net: dl2k: Remove unused macro DRV_NAME

There is no caller in tree any more.

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: wan: slic_ds26522: Remove unused macro DRV_NAME

There is no caller in tree any more.

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: Remove unused macro TIPC_NACK_INTV

There is no caller in tree any more.

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: Remove unused macro TIPC_FWD_MSG

There is no caller in tree any more.

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mptcp: Remove unused macro MPTCP_SAME_STATE

There is no caller in tree any more.

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: clean up codestyle

This is a pure codestyle cleanup patch. No functional change intended.

Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Use helper macro IP_MAX_MTU in __ip_append_data()

What 0xFFFF means here is actually the max mtu of a ip packet. Use help
macro IP_MAX_MTU here.

Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: ethernet: ti: am65-cpts: fix i2083 genf (and estf) Reconfiguration Issue

The new bit TX_GENF_CLR_EN has been added in AM65x SR2.0 to fix i2083
errata, which can be just set unconditionally for all SoCs.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'sfc-clean-up-some-W-1-build-warnings'

Edward Cree says:

====================
sfc: clean up some W=1 build warnings

A collection of minor fixes to issues flagged up by W=1.
After this series, the only remaining warnings in the sfc driver are
 some 'member missing in kerneldoc' warnings from ptp.c.
Tested by building on x86_64 and running 'ethtool -p' on an EF10 NIC;
 there was no error, but I couldn't observe the actual LED as I'm
 working remotely.

[ Incidentally, ethtool_phys_id()'s behaviour on an error return
  looks strange — if I'm reading it right, it will break out of the
  inner loop but not the outer one, and eventually return the rc
  from the last run of the inner loop.  Is this intended? ]
====================

Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

sfc: return errors from efx_mcdi_set_id_led, and de-indirect

W=1 warnings indicated that 'rc' was unused in efx_mcdi_set_id_led();
 change the function to return int instead of void and plumb the rc
 through the caller efx_ethtool_phys_id().
Since (post-Falcon) all sfc NICs use MCDI for this, there's no point in
 indirecting through a nic_type method, so remove that and just call
 efx_mcdi_set_id_led() directly.

Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

sfc: fix kernel-doc on struct efx_loopback_state

Missing 'struct' keyword caused "cannot understand function prototype"
 warnings.

Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

sfc: fix unused-but-set-variable warning in efx_farch_filter_remove_safe

Thanks to some past refactor, 'spec' is not actually used in this
 function; the code using it moved to the callee efx_farch_filter_remove.
Remove the variable to fix a W=1 warning.

Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

sfc: fix W=1 warnings in efx_farch_handle_rx_not_ok

Some of these RX-event flags aren't used at all, so remove them.
Others are used only #ifdef DEBUG to log a message; suppress the
 unused-var warnings #ifndef DEBUG with a void cast.

Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'Add-ip6_fragment-in-ipv6_stub'

wenxu says:

====================
Add ip6_fragment in ipv6_stub

Add ip6_fragment in ipv6_stub and use it in openvswitch
This version add default function eafnosupport_ipv6_fragment
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

openvswitch: using ip6_fragment in ipv6_stub

Using ipv6_stub->ipv6_fragment to avoid the netfilter dependency

Signed-off-by: wenxu <wenxu@ucloud.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv6: add ipv6_fragment hook in ipv6_stub

Add ipv6_fragment to ipv6_stub to avoid calling netfilter when
access ip6_fragment.

Signed-off-by: wenxu <wenxu@ucloud.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'gtp-minor-enhancements'

Nicolas Dichtel says:

====================
gtp: minor enhancements

The first patch removes a useless rcu lock and the second relax alloc
constraints when a PDP context is added.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

gtp: relax alloc constraint when adding a pdp

When a PDP context is added, the rtnl lock is held, thus no need to force
a GFP_ATOMIC.

Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

gtp: remove useless rcu_read_lock()

The rtnl lock is taken just the line above, no need to take the rcu also.

Fixes: 1788b8569f5d ("gtp: fix use-after-free in gtp_encap_destroy()")
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: phylink: avoid oops during initialisation

If we intend to use PCS operations, mac_pcs_get_state() will not be
implemented, so will be NULL. If we also intend to register the PCS
operations in mac_prepare() or mac_config(), then this leads to an
attempt to call NULL function pointer during phylink_start(). Avoid
this, but we must report the link is down.

Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'hinic-add-debugfs-support'

Luo bin says:

====================
hinic: add debugfs support

add debugfs node for querying sq/rq info and function table
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

hinic: add support to query function table

add debugfs node for querying function table, for example:
cat /sys/kernel/debug/hinic/0000:15:00.0/func_table/valid

Signed-off-by: Luo bin <luobin9@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

hinic: add support to query rq info

add debugfs node for querying rq info, for example:
cat /sys/kernel/debug/hinic/0000:15:00.0/RQs/0x0/rq_hw_pi

Signed-off-by: Luo bin <luobin9@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

hinic: add support to query sq info

add debugfs node for querying sq info, for example:
cat /sys/kernel/debug/hinic/0000:15:00.0/SQs/0x0/sq_pi

Signed-off-by: Luo bin <luobin9@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

netlink: policy: correct validation type check

In the policy export for binary attributes I erroneously used
a != NLA_VALIDATE_NONE comparison instead of checking for the
two possible values, which meant that if a validation function
pointer ended up aliasing the min/max as negatives, we'd hit
a warning in nla_get_range_unsigned().

Fix this to correctly check for only the two types that should
be handled here, i.e. range with or without warn-too-long.

Reported-by: syzbot+353df1490da781637624@syzkaller.appspotmail.com
Fixes: 8aa26c575fb3 ("netlink: make NLA_BINARY validation more flexible")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

netlabel: remove unused param from audit_log_format()

Commit d3b990b7f327 ("netlabel: fix problems with mapping removal")
added a check to return an error if ret_val != 0, before ret_val is
later used in a log message. Now it will unconditionally print "...
res=1". So just drop the check.

Addresses-Coverity: ("Dead code")
Fixes: d3b990b7f327 ("netlabel: fix problems with mapping removal")
Signed-off-by: Alex Dewar <alex.dewar90@gmail.com>
Acked-by: Paul Moore <paul@paul-moore.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'ionic-memory-usage-rework'

Shannon Nelson says:

====================
ionic memory usage rework

Previous review comments have suggested [1],[2] that this driver
needs to rework how queue resources are managed and reconfigured
so that we don't do a full driver reset and to better handle
potential allocation failures.  This patchset is intended to
address those comments.

The first few patches clean some general issues and
simplify some of the memory structures.  The last 4 patches
specifically address queue parameter changes without a full
ionic_stop()/ionic_open().

[1] https://lore.kernel.org/netdev/20200706103305.182bd727@kicinski-fedora-pc1c0hjn.dhcp.thefacebook.com/
[2] https://lore.kernel.org/netdev/20200724.194417.2151242753657227232.davem@davemloft.net/

v3: use PTR_ALIGN without typecast
    fix up Neel's attribution

v2: use PTR_ALIGN
    recovery if netif_set_real_num_tx/rx_queues fails
    less racy queue bring up after reconfig
    common-ize the reconfig queue stop and start
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

ionic: pull reset_queues into tx_timeout handler

Convert tx_timeout handler to not do the full reset.  As this was
the last user of ionic_reset_queues(), we can drop it.

Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>

ionic: change queue count with no reset

Add to our new ionic_reconfigure_queues() to also be able to change
the number of queues in use, and to change the queue interrupt layout
between split and combined.

Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>

ionic: change the descriptor ring length without full reset

The original way of changing ring length was to completely
tear down the lif's queue structure and then rebuild it, while
running the risk of allocations that might fail in the middle
and leave us with a broken driver.

Instead, we can set up all the new queue and descriptor
allocations first, then swap them out and delete the old
allocations.  If the new allocations fail, we report the error,
stay with the old setup and continue running.  This gives us
a safer path, and a smaller window of time where we're not
processing traffic.

Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>

ionic: change mtu without full queue rebuild

We really don't need to tear down and rebuild the whole queue structure
when changing the MTU; we can simply stop the queues, clean and refill,
then restart the queues.

Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>

ionic: use index not pointer for queue tracking

Use index counters rather than pointers for tracking head
and tail in the queues to save a little memory and to perhaps
slightly faster queue processing.

Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>

ionic: reduce contiguous memory allocation requirement

Split out the queue descriptor blocks into separate dma
allocations to make for smaller blocks.

Co-developed-by: Neel Patel <neel@pensando.io>
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>

ionic: clean up unnecessary non-static functions

ionic_open() and ionic_stop() are not referenced outside of their
defining file, so make them static.

Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>

ionic: rework and simplify handling of the queue stats block

Use a block of stats structs attached to the lif instead of
little ones attached to each qcq.  This simplifies our memory
management and gets rid of a lot of unnecessary indirection.

Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>

ionic: remove lif list concept

As we aren't yet supporting multiple lifs, we can remove
complexity by removing the list concept and related code,
to be re-engineered later when actually needed.

Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>

ionic: use kcalloc for new arrays

Use kcalloc for allocating arrays of structures.

Following along after
commit e71642009cbdA ("ionic_lif: Use devm_kcalloc() in ionic_qcq_alloc()")
there are a couple more array allocations that can be converted
to using devm_kcalloc().

Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>

ionic: fix up a couple of debug strings

Fix the queue name displayed.

Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>

ionic: set MTU floor at ETH_MIN_MTU

The NIC might tell us its minimum MTU, but let's be sure not
to use something smaller than ETH_MIN_MTU.

Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'Enable-Fiber-on-DP83822-PHY'

Dan Murphy says:

====================
Enable Fiber on DP83822 PHY

The DP83822 Ethernet PHY has the ability to connect via a Fiber port.  The
derivative PHYs DP83825 and DP83826 do not have this ability. In fiber mode
the DP83822 disables auto negotiation and has a fixed 100Mbps speed with
support for full or half duplex modes.

A devicetree binding was added to set the signal polarity for the fiber
connection.  This property is only applicable if the FX_EN strap is set in
hardware other wise the signal loss detection is disabled on the PHY.

If the FX_EN is not strapped the device can be configured to run in fiber mode
via the device tree. All be it the PHY will not perform signal loss detection.

v2 review from a long time ago can be found here - https://lore.kernel.org/patchwork/patch/1270958/
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

net: phy: DP83822: Add ability to advertise Fiber connection

The DP83822 can be configured to use a Fiber connection.  The strap
register is read to determine if the device has been configured to use
a fiber connection.  With the fiber connection the PHY can be configured
to detect whether the fiber connection is active by either a high signal
or a low signal.

Fiber mode is only applicable to the DP83822 so rework the PHY match
table so that non-fiber PHYs can still use the same driver but not call
or use any of the fiber features.

Signed-off-by: Dan Murphy <dmurphy@ti.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

dt-bindings: net: dp83822: Add TI dp83822 phy

Add a dt binding for the TI dp83822 ethernet phy device.

Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Dan Murphy <dmurphy@ti.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: add option to not create fall-back tunnels in root-ns as well

The sysctl that was added  earlier by commit 79134e6ce2c ("net: do
not create fallback tunnels for non-default namespaces") to create
fall-back only in root-ns. This patch enhances that behavior to provide
option not to create fallback tunnels in root-ns as well. Since modules
that create fallback tunnels could be built-in and setting the sysctl
value after booting is pointless, so added a kernel cmdline options to
change this default. The default setting is preseved for backward
compatibility. The kernel command line option of fb_tunnels=initns will
set the sysctl value to 1 and will create fallback tunnels only in initns
while kernel cmdline fb_tunnels=none will set the sysctl value to 2 and
fallback tunnels are skipped in every netns.

Signed-off-by: Mahesh Bandewar <maheshb@google.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Maciej Zenczykowski <maze@google.com>
Cc: Jian Yang <jianyang@google.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'Add-phylib-support-to-smsc95xx'

Andre Edich says:

====================
Add phylib support to smsc95xx

To allow to probe external PHY drivers, this patch series adds use of
phylib to the smsc95xx driver.

Changes in v5:
- Removed all phy_read calls from the smsc95xx driver.

Changes in v4:
- Removed useless inline type qualifier.

Changes in v3:
- Moved all MDI-X functionality to the corresponding phy driver;
- Removed field internal_phy from a struct smsc95xx_priv;
- Initialized field is_internal of a struct phy_device;
- Kconfig: Added selection of PHYLIB and SMSC_PHY for USB_NET_SMSC95XX.

Changes in v2:
- Moved 'net' patches from here to the separate patch series;
- Removed redundant call of the phy_start_aneg after phy_start;
- Removed netif_dbg tracing "speed, duplex, lcladv, and rmtadv";
- mdiobus: added dependency from the usbnet device;
- Moved making of the MII address from 'phy_id' and 'idx' into the
  function mii_address;
- Moved direct MDIO accesses under condition 'if (pdata->internal_phy)',
  as they only need for the internal PHY;
- To be sure, that this set of patches is git-bisectable, tested each
  sub-set of patches to be functional for both, internal and external
  PHYs, including suspend/resume test for the 'devices'
  (5.7.8-1-ARCH, Raspberry Pi 3 Model B).
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

smsc95xx: add phylib support

Generally, each PHY has their own configuration and it can be done
through an external PHY driver.  The smsc95xx driver uses only the
hard-coded internal PHY configuration.

This patch adds phylib support to probe external PHY drivers for
configuring external PHYs.

The MDI-X configuration for the internal PHYs moves from
drivers/net/usb/smsc95xx.c to drivers/net/phy/smsc.c.

Signed-off-by: Andre Edich <andre.edich@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

smsc95xx: use usbnet->driver_priv

Using `void *driver_priv` instead of `unsigned long data[]` is more
straightforward way to recover the `struct smsc95xx_priv *` from the
`struct net_device *`.

Signed-off-by: Andre Edich <andre.edich@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

smsc95xx: remove redundant function arguments

This patch removes arguments netdev and phy_id from the functions
smsc95xx_mdio_read_nopm and smsc95xx_mdio_write_nopm.  Both removed
arguments are recovered from a new argument `struct usbnet *dev`.

Signed-off-by: Andre Edich <andre.edich@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge tag 'mac80211-next-for-davem-2020-08-28' of git://git./linux/kernel/git/jberg/mac80211-next

Johannes Berg says:

====================
This time we have:
 * some code to support SAE (WPA3) offload in AP mode
 * many documentation (wording) fixes/updates
 * netlink policy updates, including the use of NLA_RANGE
   with binary attributes
 * regulatory improvements for adjacent frequency bands
 * and a few other small additions/refactorings/cleanups
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

gtp: add notification mechanism

Like all other network functions, let's notify gtp context on creation and
deletion.

Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Tested-by: Gabriel Ganne <gabriel.ganne@6wind.com>
Acked-by: Harald Welte <laforge@gnumonks.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 's390-qeth-next'

Julian Wiedmann says:

====================
s390/qeth: updates 2020-08-27

please apply the following patch series for qeth to netdev's net-next tree.

Patch 8 makes some improvements to how we handle HW address events,
avoiding some uncertainty around processing stale events after we
switched off the feature.
Except for that it's all straight-forward cleanups.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

s390/qeth: strictly order bridge address events

The current code for bridge address events has two shortcomings in its
control sequence:

1. after disabling address events via PNSO, we don't flush the remaining
   events from the event_wq. So if the feature is re-enabled fast
   enough, stale events could leak over.
2. PNSO and the events' arrival via the READ ccw device are unordered.
   So even if we flushed the workqueue, it's difficult to say whether
   the READ device might produce more events onto the workqueue
   afterwards.

Fix this by
1. explicitly fencing off the events when we no longer care, in the
   READ device's event handler. This ensures that once we flush the
   workqueue, it doesn't get additional address events.
2. Flush the workqueue after disabling the events & fencing them off.
   As the code that triggers the flush will typically hold the sbp_lock,
   we need to rework the worker code to avoid a deadlock here in case
   of a 'notifications-stopped' event. In case of lock contention,
   requeue such an event with a delay. We'll eventually aquire the lock,
   or spot that the feature has been disabled and the event can thus be
   discarded.

This leaves the theoretical race that a stale event could arrive
_after_ we re-enabled ourselves to receive events again. Such an event
would be impossible to distinguish from a 'good' event, nothing we can
do about it.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Reviewed-by: Alexandra Winter <wintera@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

s390/qeth: unify structs for bridge port state

The data returned from IPA_SBP_QUERY_BRIDGE_PORTS and
IPA_SBP_BRIDGE_PORT_STATE_CHANGE has the same format. Use a single
struct definition for it.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Reviewed-by: Alexandra Winter <wintera@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

s390/qeth: copy less data from bridge state events

Current code copies _all_ entries from the event into a worker, when we
later only need specific data from the first entry.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Reviewed-by: Alexandra Winter <wintera@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

s390/qeth: don't let HW override the configured port role

The only time that our Bridgeport role should change is when we change
the configuration ourselves. In which case we also adjust our internal
state tracking, no need to do it again when we receive the corresponding
event.

Removing the locked section helps a subsequent patch that needs to flush
the workqueue while under sbp_lock.

It would be nice to raise a warning here in case HW does weird things
after all, but this could end up generating false-positives when we
change the configuration ourselves.

Suggested-by: Alexandra Winter <wintera@linux.ibm.com>
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Reviewed-by: Alexandra Winter <wintera@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

s390/qeth: don't disable address events during initialization

A newly initialized device is disabled for address events, there's no
need to explicitly disable them.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Reviewed-by: Alexandra Winter <wintera@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

s390/qeth: make queue lock a proper spinlock

queue->state is a ternary spinlock in disguise, used by
OSA's TX completion path to lock the Output Queue and flush any pending
packets on it to the device. If the Queue is already locked by our TX
code, setting the lock word to QETH_OUT_Q_LOCKED_FLUSH lets the TX
completion code move on - the TX path will later take care of things
when it unlocks the Queue.

This sort of DIY locking is a non-starter of course, just let the
TX completion path block on the spinlock when necessary. If that ends up
causing additional latency due to lock contention, then converting
the OSA path to use xmit_more is the right way to go forward.

Also slightly expand the locked section and capture all of
qeth_do_send_packet(), so that the update for the 'bufs_pack' statistics
is done race-free.

While reworking the TX completion path's code, remove a barrier() that
doesn't make any sense.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

s390/qeth: use to_delayed_work()

Avoid poking around in the delayed_work struct's internals.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

s390/qeth: clean up qeth_l3_send_setdelmc()'s declaration

Clarify that the 'ipacmd' parameter is an enum, and thus compatible to
what qeth_ipa_alloc_cmd() expects as input.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: fix use-after-free in tipc_bcast_get_mode

Syzbot has reported those issues as:

==================================================================
BUG: KASAN: use-after-free in tipc_bcast_get_mode+0x3ab/0x400 net/tipc/bcast.c:759
Read of size 1 at addr ffff88805e6b3571 by task kworker/0:6/3850

CPU: 0 PID: 3850 Comm: kworker/0:6 Not tainted 5.8.0-rc7-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Workqueue: events tipc_net_finalize_work

Thread 1's call trace:
[...]
  kfree+0x103/0x2c0 mm/slab.c:3757 <- bcbase releasing
  tipc_bcast_stop+0x1b0/0x2f0 net/tipc/bcast.c:721
  tipc_exit_net+0x24/0x270 net/tipc/core.c:112
[...]

Thread 2's call trace:
[...]
  tipc_bcast_get_mode+0x3ab/0x400 net/tipc/bcast.c:759 <- bcbase
has already been freed by Thread 1

  tipc_node_broadcast+0x9e/0xcc0 net/tipc/node.c:1744
  tipc_nametbl_publish+0x60b/0x970 net/tipc/name_table.c:752
  tipc_net_finalize net/tipc/net.c:141 [inline]
  tipc_net_finalize+0x1fa/0x310 net/tipc/net.c:131
  tipc_net_finalize_work+0x55/0x80 net/tipc/net.c:150
[...]

==================================================================
BUG: KASAN: use-after-free in tipc_named_reinit+0xef/0x290 net/tipc/name_distr.c:344
Read of size 8 at addr ffff888052ab2000 by task kworker/0:13/30628
CPU: 0 PID: 30628 Comm: kworker/0:13 Not tainted 5.8.0-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Workqueue: events tipc_net_finalize_work
Call Trace:
 __dump_stack lib/dump_stack.c:77 [inline]
 dump_stack+0x1f0/0x31e lib/dump_stack.c:118
 print_address_description+0x66/0x5a0 mm/kasan/report.c:383
 __kasan_report mm/kasan/report.c:513 [inline]
 kasan_report+0x132/0x1d0 mm/kasan/report.c:530
 tipc_named_reinit+0xef/0x290 net/tipc/name_distr.c:344
 tipc_net_finalize+0x85/0xe0 net/tipc/net.c:138
 tipc_net_finalize_work+0x50/0x70 net/tipc/net.c:150
 process_one_work+0x789/0xfc0 kernel/workqueue.c:2269
 worker_thread+0xaa4/0x1460 kernel/workqueue.c:2415
 kthread+0x37e/0x3a0 drivers/block/aoe/aoecmd.c:1234
 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:293
[...]
Freed by task 14058:
 save_stack mm/kasan/common.c:48 [inline]
 set_track mm/kasan/common.c:56 [inline]
 kasan_set_free_info mm/kasan/common.c:316 [inline]
 __kasan_slab_free+0x114/0x170 mm/kasan/common.c:455
 __cache_free mm/slab.c:3426 [inline]
 kfree+0x10a/0x220 mm/slab.c:3757
 tipc_exit_net+0x29/0x50 net/tipc/core.c:113
 ops_exit_list net/core/net_namespace.c:186 [inline]
 cleanup_net+0x708/0xba0 net/core/net_namespace.c:603
 process_one_work+0x789/0xfc0 kernel/workqueue.c:2269
 worker_thread+0xaa4/0x1460 kernel/workqueue.c:2415
 kthread+0x37e/0x3a0 drivers/block/aoe/aoecmd.c:1234
 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:293

Fix it by calling flush_scheduled_work() to make sure the
tipc_net_finalize_work() stopped before releasing bcbase object.

Reported-by: syzbot+6ea1f7a8df64596ef4d7@syzkaller.appspotmail.com
Reported-by: syzbot+e9cc557752ab126c1b99@syzkaller.appspotmail.com
Acked-by: Jon Maloy <jmaloy@redhat.com>
Signed-off-by: Hoang Huu Le <hoang.h.le@dektech.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'Move-MDIO-drivers-into-their-own-directory'

Andrew Lunn says:

====================
Move MDIO drivers into their own directory

The phy subdirectory is getting cluttered. It has both PHY drivers and
MDIO drivers, plus a stray switch driver. Soon more PCS drivers are
likely to appear.

Move MDIO and PCS drivers into new directories. This requires fixing
up the xgene driver which uses a relative include path.

v2:
Move the subdirs to drivers/net, rather than drivers/net/phy.

v3:
Add subdirectories under include/linux for mdio and pcs

v4:
there->their
include path fix
No new kconfig prompts
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

net: phy: Sort Kconfig and Makefile

Sort the Kconfig based on the text shown in make menuconfig and sort
the Makefile by CONFIG symbol.

Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: mdio: Move MDIO drivers into a new subdirectory

Move all the MDIO drivers and multiplexers into drivers/net/mdio.  The
mdio core is however left in the phy directory, due to mutual
dependencies between the MDIO core and the PHY core.

Take this opportunity to sort the Kconfig based on the menuconfig
strings, and move the multiplexers to the end with a separating
comment.

v2:
Fix typo in commit message

Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: xgene: Move shared header file into include/linux

This header file is currently included into the ethernet driver via a
relative path into the PHY subsystem. This is bad practice, and causes
issues for the upcoming move of the MDIO driver. Move the header file
into include/linux to clean this up.

v2:
Move header to include/linux/mdio

Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/phy/mdio-i2c: Move header file to include/linux/mdio

In preparation for moving all MDIO drivers into drivers/net/mdio, move
the mdio-i2c header file into include/linux/mdio so it can be used by
both the MDIO driver and the SFP code which instantiates I2C MDIO
busses.

v2:
Add include/linux/mdio

Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: pcs: Move XPCS into new PCS subdirectory

Create drivers/net/pcs and move the Synopsys DesignWare XPCS into the
new directory. Move the header file into a subdirectory
include/linux/pcs

Start a naming convention of all PCS files use the prefix pcs-, and
rename the XPCS files to fit.

v2:
Add include/linux/pcs

v4:
Fix include path in stmmac.
Remove PCS_DEVICES to avoid new prompts

Cc: Jose Abreu <Jose.Abreu@synopsys.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

nl80211: support SAE authentication offload in AP mode

Let drivers advertise support for AP-mode SAE authentication offload
with a new NL80211_EXT_FEATURE_SAE_OFFLOAD_AP flag.

Signed-off-by: Chung-Hsien Hsu <stanley.hsu@cypress.com>
Signed-off-by: Chi-Hsien Lin <chi-hsien.lin@cypress.com>
Link: https://lore.kernel.org/r/20200817073316.33402-4-stanley.hsu@cypress.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211: rename csa counters to countdown counters

We want to reuse the functions and structs for other counters such as BSS
color change. Rename them to more generic names.

Signed-off-by: John Crispin <john@phrozen.org>
Link: https://lore.kernel.org/r/20200811080107.3615705-2-john@phrozen.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

nl80211: rename csa counter attributes countdown counters

We want to reuse the attributes for other counters such as BSS color
change. Rename them to more generic names.

Signed-off-by: John Crispin <john@phrozen.org>
Link: https://lore.kernel.org/r/20200811080107.3615705-1-john@phrozen.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

nl80211: add support for setting fixed HE rate/gi/ltf

This patch adds the nl80211 structs, definitions, policies and parsing
code required to pass fixed HE rate, GI and LTF settings.

Signed-off-by: Miles Hu <milehu@codeaurora.org>
Signed-off-by: John Crispin <john@phrozen.org>
Link: https://lore.kernel.org/r/20200804081630.2013619-1-john@phrozen.org
[fix comment]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

nl80211: fix PORT_AUTHORIZED wording to reflect behavior

The CMD_PORT_AUTHORIZED event was described as an event which indicated
a successfully completed 4-way handshake. But the behavior was
not as advertized. The only driver which uses this is brcmfmac, and
this driver only sends the event after a successful 802.1X-FT roam.

This prevents userspace applications from knowing if the 4-way completed
on:

1. Normal 802.1X connects
2. Normal PSK connections
3. FT-PSK roams

wpa_supplicant handles this incorrect behavior by just completing
the connection after association, before the 4-way has completed.
If the 4-way ends up failing it disconnects at that point.

Since this behavior appears to be expected (wpa_s handles it this
way) I have changed the wording in the API description to reflect the
actual behavior.

Signed-off-by: James Prestwood <prestwoj@gmail.com>
Link: https://lore.kernel.org/r/20200413162053.3711-1-prestwoj@gmail.com
[fix spelling of 802.1X throughout ...]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: add helper fn for adjacent rule channels

Some usable channels are located in the union of adjacent
regulatory rules, for example channel 144 in Germany.

Enable them, by also checking if a channel spans two adjacent
regulatory rules/frequency ranges.

All flags involved are disabling things, therefore we can build
the maximum by or-ing them together. Furthermore, take the maximum
of DFS CAC time values and the minimum of allowed power of both
adjacent channels in order to comply with both regulatory rules at
the same time.

Signed-off-by: Markus Theil <markus.theil@tu-ilmenau.de>
Link: https://lore.kernel.org/r/20200803144353.305538-2-markus.theil@tu-ilmenau.de
[remove unrelated comment changes]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: add helper fn for single rule channels

As a preparation to handle adjacent rule channels,
factor out handling channels located in a single
regulatory rule.

Signed-off-by: Markus Theil <markus.theil@tu-ilmenau.de>
Link: https://lore.kernel.org/r/20200803144353.305538-1-markus.theil@tu-ilmenau.de
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

nl80211: use NLA_POLICY_RANGE(NLA_BINARY, ...) for a few attributes

We have a few attributes with minimum and maximum lengths that are
not the same, use the new feature of being able to specify both in
the policy to validate them, removing code and allowing this to be
advertised to userspace in the policy export.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://lore.kernel.org/r/20200819085642.8f12ffa14f33.I9d948d59870e521febcd79bb4a986b1de1dca47b@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

nl80211: clean up code/policy a bit

Use the policy to validate minimum and exact lengths in
some attributes that weren't previously covered in the
right ways, and remove associated validation code.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://lore.kernel.org/r/20200805154714.3ba1e233cfa0.I5dc8109b7ab5c3f4ae925f903a30cc9b35753262@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

net: wireless: Convert to use the preferred fallthrough macro

Convert the uses of fallthrough comments to fallthrough macro.

Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Link: https://lore.kernel.org/r/20200822082323.45495-1-linmiaohe@huawei.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

net: wireless: wext_compat.c: delete duplicated word

Drop the repeated word "be".

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Johannes Berg <johannes.berg@intel.com>
Cc: Kalle Valo <kvalo@codeaurora.org>
Cc: linux-wireless@vger.kernel.org
Link: https://lore.kernel.org/r/20200822231953.465-8-rdunlap@infradead.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

net: wireless: sme.c: delete duplicated word

Drop the repeated word "is".

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Johannes Berg <johannes.berg@intel.com>
Cc: Kalle Valo <kvalo@codeaurora.org>
Cc: linux-wireless@vger.kernel.org
Link: https://lore.kernel.org/r/20200822231953.465-7-rdunlap@infradead.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

net: wireless: scan.c: delete or fix duplicated words

Drop repeated word "stored".
Change "is is" to "it is".

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Johannes Berg <johannes.berg@intel.com>
Cc: Kalle Valo <kvalo@codeaurora.org>
Cc: linux-wireless@vger.kernel.org
Link: https://lore.kernel.org/r/20200822231953.465-6-rdunlap@infradead.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

net: wireless: reg.c: delete duplicated words + fix punctuation

Drop duplicated words "was" and "does".
Fix "let's" apostrophe.

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Johannes Berg <johannes.berg@intel.com>
Cc: Kalle Valo <kvalo@codeaurora.org>
Cc: linux-wireless@vger.kernel.org
Link: https://lore.kernel.org/r/20200822231953.465-5-rdunlap@infradead.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

net: wireless: delete duplicated word + fix grammar

Drop the repeated word "Return" + fix verb.

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Johannes Berg <johannes.berg@intel.com>
Cc: Kalle Valo <kvalo@codeaurora.org>
Cc: linux-wireless@vger.kernel.org
Link: https://lore.kernel.org/r/20200822231953.465-4-rdunlap@infradead.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

net: mac80211: mesh.h: delete duplicated word

Drop the repeated word "address".

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Johannes Berg <johannes.berg@intel.com>
Cc: Kalle Valo <kvalo@codeaurora.org>
Cc: linux-wireless@vger.kernel.org
Link: https://lore.kernel.org/r/20200822231953.465-3-rdunlap@infradead.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

net: mac80211: agg-rx.c: fix duplicated words

Change "If if" to "If it".

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Johannes Berg <johannes.berg@intel.com>
Cc: Kalle Valo <kvalo@codeaurora.org>
Cc: linux-wireless@vger.kernel.org
Link: https://lore.kernel.org/r/20200822231953.465-2-rdunlap@infradead.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

Merge branch 'drivers-net-constify-static-ops-variables'

Rikard Falkeborn says:

====================
drivers/net: constify static ops-variables

This series constifies a number of static ops variables, to allow the
compiler to put them in read-only memory. Compile-tested only.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

net: ath11k: constify ath11k_thermal_ops

The only usage of ath11k_thermal_ops is to pass its address to
thermal_cooling_device_register() which takes a const pointer. Make it
const to allow the compiler to put it in read-only memory.

Signed-off-by: Rikard Falkeborn <rikard.falkeborn@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: phy: mscc: macsec: constify vsc8584_macsec_ops

The only usage of vsc8584_macsec_ops is to assign its address to the
macsec_ops field in the phydev struct, which is a const pointer. Make it
const to allow the compiler to put it in read-only memory.

Signed-off-by: Rikard Falkeborn <rikard.falkeborn@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: phy: at803x: constify static regulator_ops

The only usage of vddio_regulator_ops and vddh_regulator_ops is to
assign their address to the ops field in the regulator_desc struct,
which is a const pointer. Make them const to allow the compiler to
put them in read-only memory.

Signed-off-by: Rikard Falkeborn <rikard.falkeborn@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: renesas: sh_eth: constify bb_ops

The only usage of bb_ops is to assign its address to the ops field in
the mdiobb_ctrl struct, which is a const pointer. Make it const to allow
the compiler to put it in read-only memory.

Signed-off-by: Rikard Falkeborn <rikard.falkeborn@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: ethernet: ravb: constify bb_ops

The only usage of bb_ops is to assign its address to the ops field in
the mdiobb_ctrl struct, which is a const pointer. Make it const to allow
the compiler to put it in read-only memory.

Signed-off-by: Rikard Falkeborn <rikard.falkeborn@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: ethernet: qualcomm: constify qca_serdev_ops

The only usage of qca_serdev_ops is to pass its address to
serdev_device_set_client_ops() which takes a const pointer. Make it
const to allow the compiler to put it in read-only memory.

Signed-off-by: Rikard Falkeborn <rikard.falkeborn@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: ipa: remove duplicate include

Remove linux/notifier.h which is included more than once

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wang Hai <wanghai38@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'refactoring-of-ibmvnic-code'

Lijun Pan says:

====================
refactoring of ibmvnic code

This patch series refactor reset_init and init functions,
and make some other cosmetic changes to make the code
easier to read and debug. v2 removes __func__ and v1's 1/5.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

ibmvnic: merge ibmvnic_reset_init and ibmvnic_init

These two functions share the majority of the code, hence merge
them together. In the meanwhile, add a reset pass-in parameter
to differentiate them. Thus, the code is easier to read and to tell
the difference between reset_init and regular init.

Signed-off-by: Lijun Pan <ljp@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ibmvnic: remove never executed if statement

At the beginning of the function, from_passive_init is set false by
"adapter->from_passive_init = false;",
hence the if statement will never run.

Signed-off-by: Lijun Pan <ljp@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ibmvnic: improve ibmvnic_init and ibmvnic_reset_init

When H_SEND_CRQ command returns with H_CLOSED, it means the
server's CRQ is not ready yet. Instead of resetting immediately,
we wait for the server to launch passive init.
ibmvnic_init() and ibmvnic_reset_init() should also return the
error code from ibmvnic_send_crq_init() call.

Signed-off-by: Lijun Pan <ljp@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ibmvnic: compare adapter->init_done_rc with more readable ibmvnic_rc_codes

Instead of comparing (adapter->init_done_rc == 1), let it
be (adapter->init_done_rc == PARTIALSUCCESS).

Signed-off-by: Lijun Pan <ljp@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'ipv4-nexthop-Various-improvements'

Ido Schimmel says:

====================
ipv4: nexthop: Various improvements

This patch set contains various improvements that I made to the nexthop
object code while studying it towards my upcoming changes.

While patches #4 and #6 fix bugs, they are not regressions (never
worked). They also do not occur to me as critical issues, which is why I
am targeting them at net-next.

Tested with fib_nexthops.sh:

Tests passed: 134
Tests failed:   0
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

selftests: fib_nexthops: Test IPv6 route with group after replacing IPv4 nexthops

Test that an IPv6 route can not use a nexthop group with mixed IPv4 and
IPv6 nexthops, but can use it after replacing the IPv4 nexthops with
IPv6 nexthops.

Output without previous patch:

# ./fib_nexthops.sh -t ipv6_fcnal_runtime

IPv6 functional runtime
-----------------------
TEST: Route add                                                     [ OK ]
TEST: Route delete                                                  [ OK ]
TEST: Ping with nexthop                                             [ OK ]
TEST: Ping - multipath                                              [ OK ]
TEST: Ping - blackhole                                              [ OK ]
TEST: Ping - blackhole replaced with gateway                        [ OK ]
TEST: Ping - gateway replaced by blackhole                          [ OK ]
TEST: Ping - group with blackhole                                   [ OK ]
TEST: Ping - group blackhole replaced with gateways                 [ OK ]
TEST: IPv6 route with device only nexthop                           [ OK ]
TEST: IPv6 multipath route with nexthop mix - dev only + gw         [ OK ]
TEST: IPv6 route can not have a v4 gateway                          [ OK ]
TEST: Nexthop replace - v6 route, v4 nexthop                        [ OK ]
TEST: Nexthop replace of group entry - v6 route, v4 nexthop         [ OK ]
TEST: IPv6 route can not have a group with v4 and v6 gateways       [ OK ]
TEST: IPv6 route can not have a group with v4 and v6 gateways       [ OK ]
TEST: IPv6 route using a group after removing v4 gateways           [ OK ]
TEST: IPv6 route can not have a group with v4 and v6 gateways       [ OK ]
TEST: IPv6 route can not have a group with v4 and v6 gateways       [ OK ]
TEST: IPv6 route using a group after replacing v4 gateways          [FAIL]
TEST: Nexthop with default route and rpfilter                       [ OK ]
TEST: Nexthop with multipath default route and rpfilter             [ OK ]

Tests passed:  21
Tests failed:   1

Output with previous patch:

# ./fib_nexthops.sh -t ipv6_fcnal_runtime

IPv6 functional runtime
-----------------------
TEST: Route add                                                     [ OK ]
TEST: Route delete                                                  [ OK ]
TEST: Ping with nexthop                                             [ OK ]
TEST: Ping - multipath                                              [ OK ]
TEST: Ping - blackhole                                              [ OK ]
TEST: Ping - blackhole replaced with gateway                        [ OK ]
TEST: Ping - gateway replaced by blackhole                          [ OK ]
TEST: Ping - group with blackhole                                   [ OK ]
TEST: Ping - group blackhole replaced with gateways                 [ OK ]
TEST: IPv6 route with device only nexthop                           [ OK ]
TEST: IPv6 multipath route with nexthop mix - dev only + gw         [ OK ]
TEST: IPv6 route can not have a v4 gateway                          [ OK ]
TEST: Nexthop replace - v6 route, v4 nexthop                        [ OK ]
TEST: Nexthop replace of group entry - v6 route, v4 nexthop         [ OK ]
TEST: IPv6 route can not have a group with v4 and v6 gateways       [ OK ]
TEST: IPv6 route can not have a group with v4 and v6 gateways       [ OK ]
TEST: IPv6 route using a group after removing v4 gateways           [ OK ]
TEST: IPv6 route can not have a group with v4 and v6 gateways       [ OK ]
TEST: IPv6 route can not have a group with v4 and v6 gateways       [ OK ]
TEST: IPv6 route using a group after replacing v4 gateways          [ OK ]
TEST: Nexthop with default route and rpfilter                       [ OK ]
TEST: Nexthop with multipath default route and rpfilter             [ OK ]

Tests passed:  22
Tests failed:   0

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: nexthop: Correctly update nexthop group when replacing a nexthop

Each nexthop group contains an indication if it has IPv4 nexthops
('has_v4'). Its purpose is to prevent IPv6 routes from using groups with
IPv4 nexthops.

However, the indication is not updated when a nexthop is replaced. This
results in the kernel wrongly rejecting IPv6 routes from pointing to
groups that only contain IPv6 nexthops. Example:

# ip nexthop replace id 1 via 192.0.2.2 dev dummy10
# ip nexthop replace id 10 group 1
# ip nexthop replace id 1 via 2001:db8:1::2 dev dummy10
# ip route replace 2001:db8:10::/64 nhid 10
Error: IPv6 routes can not use an IPv4 nexthop.

Solve this by iterating over all the nexthop groups that the replaced
nexthop is a member of and potentially update their IPv4 indication
according to the new set of member nexthops.

Avoid wasting cycles by only performing the update in case an IPv4
nexthop is replaced by an IPv6 nexthop.

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

selftests: fib_nexthops: Test IPv6 route with group after removing IPv4 nexthops

Test that an IPv6 route can not use a nexthop group with mixed IPv4 and
IPv6 nexthops, but can use it after deleting the IPv4 nexthops.

Output without previous patch:

# ./fib_nexthops.sh -t ipv6_fcnal_runtime

IPv6 functional runtime
-----------------------
TEST: Route add                                                     [ OK ]
TEST: Route delete                                                  [ OK ]
TEST: Ping with nexthop                                             [ OK ]
TEST: Ping - multipath                                              [ OK ]
TEST: Ping - blackhole                                              [ OK ]
TEST: Ping - blackhole replaced with gateway                        [ OK ]
TEST: Ping - gateway replaced by blackhole                          [ OK ]
TEST: Ping - group with blackhole                                   [ OK ]
TEST: Ping - group blackhole replaced with gateways                 [ OK ]
TEST: IPv6 route with device only nexthop                           [ OK ]
TEST: IPv6 multipath route with nexthop mix - dev only + gw         [ OK ]
TEST: IPv6 route can not have a v4 gateway                          [ OK ]
TEST: Nexthop replace - v6 route, v4 nexthop                        [ OK ]
TEST: Nexthop replace of group entry - v6 route, v4 nexthop         [ OK ]
TEST: IPv6 route can not have a group with v4 and v6 gateways       [ OK ]
TEST: IPv6 route can not have a group with v4 and v6 gateways       [ OK ]
TEST: IPv6 route using a group after deleting v4 gateways           [FAIL]
TEST: Nexthop with default route and rpfilter                       [ OK ]
TEST: Nexthop with multipath default route and rpfilter             [ OK ]

Tests passed:  18
Tests failed:   1

Output with previous patch:

bash-5.0# ./fib_nexthops.sh -t ipv6_fcnal_runtime

IPv6 functional runtime
-----------------------
TEST: Route add                                                     [ OK ]
TEST: Route delete                                                  [ OK ]
TEST: Ping with nexthop                                             [ OK ]
TEST: Ping - multipath                                              [ OK ]
TEST: Ping - blackhole                                              [ OK ]
TEST: Ping - blackhole replaced with gateway                        [ OK ]
TEST: Ping - gateway replaced by blackhole                          [ OK ]
TEST: Ping - group with blackhole                                   [ OK ]
TEST: Ping - group blackhole replaced with gateways                 [ OK ]
TEST: IPv6 route with device only nexthop                           [ OK ]
TEST: IPv6 multipath route with nexthop mix - dev only + gw         [ OK ]
TEST: IPv6 route can not have a v4 gateway                          [ OK ]
TEST: Nexthop replace - v6 route, v4 nexthop                        [ OK ]
TEST: Nexthop replace of group entry - v6 route, v4 nexthop         [ OK ]
TEST: IPv6 route can not have a group with v4 and v6 gateways       [ OK ]
TEST: IPv6 route can not have a group with v4 and v6 gateways       [ OK ]
TEST: IPv6 route using a group after deleting v4 gateways           [ OK ]
TEST: Nexthop with default route and rpfilter                       [ OK ]
TEST: Nexthop with multipath default route and rpfilter             [ OK ]

Tests passed:  19
Tests failed:   0

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: nexthop: Correctly update nexthop group when removing a nexthop

Each nexthop group contains an indication if it has IPv4 nexthops
('has_v4'). Its purpose is to prevent IPv6 routes from using groups with
IPv4 nexthops.

However, the indication is not updated when a nexthop is removed. This
results in the kernel wrongly rejecting IPv6 routes from pointing to
groups that only contain IPv6 nexthops. Example:

# ip nexthop replace id 1 via 192.0.2.2 dev dummy10
# ip nexthop replace id 2 via 2001:db8:1::2 dev dummy10
# ip nexthop replace id 10 group 1/2
# ip nexthop del id 1
# ip route replace 2001:db8:10::/64 nhid 10
Error: IPv6 routes can not use an IPv4 nexthop.

Solve this by updating the indication according to the new set of
member nexthops.

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>