The introduction of this schedule point was done in commit
2ba2506c ("[NET]: Add preemption point in qdisc_run")
at a time the loop was not bounded.

Then later in commit d5b8aa1d ("net_sched: fix dequeuer fairness")
we added a limit on the number of packets.

Now is the time to remove the schedule point, since the default
limit of 64 packets matches the number of packets a typical NAPI
poll can process in a row.

This solves a latency problem for most TCP receivers under moderate load :

1) host receives a packet.
   NET_RX_SOFTIRQ is raised by NIC hard IRQ handler

2) __do_softirq() does its first loop, handling NET_RX_SOFTIRQ
   and calling the driver napi->loop() function

3) TCP stores the skb in socket receive queue:

4) TCP calls sk->sk_data_ready() and wakeups a user thread
   waiting for EPOLLIN (as a result, need_resched() might now be true)

5) TCP cooks an ACK and sends it.

6) qdisc_run() processes one packet from qdisc, and sees need_resched(),
   this raises NET_TX_SOFTIRQ (even if there are no more packets in
   the qdisc)

Then we go back to the __do_softirq() in 2), and we see that new
softirqs were raised. Since need_resched() is true, we end up waking
ksoftirqd in this path :

    if (pending) {
            if (time_before(jiffies, end) && !need_resched() &&
                --max_restart)
                    goto restart;

            wakeup_softirqd();
    }

So we have many wakeups of ksoftirqd kernel threads,
and more calls to qdisc_run() with associated lock overhead.

Note that another way to solve the issue would be to change TCP
to first send the ACK packet, then signal the EPOLLIN,
but this changes P99 latencies, as sending the ACK packet
can add a long delay.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The struct __dsa_skb_cb is supposed to span the entire 48-byte skb
control block, while the struct dsa_skb_cb only the portion of it which
is used by the DSA core (the rest is available as private data to
drivers).

The DSA_SKB_CB and __DSA_SKB_CB helpers are supposed to help retrieve
this pointer based on a skb, but it turns out there is nobody directly
interested in the struct __dsa_skb_cb in the kernel. So remove it.
Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Vladimir Oltean says:

====================
SJA1105 DSA coding style cleanup

This series provides some mechanical cleanup patches related to function
names and prototypes.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

The most commonly called function in the driver is long due for a
rename. The "packed" word is redundant (it doesn't make sense to
transfer an unpacked structure, since that is in CPU endianness yadda
yadda), and the "spi" word is also redundant since argument 2 of the
function is SPI_READ or SPI_WRITE.

As for the sja1105_spi_send_long_packed_buf function, it is only being
used from sja1105_spi.c, so remove its global prototype.
Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Having a function that takes a variable number of unpacked bytes which
it generically calls an "int" is confusing and makes auditing patches
next to impossible.

We only use spi_send_int with the int sizes of 32 and 64 bits. So just
make the spi_send_int function less generic and replace it with the
appropriate two explicit functions, which can now type-check the int
pointer type.

Note that there is still a small weirdness in the u32 function, which
has to convert it to a u64 temporary. This is because of how the packing
API works at the moment, but the weirdness is at least hidden from
callers of sja1105_xfer_u32 now.
Suggested-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Let the compiler decide.
Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Paulo Alcantara says:

====================
Experimental SMB rootfs support

This patch series enables Linux to mount root file systems over the
network by utilizing SMB protocol.

Upstream commit 8eecd1c2 ("cifs: Add support for root file
systems") introduced a new CONFIG_CIFS_ROOT option, a virtual device
(Root_CIFS) and a kernel cmdline parameter "cifsroot=" which tells the
kernel to actually mount the root filesystem over a SMB share.

The feature relies on ipconfig to set up the network prior to mounting
the rootfs, so when it is set along with "cifsroot=" parameter:

    (1) cifs_root_setup() parses all necessary data out of "cifsroot="
    parameter for the init process know how to mount the SMB rootfs
    (e.g. SMB server address, mount options).

    (2) If DHCP failed for some reason in ipconfig, we keep retrying
    forever as we have nowhere to go for NFS or SMB root
    filesystems (see PATCH 2/2). Otherwise go to (3).

    (3) mount_cifs_root() is then called by mount_root() (ROOT_DEV ==
    Root_CIFS), retrieves early parsed data from (1), then attempt to
    mount SMB rootfs by CIFSROOT_RETRY_MAX times at most (see PATCH
    1/2).

    (4) If all attempts failed, fall back to floppy drive, otherwise
    continue the boot process with rootfs mounted over a SMB share.

My idea was to keep the same behavior of nfsroot - as it seems to work
for most users so far.

For more information on how this feature works, see
Documentation/filesystems/cifs/cifsroot.txt.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

The experimental root file system support in cifs.ko relies on
ipconfig to set up the network stack and then accessing the SMB share
that contains the rootfs files.
Signed-off-by: Paulo Alcantara (SUSE) <pc@cjr.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add a new virtual device named /dev/cifs (0xfe) to tell the kernel to
mount the root file system over the network by using SMB protocol.

cifs_root_data() will be responsible to retrieve the parsed
information of the new command-line option (cifsroot=) and then call
do_mount_root() with the appropriate mount options for cifs.ko.
Signed-off-by: Paulo Alcantara (SUSE) <pc@cjr.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>

Shannon Nelson says:

====================
ionic: driver updates

These patches are a few updates to clean up some code
issues and add an ethtool feature.

v3: drop the Fixes tags as they really aren't fixing bugs
    simplify ionic_lif_quiesce() as no return is necessary

v2: add cover letter
    edit a couple of patch descriptions for clarity
      and add Fixes: tags
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Even though we've already turned off the queue activity with
the ionic_qcq_disable(), we need to wait for any device queues
that are processing packets to drain down before we try to
flush our packets and tear down the queues.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>

Wire up the --set-fec and --show-fec features in the ethtool
callbacks and pull the related code out of set_link_ksettings.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>

The user's request for an interrupt coalescing value gets
translated into a hardware value to be used with the NIC,
and was getting reported back based on the hw value, which,
due to hw tic resolution, could be reported as a different
number than what the user originally asked for.  This code
now tracks both the user request and what was put into the
hardware so we can report back to the user what they
requested.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>

Replace the open-coded ionic_wait_for_bit() with the
kernel's wait_on_bit_lock().
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>

There is no need for a goto in this bit of code.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>

Jiri Pirko says:

====================
net: introduce per-netns netdevice notifiers and use them in mlxsw

Some drivers, like mlxsw, are not interested in notifications coming in
for netdevices from other network namespaces. So introduce per-netns
notifiers and allow to reduce overhead by listening only for
notifications from the same netns.

This is also a preparation for upcoming patchset "devlink: allow devlink
instances to change network namespace". This resolves deadlock during
reload mlxsw into initial netns made possible by
328fbe74 ("net: Close race between {un, }register_netdevice_notifier() and setup_net()/cleanup_net()").
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

The mlxsw_sp instance is not interested in events happening in other
network namespaces. So use "_net" variants for netdevice notifier
registration/unregistration and get only events which are happening in
the net the instance is in.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Often the code for example in drivers is interested in getting notifier
call only from certain network namespace. In addition to the existing
global netdevice notifier chain introduce per-netns chains and allow
users to register to that. Eventually this would eliminate unnecessary
overhead in case there are many netdevices in many network namespaces.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Push iterations over net namespaces and netdevices from
register_netdevice_notifier() and unregister_netdevice_notifier()
into helper functions. Along with that introduce continue_reverse macros
to make the code a bit nicer allowing to get rid of "last" marks.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Use a guard clause in tx_bottom() to reduce the indentation of the
do-while loop.

Also, fix a couple of spelling and grammatical mistakes in the
r8152_csum_workaround() function comment.

Change-Id: I460befde150ad92248fd85b0f189ec2df2ab8431
Signed-off-by: Prashant Malani <pmalani@chromium.org>
Reviewed-by: Grant Grundler <grundler@chromium.org>
Acked-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds support for MSG_PEEK. In such a case, packets are not
removed from the rx_queue and credit updates are not sent.
Signed-off-by: Matias Ezequiel Vara Larsen <matiasevara@gmail.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Tested-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Fix NULL-pointer dereference on tty open due to a failure to handle a
missing interrupt-in endpoint when probing modem ports:

	BUG: kernel NULL pointer dereference, address: 0000000000000006
	...
	RIP: 0010:tiocmget_submit_urb+0x1c/0xe0 [hso]
	...
	Call Trace:
	hso_start_serial_device+0xdc/0x140 [hso]
	hso_serial_open+0x118/0x1b0 [hso]
	tty_open+0xf1/0x490

Fixes: 542f5482 ("tty: Modem functions for the HSO driver")
Signed-off-by: Johan Hovold <johan@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Convert Renesas Electronics SH EtherMAC bindings documentation to
json-schema.  Also name bindings documentation file according to the compat
string being documented.
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Reviewed-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Adopt and integrate the feature to pass the MAC address via device tree
from asix_device.c (03fc5d4f) also to other ax88179 based asix chips.
E.g. the bootloader fills in local-mac-address and the driver will then
pick up and use this MAC address.
Signed-off-by: Peter Fink <pfink@christ-es.de>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

Just put related code together to ease code reading: the memcpy() is
related to the nla_reserve().
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Jiri Pirko says:

====================
net: introduce alternative names for network interfaces

In the past, there was repeatedly discussed the IFNAMSIZ (16) limit for
netdevice name length. Now when we have PF and VF representors
with port names like "pfXvfY", it became quite common to hit this limit:
0123456789012345
enp131s0f1npf0vf6
enp131s0f1npf0vf22

Udev cannot rename these interfaces out-of-the-box and user needs to
create custom rules to handle them.

Also, udev has multiple schemes of netdev names. From udev code:
 * Type of names:
 *   b<number>                             - BCMA bus core number
 *   c<bus_id>                             - bus id of a grouped CCW or CCW device,
 *                                           with all leading zeros stripped [s390]
 *   o<index>[n<phys_port_name>|d<dev_port>]
 *                                         - on-board device index number
 *   s<slot>[f<function>][n<phys_port_name>|d<dev_port>]
 *                                         - hotplug slot index number
 *   x<MAC>                                - MAC address
 *   [P<domain>]p<bus>s<slot>[f<function>][n<phys_port_name>|d<dev_port>]
 *                                         - PCI geographical location
 *   [P<domain>]p<bus>s<slot>[f<function>][u<port>][..][c<config>][i<interface>]
 *                                         - USB port number chain
 *   v<slot>                               - VIO slot number (IBM PowerVM)
 *   a<vendor><model>i<instance>           - Platform bus ACPI instance id
 *   i<addr>n<phys_port_name>              - Netdevsim bus address and port name

One device can be often renamed by multiple patterns at the
same time (e.g. pci address/mac).

This patchset introduces alternative names for network interfaces.
Main goal is to:
1) Overcome the IFNAMSIZ limitation (altname limitation is 128 bytes)
2) Allow to have multiple names at the same time (multiple udev patterns)
3) Allow to use alternative names as handle for commands

The patchset introduces two new commands to add/delete list of properties.
Currently only alternative names are implemented but the ifrastructure
could be easily extended later on. This is very similar to the list of vlan
and tunnels being added/deleted to/from bridge ports.

See following examples.

$ ip link
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
2: dummy0: <BROADCAST,NOARP> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
    link/ether ae:67:a9:67:46:86 brd ff:ff:ff:ff:ff:ff

-> Add alternative names for dummy0:

$ ip link prop add dummy0 altname someothername
$ ip link prop add dummy0 altname someotherveryveryveryverylongname
$ ip link
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
2: dummy0: <BROADCAST,NOARP> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
    link/ether ae:67:a9:67:46:86 brd ff:ff:ff:ff:ff:ff
    altname someothername
    altname someotherveryveryveryverylongname
$ ip link show someotherveryveryveryverylongname
2: dummy0: <BROADCAST,NOARP> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
    link/ether ae:67:a9:67:46:86 brd ff:ff:ff:ff:ff:ff
    altname someothername
    altname someotherveryveryveryverylongname

-> Add bridge brx, add it's alternative name and use alternative names to
   do enslavement.

$ ip link add name brx type bridge
$ ip link prop add brx altname mypersonalsuperspecialbridge
$ ip link set someotherveryveryveryverylongname master mypersonalsuperspecialbridge
$ ip link
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
2: dummy0: <BROADCAST,NOARP> mtu 1500 qdisc noop master brx state DOWN mode DEFAULT group default qlen 1000
    link/ether ae:67:a9:67:46:86 brd ff:ff:ff:ff:ff:ff
    altname someothername
    altname someotherveryveryveryverylongname
3: brx: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
    link/ether ae:67:a9:67:46:86 brd ff:ff:ff:ff:ff:ff
    altname mypersonalsuperspecialbridge

-> Add ipv4 address to the bridge using alternative name:

$ ip addr add 192.168.0.1/24 dev mypersonalsuperspecialbridge
$ ip addr show mypersonalsuperspecialbridge
3: brx: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN group default qlen 1000
    link/ether ae:67:a9:67:46:86 brd ff:ff:ff:ff:ff:ff
    altname mypersonalsuperspecialbridge
    inet 192.168.0.1/24 scope global brx
       valid_lft forever preferred_lft forever

-> Delete one of dummy0 alternative names:

$ ip link prop del dummy0 altname someotherveryveryveryverylongname
$ ip link
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
2: dummy0: <BROADCAST,NOARP> mtu 1500 qdisc noop master brx state DOWN mode DEFAULT group default qlen 1000
    link/ether ae:67:a9:67:46:86 brd ff:ff:ff:ff:ff:ff
    altname someothername
3: brx: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
    link/ether ae:67:a9:67:46:86 brd ff:ff:ff:ff:ff:ff
    altname mypersonalsuperspecialbridge

-> Add multiple alternative names at once

$ ip link prop add dummy0 altname a altname b altname c altname d
$ ip link
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
2: dummy0: <BROADCAST,NOARP> mtu 1500 qdisc noop master brx state DOWN mode DEFAULT group default qlen 1000
    link/ether ae:67:a9:67:46:86 brd ff:ff:ff:ff:ff:ff
    altname someothername
    altname a
    altname b
    altname c
    altname d
3: brx: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
    link/ether ae:67:a9:67:46:86 brd ff:ff:ff:ff:ff:ff
    altname mypersonalsuperspecialbridge
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Extend the basic rtnetlink commands to use alternative interface names
as a handle instead of ifindex and ifname.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Introduce helper function rtnl_get_dev() that gets net_device structure
instance pointer according to passed ifname or ifname attribute.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

__rtnl_newlink() code flow is a bit different around tb[IFLA_IFNAME]
processing comparing to the other places. Change that to be unified with
the rest.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Extend exiting getlink info message with list of properties. Now the
only ones are alternative names.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add two commands to add and delete list of link properties. Implement
the first property type along - alternative ifnames.
Each net device can have multiple alternative names.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Introduce name_node structure to hold name of device and put it into
hashlist instead of putting there struct net_device directly. Add a
necessary infrastructure to manipulate the hashlist. This prepares
the code to use the same hashlist for alternative names introduced
later in this set.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Name hashlist is going to be used for more than just dev->name, so use
rather index hashlist for iteration over net_device instances.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp_twsk_unique() has a hard coded assumption about ipv4 loopback
being 127/8

Lets instead use the standard ipv4_is_loopback() method,
in a new ipv6_addr_v4mapped_loopback() helper.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Function netif_schedule_queue() has a hardcoded comparison between queue
state and any xoff flag. This comparison does the same thing as method
netif_xmit_stopped(). In terms of code clarity, it is better. See other
methods like: generic_xdp_tx() and dev_direct_xmit().
Signed-off-by: Julio Faracco <jcfaracco@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The same for-loop check for the LINK_LIST_READY bit of an OOB_CTRL
register is used in several places. Factor these out into a single
function to reduce the lines of code.

Change-Id: I20e8f327045a72acc0a83e2d145ae2993ab62915
Signed-off-by: Prashant Malani <pmalani@chromium.org>
Reviewed-by: Grant Grundler <grundler@chromium.org>
Acked-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Pull networking fixes from David Miller:

 1) Sanity check URB networking device parameters to avoid divide by
    zero, from Oliver Neukum.

 2) Disable global multicast filter in NCSI, otherwise LLDP and IPV6
    don't work properly. Longer term this needs a better fix tho. From
    Vijay Khemka.

 3) Small fixes to selftests (use ping when ping6 is not present, etc.)
    from David Ahern.

 4) Bring back rt_uses_gateway member of struct rtable, it's semantics
    were not well understood and trying to remove it broke things. From
    David Ahern.

 5) Move usbnet snaity checking, ignore endpoints with invalid
    wMaxPacketSize. From Bjørn Mork.

 6) Missing Kconfig deps for sja1105 driver, from Mao Wenan.

 7) Various small fixes to the mlx5 DR steering code, from Alaa Hleihel,
    Alex Vesker, and Yevgeny Kliteynik

 8) Missing CAP_NET_RAW checks in various places, from Ori Nimron.

 9) Fix crash when removing sch_cbs entry while offloading is enabled,
    from Vinicius Costa Gomes.

10) Signedness bug fixes, generally in looking at the result given by
    of_get_phy_mode() and friends. From Dan Crapenter.

11) Disable preemption around BPF_PROG_RUN() calls, from Eric Dumazet.

12) Don't create VRF ipv6 rules if ipv6 is disabled, from David Ahern.

13) Fix quantization code in tcp_bbr, from Kevin Yang.

* git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (127 commits)
  net: tap: clean up an indentation issue
  nfp: abm: fix memory leak in nfp_abm_u32_knode_replace
  tcp: better handle TCP_USER_TIMEOUT in SYN_SENT state
  sk_buff: drop all skb extensions on free and skb scrubbing
  tcp_bbr: fix quantization code to not raise cwnd if not probing bandwidth
  mlxsw: spectrum_flower: Fail in case user specifies multiple mirror actions
  Documentation: Clarify trap's description
  mlxsw: spectrum: Clear VLAN filters during port initialization
  net: ena: clean up indentation issue
  NFC: st95hf: clean up indentation issue
  net: phy: micrel: add Asym Pause workaround for KSZ9021
  net: socionext: ave: Avoid using netdev_err() before calling register_netdev()
  ptp: correctly disable flags on old ioctls
  lib: dimlib: fix help text typos
  net: dsa: microchip: Always set regmap stride to 1
  nfp: flower: fix memory leak in nfp_flower_spawn_vnic_reprs
  nfp: flower: prevent memory leak in nfp_flower_spawn_phy_reprs
  net/sched: Set default of CONFIG_NET_TC_SKB_EXT to N
  vrf: Do not attempt to create IPv6 mcast rule if IPv6 is disabled
  net: sched: sch_sfb: don't call qdisc_put() while holding tree lock
  ...

Merge hugepage allocation updates from David Rientjes:
 "We (mostly Linus, Andrea, and myself) have been discussing offlist how
  to implement a sane default allocation strategy for hugepages on NUMA
  platforms.

  With these reverts in place, the page allocator will happily allocate
  a remote hugepage immediately rather than try to make a local hugepage
  available. This incurs a substantial performance degradation when
  memory compaction would have otherwise made a local hugepage
  available.

  This series reverts those reverts and attempts to propose a more sane
  default allocation strategy specifically for hugepages. Andrea
  acknowledges this is likely to fix the swap storms that he originally
  reported that resulted in the patches that removed __GFP_THISNODE from
  hugepage allocations.

  The immediate goal is to return 5.3 to the behavior the kernel has
  implemented over the past several years so that remote hugepages are
  not immediately allocated when local hugepages could have been made
  available because the increased access latency is untenable.

  The next goal is to introduce a sane default allocation strategy for
  hugepages allocations in general regardless of the configuration of
  the system so that we prevent thrashing of local memory when
  compaction is unlikely to succeed and can prefer remote hugepages over
  remote native pages when the local node is low on memory."

Note on timing: this reverts the hugepage VM behavior changes that got
introduced fairly late in the 5.3 cycle, and that fixed a huge
performance regression for certain loads that had been around since
4.18.

Andrea had this note:

 "The regression of 4.18 was that it was taking hours to start a VM
  where 3.10 was only taking a few seconds, I reported all the details
  on lkml when it was finally tracked down in August 2018.

     https://lore.kernel.org/linux-mm/20180820032640.9896-2-aarcange@redhat.com/

  __GFP_THISNODE in MADV_HUGEPAGE made the above enterprise vfio
  workload degrade like in the "current upstream" above. And it still
  would have been that bad as above until 5.3-rc5"

where the bad behavior ends up happening as you fill up a local node,
and without that change, you'd get into the nasty swap storm behavior
due to compaction working overtime to make room for more memory on the
nodes.

As a result 5.3 got the two performance fix reverts in rc5.

However, David Rientjes then noted that those performance fixes in turn
regressed performance for other loads - although not quite to the same
degree.  He suggested reverting the reverts and instead replacing them
with two small changes to how hugepage allocations are done (patch
descriptions rephrased by me):

 - "avoid expensive reclaim when compaction may not succeed": just admit
   that the allocation failed when you're trying to allocate a huge-page
   and compaction wasn't successful.

 - "allow hugepage fallback to remote nodes when madvised": when that
   node-local huge-page allocation failed, retry without forcing the
   local node.

but by then I judged it too late to replace the fixes for a 5.3 release.
So 5.3 was released with behavior that harked back to the pre-4.18 logic.

But now we're in the merge window for 5.4, and we can see if this
alternate model fixes not just the horrendous swap storm behavior, but
also restores the performance regression that the late reverts caused.

Fingers crossed.

* emailed patches from David Rientjes <rientjes@google.com>:
  mm, page_alloc: allow hugepage fallback to remote nodes when madvised
  mm, page_alloc: avoid expensive reclaim when compaction may not succeed
  Revert "Revert "Revert "mm, thp: consolidate THP gfp handling into alloc_hugepage_direct_gfpmask""
  Revert "Revert "mm, thp: restore node-local hugepage allocations""

For systems configured to always try hard to allocate transparent
hugepages (thp defrag setting of "always") or for memory that has been
explicitly madvised to MADV_HUGEPAGE, it is often better to fallback to
remote memory to allocate the hugepage if the local allocation fails
first.

The point is to allow the initial call to __alloc_pages_node() to attempt
to defragment local memory to make a hugepage available, if possible,
rather than immediately fallback to remote memory.  Local hugepages will
always have a better access latency than remote (huge)pages, so an attempt
to make a hugepage available locally is always preferred.

If memory compaction cannot be successful locally, however, it is likely
better to fallback to remote memory.  This could take on two forms: either
allow immediate fallback to remote memory or do per-zone watermark checks.
It would be possible to fallback only when per-zone watermarks fail for
order-0 memory, since that would require local reclaim for all subsequent
faults so remote huge allocation is likely better than thrashing the local
zone for large workloads.

In this case, it is assumed that because the system is configured to try
hard to allocate hugepages or the vma is advised to explicitly want to try
hard for hugepages that remote allocation is better when local allocation
and memory compaction have both failed.
Signed-off-by: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Stefan Priebe - Profihost AG <s.priebe@profihost.ag>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Memory compaction has a couple significant drawbacks as the allocation
order increases, specifically:

 - isolate_freepages() is responsible for finding free pages to use as
   migration targets and is implemented as a linear scan of memory
   starting at the end of a zone,

 - failing order-0 watermark checks in memory compaction does not account
   for how far below the watermarks the zone actually is: to enable
   migration, there must be *some* free memory available.  Per the above,
   watermarks are not always suffficient if isolate_freepages() cannot
   find the free memory but it could require hundreds of MBs of reclaim to
   even reach this threshold (read: potentially very expensive reclaim with
   no indication compaction can be successful), and

 - if compaction at this order has failed recently so that it does not even
   run as a result of deferred compaction, looping through reclaim can often
   be pointless.

For hugepage allocations, these are quite substantial drawbacks because
these are very high order allocations (order-9 on x86) and falling back to
doing reclaim can potentially be *very* expensive without any indication
that compaction would even be successful.

Reclaim itself is unlikely to free entire pageblocks and certainly no
reliance should be put on it to do so in isolation (recall lumpy reclaim).
This means we should avoid reclaim and simply fail hugepage allocation if
compaction is deferred.

It is also not helpful to thrash a zone by doing excessive reclaim if
compaction may not be able to access that memory.  If order-0 watermarks
fail and the allocation order is sufficiently large, it is likely better
to fail the allocation rather than thrashing the zone.
Signed-off-by: David Rientjes <rientjes@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Stefan Priebe - Profihost AG <s.priebe@profihost.ag>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>