwei Ge <ge.changwei@h3c.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agoocfs2: remove ocfs2_is_o2cb_active()
Gang He [Fri, 2 Nov 2018 22:48:03 +0000 (15:48 -0700)]
ocfs2: remove ocfs2_is_o2cb_active()

Remove ocfs2_is_o2cb_active().  We have similar functions to identify
which cluster stack is being used via osb->osb_cluster_stack.

Secondly, the current implementation of ocfs2_is_o2cb_active() is not
totally safe.  Based on the design of stackglue, we need to get
ocfs2_stack_lock before using ocfs2_stack related data structures, and
that active_stack pointer can be NULL in the case of mount failure.

Link: http://lkml.kernel.org/r/1495441079-11708-1-git-send-email-ghe@suse.com
Signed-off-by: Gang He <ghe@suse.com>
Reviewed-by: Joseph Qi <jiangqi903@gmail.com>
Reviewed-by: Eric Ren <zren@suse.com>
Acked-by: Changwei Ge <ge.changwei@h3c.com>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agomm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings
Andrea Arcangeli [Fri, 2 Nov 2018 22:47:59 +0000 (15:47 -0700)]
mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings

THP allocation might be really disruptive when allocated on NUMA system
with the local node full or hard to reclaim.  Stefan has posted an
allocation stall report on 4.12 based SLES kernel which suggests the
same issue:

  kvm: page allocation stalls for 194572ms, order:9, mode:0x4740ca(__GFP_HIGHMEM|__GFP_IO|__GFP_FS|__GFP_COMP|__GFP_NOMEMALLOC|__GFP_HARDWALL|__GFP_THISNODE|__GFP_MOVABLE|__GFP_DIRECT_RECLAIM), nodemask=(null)
  kvm cpuset=/ mems_allowed=0-1
  CPU: 10 PID: 84752 Comm: kvm Tainted: G        W 4.12.0+98-ph <a href="/view.php?id=1" title="[geschlossen] Integration Ramdisk" class="resolved">0000001</a> SLE15 (unreleased)
  Hardware name: Supermicro SYS-1029P-WTRT/X11DDW-NT, BIOS 2.0 12/05/2017
  Call Trace:
   dump_stack+0x5c/0x84
   warn_alloc+0xe0/0x180
   __alloc_pages_slowpath+0x820/0xc90
   __alloc_pages_nodemask+0x1cc/0x210
   alloc_pages_vma+0x1e5/0x280
   do_huge_pmd_wp_page+0x83f/0xf00
   __handle_mm_fault+0x93d/0x1060
   handle_mm_fault+0xc6/0x1b0
   __do_page_fault+0x230/0x430
   do_page_fault+0x2a/0x70
   page_fault+0x7b/0x80
   [...]
  Mem-Info:
  active_anon:126315487 inactive_anon:1612476 isolated_anon:5
   active_file:60183 inactive_file:245285 isolated_file:0
   unevictable:15657 dirty:286 writeback:1 unstable:0
   slab_reclaimable:75543 slab_unreclaimable:2509111
   mapped:81814 shmem:31764 pagetables:370616 bounce:0
   free:32294031 free_pcp:6233 free_cma:0
  Node 0 active_anon:254680388kB inactive_anon:1112760kB active_file:240648kB inactive_file:981168kB unevictable:13368kB isolated(anon):0kB isolated(file):0kB mapped:280240kB dirty:1144kB writeback:0kB shmem:95832kB shmem_thp: 0kB shmem_pmdmapped: 0kB anon_thp: 81225728kB writeback_tmp:0kB unstable:0kB all_unreclaimable? no
  Node 1 active_anon:250583072kB inactive_anon:5337144kB active_file:84kB inactive_file:0kB unevictable:49260kB isolated(anon):20kB isolated(file):0kB mapped:47016kB dirty:0kB writeback:4kB shmem:31224kB shmem_thp: 0kB shmem_pmdmapped: 0kB anon_thp: 31897600kB writeback_tmp:0kB unstable:0kB all_unreclaimable? no

The defrag mode is "madvise" and from the above report it is clear that
the THP has been allocated for MADV_HUGEPAGA vma.

Andrea has identified that the main source of the problem is
__GFP_THISNODE usage:

: The problem is that direct compaction combined with the NUMA
: __GFP_THISNODE logic in mempolicy.c is telling reclaim to swap very
: hard the local node, instead of failing the allocation if there's no
: THP available in the local node.
:
: Such logic was ok until __GFP_THISNODE was added to the THP allocation
: path even with MPOL_DEFAULT.
:
: The idea behind the __GFP_THISNODE addition, is that it is better to
: provide local memory in PAGE_SIZE units than to use remote NUMA THP
: backed memory. That largely depends on the remote latency though, on
: threadrippers for example the overhead is relatively low in my
: experience.
:
: The combination of __GFP_THISNODE and __GFP_DIRECT_RECLAIM results in
: extremely slow qemu startup with vfio, if the VM is larger than the
: size of one host NUMA node. This is because it will try very hard to
: unsuccessfully swapout get_user_pages pinned pages as result of the
: __GFP_THISNODE being set, instead of falling back to PAGE_SIZE
: allocations and instead of trying to allocate THP on other nodes (it
: would be even worse without vfio type1 GUP pins of course, except it'd
: be swapping heavily instead).

Fix this by removing __GFP_THISNODE for THP requests which are
requesting the direct reclaim.  This effectivelly reverts 5265047ac301
on the grounds that the zone/node reclaim was known to be disruptive due
to premature reclaim when there was memory free.  While it made sense at
the time for HPC workloads without NUMA awareness on rare machines, it
was ultimately harmful in the majority of cases.  The existing behaviour
is similar, if not as widespare as it applies to a corner case but
crucially, it cannot be tuned around like zone_reclaim_mode can.  The
default behaviour should always be to cause the least harm for the
common case.

If there are specialised use cases out there that want zone_reclaim_mode
in specific cases, then it can be built on top.  Longterm we should
consider a memory policy which allows for the node reclaim like behavior
for the specific memory ranges which would allow a

[1] http://lkml.kernel.org/r/20180820032204.9591-1-aarcange@redhat.com

Mel said:

: Both patches look correct to me but I'm responding to this one because
: it's the fix.  The change makes sense and moves further away from the
: severe stalling behaviour we used to see with both THP and zone reclaim
: mode.
:
: I put together a basic experiment with usemem configured to reference a
: buffer multiple times that is 80% the size of main memory on a 2-socket
: box with symmetric node sizes and defrag set to "always".  The defrag
: setting is not the default but it would be functionally similar to
: accessing a buffer with madvise(MADV_HUGEPAGE).  Usemem is configured to
: reference the buffer multiple times and while it's not an interesting
: workload, it would be expected to complete reasonably quickly as it fits
: within memory.  The results were;
:
: usemem
:                                   vanilla           noreclaim-v1
: Amean     Elapsd-1       42.78 (   0.00%)       26.87 (  37.18%)
: Amean     Elapsd-3       27.55 (   0.00%)        7.44 (  73.00%)
: Amean     Elapsd-4        5.72 (   0.00%)        5.69 (   0.45%)
:
: This shows the elapsed time in seconds for 1 thread, 3 threads and 4
: threads referencing buffers 80% the size of memory.  With the patches
: applied, it's 37.18% faster for the single thread and 73% faster with two
: threads.  Note that 4 threads showing little difference does not indicate
: the problem is related to thread counts.  It's simply the case that 4
: threads gets spread so their workload mostly fits in one node.
:
: The overall view from /proc/vmstats is more startling
:
:                          4.19.0-rc1  4.19.0-rc1
:                             vanillanoreclaim-v1r1
: Minor Faults               35593425      708164
: Major Faults                 484088          36
: Swap Ins                    3772837           0
: Swap Outs                   3932295           0
:
: Massive amounts of swap in/out without the patch
:
: Direct pages scanned        6013214           0
: Kswapd pages scanned              0           0
: Kswapd pages reclaimed            0           0
: Direct pages reclaimed      4033009           0
:
: Lots of reclaim activity without the patch
:
: Kswapd efficiency              100%        100%
: Kswapd velocity               0.000       0.000
: Direct efficiency               67%        100%
: Direct velocity           11191.956       0.000
:
: Mostly from direct reclaim context as you'd expect without the patch.
:
: Page writes by reclaim  3932314.000       0.000
: Page writes file                 19           0
: Page writes anon            3932295           0
: Page reclaim immediate        42336           0
:
: Writes from reclaim context is never good but the patch eliminates it.
:
: We should never have default behaviour to thrash the system for such a
: basic workload.  If zone reclaim mode behaviour is ever desired but on a
: single task instead of a global basis then the sensible option is to build
: a mempolicy that enforces that behaviour.

This was a severe regression compared to previous kernels that made
important workloads unusable and it starts when __GFP_THISNODE was
added to THP allocations under MADV_HUGEPAGE.  It is not a significant
risk to go to the previous behavior before __GFP_THISNODE was added, it
worked like that for years.

This was simply an optimization to some lucky workloads that can fit in
a single node, but it ended up breaking the VM for others that can't
possibly fit in a single node, so going back is safe.

[mhocko@suse.com: rewrote the changelog based on the one from Andrea]
Link: http://lkml.kernel.org/r/20180925120326.24392-2-mhocko@kernel.org
Fixes: 5265047ac301 ("mm, thp: really limit transparent hugepage allocation to local node")
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reported-by: Stefan Priebe <s.priebe@profihost.ag>
Debugged-by: Andrea Arcangeli <aarcange@redhat.com>
Reported-by: Alex Williamson <alex.williamson@redhat.com>
Reviewed-by: Mel Gorman <mgorman@techsingularity.net>
Tested-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: <stable@vger.kernel.org> [4.1+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agoinclude/linux/notifier.h: SRCU: fix ctags
Sam Protsenko [Fri, 2 Nov 2018 22:47:53 +0000 (15:47 -0700)]
include/linux/notifier.h: SRCU: fix ctags

ctags indexing ("make tags" command) throws this warning:

    ctags: Warning: include/linux/notifier.h:125:
    null expansion of name pattern "\1"

This is the result of DEFINE_PER_CPU() macro expansion.  Fix that by
getting rid of line break.

Similar fix was already done in commit 25528213fe9f ("tags: Fix
DEFINE_PER_CPU expansions"), but this one probably wasn't noticed.

Link: http://lkml.kernel.org/r/20181030202808.28027-1-semen.protsenko@linaro.org
Fixes: 9c80172b902d ("kernel/SRCU: provide a static initializer")
Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Andy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agomm: handle no memcg case in memcg_kmem_charge() properly
Roman Gushchin [Fri, 2 Nov 2018 22:47:49 +0000 (15:47 -0700)]
mm: handle no memcg case in memcg_kmem_charge() properly

Mike Galbraith reported a regression caused by the commit 9b6f7e163cd0
("mm: rework memcg kernel stack accounting") on a system with
"cgroup_disable=memory" boot option: the system panics with the following
stack trace:

  BUG: unable to handle kernel NULL pointer dereference at 00000000000000f8
  PGD 0 P4D 0
  Oops: 0002 [#1] PREEMPT SMP PTI
  CPU: 0 PID: 1 Comm: systemd Not tainted 4.19.0-preempt+ #410
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS ?-20180531_142017-buildhw-08.phx2.fed4
  RIP: 0010:page_counter_try_charge+0x22/0xc0
  Code: 41 5d c3 c3 0f 1f 40 00 0f 1f 44 00 00 48 85 ff 0f 84 a7 00 00 00 41 56 48 89 f8 49 89 fe 49
  Call Trace:
   try_charge+0xcb/0x780
   memcg_kmem_charge_memcg+0x28/0x80
   memcg_kmem_charge+0x8b/0x1d0
   copy_process.part.41+0x1ca/0x2070
   _do_fork+0xd7/0x3d0
   do_syscall_64+0x5a/0x180
   entry_SYSCALL_64_after_hwframe+0x49/0xbe

The problem occurs because get_mem_cgroup_from_current() returns the NULL
pointer if memory controller is disabled.  Let's check if this is a case
at the beginning of memcg_kmem_charge() and just return 0 if
mem_cgroup_disabled() returns true.  This is how we handle this case in
many other places in the memory controller code.

Link: http://lkml.kernel.org/r/20181029215123.17830-1-guro@fb.com
Fixes: 9b6f7e163cd0 ("mm: rework memcg kernel stack accounting")
Signed-off-by: Roman Gushchin <guro@fb.com>
Reported-by: Mike Galbraith <efault@gmx.de>
Acked-by: Rik van Riel <riel@surriel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: Shakeel Butt <shakeelb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agonetfilter: conntrack: fix calculation of next bucket number in early_drop
Vasily Khoruzhick [Thu, 25 Oct 2018 19:15:43 +0000 (12:15 -0700)]
netfilter: conntrack: fix calculation of next bucket number in early_drop

If there's no entry to drop in bucket that corresponds to the hash,
early_drop() should look for it in other buckets. But since it increments
hash instead of bucket number, it actually looks in the same bucket 8
times: hsize is 16k by default (14 bits) and hash is 32-bit value, so
reciprocal_scale(hash, hsize) returns the same value for hash..hash+7 in
most cases.

Fix it by increasing bucket number instead of hash and rename _hash
to bucket to avoid future confusion.

Fixes: 3e86638e9a0b ("netfilter: conntrack: consider ct netns in early_drop logic")
Cc: <stable@vger.kernel.org> # v4.7+
Signed-off-by: Vasily Khoruzhick <vasilykh@arista.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
6 years agonetfilter: nft_compat: ebtables 'nat' table is normal chain type
Florian Westphal [Fri, 2 Nov 2018 10:33:37 +0000 (11:33 +0100)]
netfilter: nft_compat: ebtables 'nat' table is normal chain type

Unlike ip(6)tables, the ebtables nat table has no special properties.
This bug causes 'ebtables -A' to fail when using a target such as
'snat' (ebt_snat target sets ".table = "nat"').  Targets that have
no table restrictions work fine.

Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
6 years agonetfilter: nfnetlink_cttimeout: pass default timeout policy to obj_to_nlattr
Pablo Neira Ayuso [Thu, 1 Nov 2018 23:14:00 +0000 (00:14 +0100)]
netfilter: nfnetlink_cttimeout: pass default timeout policy to obj_to_nlattr

Otherwise, we hit a NULL pointer deference since handlers always assume
default timeout policy is passed.

  netlink: 24 bytes leftover after parsing attributes in process `syz-executor2'.
  kasan: CONFIG_KASAN_INLINE enabled
  kasan: GPF could be caused by NULL-ptr deref or user memory access
  general protection fault: 0000 [#1] PREEMPT SMP KASAN
  CPU: 0 PID: 9575 Comm: syz-executor1 Not tainted 4.19.0+ #312
  Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
  RIP: 0010:icmp_timeout_obj_to_nlattr+0x77/0x170 net/netfilter/nf_conntrack_proto_icmp.c:297

Fixes: c779e849608a ("netfilter: conntrack: remove get_timeout() indirection")
Reported-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
6 years agonetfilter: conntrack: add nf_{tcp,udp,sctp,icmp,dccp,icmpv6,generic}_pernet()
Pablo Neira Ayuso [Thu, 1 Nov 2018 23:11:34 +0000 (00:11 +0100)]
netfilter: conntrack: add nf_{tcp,udp,sctp,icmp,dccp,icmpv6,generic}_pernet()

Expose these functions to access conntrack protocol tracker netns area,
nfnetlink_cttimeout needs this.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
6 years agonetfilter: ipset: Fix calling ip_set() macro at dumping
Jozsef Kadlecsik [Tue, 30 Oct 2018 21:43:42 +0000 (22:43 +0100)]
netfilter: ipset: Fix calling ip_set() macro at dumping

The ip_set() macro is called when either ip_set_ref_lock held only
or no lock/nfnl mutex is held at dumping. Take this into account
properly. Also, use Pablo's suggestion to use rcu_dereference_raw(),
the ref_netlink protects the set.

Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
6 years agonetfilter: xt_IDLETIMER: add sysfs filename checking routine
Taehee Yoo [Sat, 20 Oct 2018 15:00:08 +0000 (00:00 +0900)]
netfilter: xt_IDLETIMER: add sysfs filename checking routine

When IDLETIMER rule is added, sysfs file is created under
/sys/class/xt_idletimer/timers/
But some label name shouldn't be used.
".", "..", "power", "uevent", "subsystem", etc...
So that sysfs filename checking routine is needed.

test commands:
   %iptables -I INPUT -j IDLETIMER --timeout 1 --label "power"

splat looks like:
[95765.423132] sysfs: cannot create duplicate filename '/devices/virtual/xt_idletimer/timers/power'
[95765.433418] CPU: 0 PID: 8446 Comm: iptables Not tainted 4.19.0-rc6+ #20
[95765.449755] Call Trace:
[95765.449755]  dump_stack+0xc9/0x16b
[95765.449755]  ? show_regs_print_info+0x5/0x5
[95765.449755]  sysfs_warn_dup+0x74/0x90
[95765.449755]  sysfs_add_file_mode_ns+0x352/0x500
[95765.449755]  sysfs_create_file_ns+0x179/0x270
[95765.449755]  ? sysfs_add_file_mode_ns+0x500/0x500
[95765.449755]  ? idletimer_tg_checkentry+0x3e5/0xb1b [xt_IDLETIMER]
[95765.449755]  ? rcu_read_lock_sched_held+0x114/0x130
[95765.449755]  ? __kmalloc_track_caller+0x211/0x2b0
[95765.449755]  ? memcpy+0x34/0x50
[95765.449755]  idletimer_tg_checkentry+0x4e2/0xb1b [xt_IDLETIMER]
[ ... ]

Fixes: 0902b469bd25 ("netfilter: xtables: idletimer target implementation")
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>