sfrench/cifs-2.6.git
5 years agoMerge branch 'Convert-mv88e6060-to-mdio-device'
David S. Miller [Wed, 1 May 2019 03:15:45 +0000 (23:15 -0400)]
Merge branch 'Convert-mv88e6060-to-mdio-device'

Andrew Lunn says:

====================
Convert mv88e6060 to mdio device

This patchset builds upon the previous patches to mv88e6060. It adds
support for probing the switch as an MDIO device and then removes the
legacy probe method. Since this is the last device supporting legacy
probe, this allows legacy probe to be removed, originally planned to
be removed in 4.17, but took a bit longer.

This change to the mv88e6060 is more risky than the previous
patchset. Some attempts to test it have been made, by hacking the
driver to match on an mv88e6352 so that it probes. These changes are
all about probe, so it is a reasonable test. But testing on a real
mv88e6060 would be great.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodt-bindings: net: DSA: Remove legacy binding
Andrew Lunn [Sun, 28 Apr 2019 00:56:24 +0000 (02:56 +0200)]
dt-bindings: net: DSA: Remove legacy binding

Now that the code to support the legacy binding has been removed,
remove the documentation for it.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: Remove legacy probing support
Andrew Lunn [Sun, 28 Apr 2019 00:56:23 +0000 (02:56 +0200)]
net: dsa: Remove legacy probing support

Now that all drivers can be probed using more traditional methods,
remove the legacy probe code.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: mv88e6060: Remove support for legacy probing
Andrew Lunn [Sun, 28 Apr 2019 00:56:22 +0000 (02:56 +0200)]
net: dsa: mv88e6060: Remove support for legacy probing

Now that the driver can be probed as an mdio device, remove the legacy
DSA platform device probing.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: mv88e6060: Support probing as an mdio device
Andrew Lunn [Sun, 28 Apr 2019 00:56:21 +0000 (02:56 +0200)]
net: dsa: mv88e6060: Support probing as an mdio device

Probing DSA devices as platform devices has been superseded by using
normal bus drivers. Add support for probing the mv88e6060 device as an
mdio device.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'dsa-core-vlan'
David S. Miller [Wed, 1 May 2019 03:05:30 +0000 (23:05 -0400)]
Merge branch 'dsa-core-vlan'

Vladimir Oltean says:

====================
Improvements to DSA core VLAN manipulation

In preparation of submitting the NXP SJA1105 driver, the Broadcom b53
and Mediatek mt7530 drivers have been found to apply some VLAN
workarounds that are needed in the new driver as well.

Therefore this patchset is mostly simply promoting the DSA driver
workarounds for VLAN to the generic code.

The b53 driver was applying a few workarounds in order to convince DSA
that its vlan_filtering setting is not really per-port. This is now
simply set by the driver via a DSA variable at probe time. The sja1105
driver will be a second user of this.

The mt7530 was also keeping track of when the .port_vlan_filtering
callback was being called. Remove the kept state from this driver
and simplify dealing with vlan_filtering in the generic case.

TODO:

Find the best way to deal generically with the situation described below
(discussion at https://lkml.org/lkml/2019/4/16/1355):

> > +Segregating the switch ports in multiple bridges is supported (e.g. 2 + 2), but
> > +all bridges should have the same level of VLAN awareness (either both have
> > +``vlan_filtering`` 0, or both 1). Also an inevitable limitation of the fact
> > +that VLAN awareness is global at the switch level is that once a bridge with
> > +``vlan_filtering`` enslaves at least one switch port, the other un-bridged
> > +ports are no longer available for standalone traffic termination.
>
> That is quite a limitation that I don't think I had fully grasped until
> reading your different patches. Since enslaving ports into a bridge
> comes after the network device was already made available for use, maybe
> you should force the carrier down or something along those lines as soon
> as a port is enslaved into a bridge with vlan_filtering=1 to make this
> more predictable for the user?
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: Add more convenient functions for installing port VLANs
Vladimir Oltean [Sun, 28 Apr 2019 18:45:54 +0000 (21:45 +0300)]
net: dsa: Add more convenient functions for installing port VLANs

This hides the need to perform a two-phase transaction and construct a
switchdev_obj_port_vlan struct.

Call graph (including a function that will be introduced in a follow-up
patch) looks like this now (same for the *_vlan_del function):

dsa_slave_vlan_rx_add_vid   dsa_port_setup_8021q_tagging
            |                        |
            |                        |
            |          +-------------+
            |          |
            v          v
           dsa_port_vid_add      dsa_slave_port_obj_add
                  |                         |
                  +-------+         +-------+
                          |         |
                          v         v
                       dsa_port_vlan_add

Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: b53: Use vlan_filtering property from dsa_switch
Vladimir Oltean [Sun, 28 Apr 2019 18:45:53 +0000 (21:45 +0300)]
net: dsa: b53: Use vlan_filtering property from dsa_switch

While possible (and safe) to use the newly introduced
dsa_port_is_vlan_filtering helper, fabricating a dsa_port pointer is a
bit awkward, so simply retrieve this from the dsa_switch structure.

Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: b53: Let DSA call .port_vlan_filtering only when necessary
Vladimir Oltean [Sun, 28 Apr 2019 18:45:52 +0000 (21:45 +0300)]
net: dsa: b53: Let DSA call .port_vlan_filtering only when necessary

Since DSA has recently learned to treat better with drivers that set
vlan_filtering_is_global, doing this is no longer required.

Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: Skip calling .port_vlan_filtering on no change
Vladimir Oltean [Sun, 28 Apr 2019 18:45:51 +0000 (21:45 +0300)]
net: dsa: Skip calling .port_vlan_filtering on no change

Even if VLAN filtering is global, DSA will call this callback once per
each port. Drivers should not have to compare the global state with the
requested change. So let DSA do it.

Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: mt7530: Use the DSA vlan_filtering helper function
Vladimir Oltean [Sun, 28 Apr 2019 18:45:50 +0000 (21:45 +0300)]
net: dsa: mt7530: Use the DSA vlan_filtering helper function

This was recently introduced, so keeping state inside the driver is no
longer necessary.

Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Suggested-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: Add helper function to retrieve VLAN awareness setting
Vladimir Oltean [Sun, 28 Apr 2019 18:45:49 +0000 (21:45 +0300)]
net: dsa: Add helper function to retrieve VLAN awareness setting

Since different types of hardware may or may not support this setting
per-port, DSA keeps it either in dsa_switch or in dsa_port.

While drivers may know the characteristics of their hardware and
retrieve it from the correct place without the need of helpers, it is
cumbersone to find out an unambigous answer from generic DSA code.

Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: Keep the vlan_filtering setting in dsa_switch if it's global
Vladimir Oltean [Sun, 28 Apr 2019 18:45:48 +0000 (21:45 +0300)]
net: dsa: Keep the vlan_filtering setting in dsa_switch if it's global

The current behavior is not as obvious as one would assume (which is
that, if the driver set vlan_filtering_is_global = 1, then checking any
dp->vlan_filtering would yield the same result). Only the ports which
are actively enslaved into a bridge would have vlan_filtering set.

This makes it tricky for drivers to check what the global state is.
So fix this and make the struct dsa_switch hold this global setting.

Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: mt7530: Let DSA handle the unsetting of vlan_filtering
Vladimir Oltean [Sun, 28 Apr 2019 18:45:47 +0000 (21:45 +0300)]
net: dsa: mt7530: Let DSA handle the unsetting of vlan_filtering

The driver, recognizing that the .port_vlan_filtering callback was never
coming after the port left its parent bridge, decided to take that duty
in its own hands. DSA now takes care of this condition, so fix that.

Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: Unset vlan_filtering when ports leave the bridge
Vladimir Oltean [Sun, 28 Apr 2019 18:45:46 +0000 (21:45 +0300)]
net: dsa: Unset vlan_filtering when ports leave the bridge

When ports are standalone (after they left the bridge), they should have
no VLAN filtering semantics (they should pass all traffic to the CPU).
Currently this is not true for switchdev drivers, because the bridge
"forgets" to unset that.

Normally one would think that doing this at the bridge layer would be a
better idea, i.e. call br_vlan_filter_toggle() from br_del_if(), similar
to how nbp_vlan_init() is called from br_add_if().

However what complicates that approach, and makes this one preferable,
is the fact that for the bridge core, vlan_filtering is a per-bridge
setting, whereas for switchdev/DSA it is per-port. Also there are
switches where the setting is per the entire device, and unsetting
vlan_filtering one by one, for each leaving port, would not be possible
from the bridge core without a certain level of awareness. So do this in
DSA and let drivers be unaware of it.

Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: b53: Let DSA handle mismatched VLAN filtering settings
Vladimir Oltean [Sun, 28 Apr 2019 18:45:45 +0000 (21:45 +0300)]
net: dsa: b53: Let DSA handle mismatched VLAN filtering settings

The DSA core is now able to do this check prior to calling the
.port_vlan_filtering callback, so tell it that VLAN filtering is global
for this particular hardware.

Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Suggested-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: Be aware of switches where VLAN filtering is a global setting
Vladimir Oltean [Sun, 28 Apr 2019 18:45:44 +0000 (21:45 +0300)]
net: dsa: Be aware of switches where VLAN filtering is a global setting

On some switches, the action of whether to parse VLAN frame headers and use
that information for ingress admission is configurable, but not per
port. Such is the case for the Broadcom BCM53xx and the NXP SJA1105
families, for example. In that case, DSA can prevent the bridge core
from trying to apply different VLAN filtering settings on net devices
that belong to the same switch.

Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Suggested-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: Store vlan_filtering as a property of dsa_port
Vladimir Oltean [Sun, 28 Apr 2019 18:45:43 +0000 (21:45 +0300)]
net: dsa: Store vlan_filtering as a property of dsa_port

This allows drivers to query the VLAN setting imposed by the bridge
driver directly from DSA, instead of keeping their own state based on
the .port_vlan_filtering callback.

Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: Fix pharse -> phase typo
Vladimir Oltean [Sun, 28 Apr 2019 18:45:42 +0000 (21:45 +0300)]
net: dsa: Fix pharse -> phase typo

Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec...
David S. Miller [Tue, 30 Apr 2019 13:26:13 +0000 (09:26 -0400)]
Merge branch 'master' of git://git./linux/kernel/git/klassert/ipsec-next

Steffen Klassert says:

====================
pull request (net-next): ipsec-next 2019-04-30

1) A lot of work to remove indirections from the xfrm code.
   From Florian Westphal.

2) Support ESP offload in combination with gso partial.
   From Boris Pismenny.

3) Remove some duplicated code from vti4.
   From Jeremy Sowden.

Please note that there is merge conflict

between commit:

8742dc86d0c7 ("xfrm4: Fix uninitialized memory read in _decode_session4")

from the ipsec tree and commit:

c53ac41e3720 ("xfrm: remove decode_session indirection from afinfo_policy")

from the ipsec-next tree. The merge conflict will appear
when those trees get merged during the merge window.
The conflict can be solved as it is done in linux-next:

https://lkml.org/lkml/2019/4/25/1207

Please pull or let me know if there are problems.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: micrel: make sure the factory test bit is cleared
Antoine Tenart [Fri, 26 Apr 2019 16:41:23 +0000 (18:41 +0200)]
net: phy: micrel: make sure the factory test bit is cleared

The KSZ8081 PHY has a factory test mode which is set at the de-assertion
of the reset line based on the RXER (KSZ8081RNA/RND) or TXC
(KSZ8081MNX/RNB) pin. If a pull-down is missing, or if the pin has a
pull-up, the factory test mode should be cleared by manually writing a 0
(according to the datasheet). This patch makes sure this factory test
bit is cleared in config_init().

Tested-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: avoid unneeded MDIO reads in genphy_read_status
Heiner Kallweit [Wed, 24 Apr 2019 19:49:30 +0000 (21:49 +0200)]
net: phy: avoid unneeded MDIO reads in genphy_read_status

Considering that in polling mode each link drop will be latched,
settings can't have changed if link was up and is up.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'dsa-tag-modules'
David S. Miller [Sun, 28 Apr 2019 23:41:01 +0000 (19:41 -0400)]
Merge branch 'dsa-tag-modules'

Andrew Lunn says:

====================
Make DSA tag drivers kernel modules

Historically, DSA tag drivers have been compiled into the kernel as
part of the DSA core. With the growing number of tag drivers, it makes
sense to allow this driver code to be compiled as a module, and loaded
on demand.

v2
--
Move name to end of structure, keeping the hot entries at the beginning.
More tag protocol to end of structure to keep hot members at the beginning.
Fix indent of #endif
Rewrite to move list pointer into a new structure
void functions, since there cannot be errors
Fix fall-through comment
Reorder patch for unused symbols to before tag drivers can be modules
tab/space cleanup
Help text wording
NET_DSA_TAG_BRCM_COMMON and NET_DSA_TAG_KZS_COMMON hidden

v3
--
boilerplate: Move kdoc next to macro
boilerplate: Fix THIS_MODULE indentation
Kconfig: More tabification
Kconfig: Punctuation

v4
--
Cover note {H}istorically
Kconfig: trailer
====================

Tested-by: Vivien Didelot <vivien.didelot@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodsa: Allow tag drivers to be built as modules
Andrew Lunn [Sun, 28 Apr 2019 17:37:23 +0000 (19:37 +0200)]
dsa: Allow tag drivers to be built as modules

Make the CONFIG symbols tristate and add help text.

The broadcom and Microchip KSZ tag drivers support two different
tagging protocols in one driver. Add a configuration option for the
drivers, and then options to select the protocol.

Create a submenu for the tagging drivers.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
v2:
tab/space cleanup
Help text wording
NET_DSA_TAG_BRCM_COMMON and NET_DSA_TAG_KZS_COMMON hidden

v3:
More tabification
Punctuation

v4:
trailler->trailer

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodsa: tag_brcm: Avoid unused symbols
Andrew Lunn [Sun, 28 Apr 2019 17:37:22 +0000 (19:37 +0200)]
dsa: tag_brcm: Avoid unused symbols

It is possible that the driver is compiled with both
CONFIG_NET_DSA_TAG_BRCM and CONFIG_NET_DSA_TAG_BRCM_PREPEND disabled.
This results in warnings about unused symbols. Add some conditional
compilation to avoid this.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
v2
Reorder patch to before tag drivers can be modules

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodsa: Cleanup unneeded table and make tag structures static
Andrew Lunn [Sun, 28 Apr 2019 17:37:21 +0000 (19:37 +0200)]
dsa: Cleanup unneeded table and make tag structures static

Now that tag drivers dynamically register, we don't need the static
table. Remove it. This also means the tag driver structures can be
made static.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodsa: Make use of the list of tag drivers
Andrew Lunn [Sun, 28 Apr 2019 17:37:20 +0000 (19:37 +0200)]
dsa: Make use of the list of tag drivers

Implement the _get and _put functions to make use of the list of tag
drivers. Also, trigger the loading of the module, based on the alias
information. The _get function takes a reference on the tag driver, so
it cannot be unloaded, and the _put function releases the reference.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
v2:
Make tag_driver_register void

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodsa: Add stub tag driver put method
Andrew Lunn [Sun, 28 Apr 2019 17:37:19 +0000 (19:37 +0200)]
dsa: Add stub tag driver put method

When a DSA switch driver is unloaded, the lock on the tag driver
should be released so the module can be unloaded. Add the needed calls,
but leave the actual release code as a stub.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
v2

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodsa: Rename dsa_resolve_tag_protocol() to _get ready for locking
Andrew Lunn [Sun, 28 Apr 2019 17:37:18 +0000 (19:37 +0200)]
dsa: Rename dsa_resolve_tag_protocol() to _get ready for locking

dsa_resolve_tag_protocol() is used to find the tagging driver needed
by a switch driver. When the tagging drivers become modules, it will
be necassary to take a reference on the module to prevent it being
unloaded. So rename this function to _get() to indicate it has some
locking properties.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodsa: Register the none tagger ops
Andrew Lunn [Sun, 28 Apr 2019 17:37:17 +0000 (19:37 +0200)]
dsa: Register the none tagger ops

The none tagger is special in that it does not live in a tag_*.c file,
but is within the core. Register/unregister when DSA is
loaded/unloaded.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodsa: Keep link list of tag drivers
Andrew Lunn [Sun, 28 Apr 2019 17:37:16 +0000 (19:37 +0200)]
dsa: Keep link list of tag drivers

Let the tag drivers register themselves with the DSA core, keeping
them in a linked list.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
v2

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodsa: Add boilerplate helper to register DSA tag driver modules
Andrew Lunn [Sun, 28 Apr 2019 17:37:15 +0000 (19:37 +0200)]
dsa: Add boilerplate helper to register DSA tag driver modules

A DSA tag driver module will need to register the tag protocols it
implements with the DSA core. Add macros containing this boiler plate.

The registration/unregistration code is currently just a stub. A Later
patch will add the real implementation.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
v2
Fix indent of #endif
Rewrite to move list pointer into a new structure
v3
Move kdoc next to macro
Fix THIS_MODULE indentation

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodsa: Add TAG protocol to tag ops
Andrew Lunn [Sun, 28 Apr 2019 17:37:14 +0000 (19:37 +0200)]
dsa: Add TAG protocol to tag ops

In order that we can match the tagging protocol a switch driver
request to the tagger, we need to know what protocol the tagger
supports. Add this information to the ops structure.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
v2
More tag protocol to end of structure to keep hot members at the beginning.

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodsa: Add MODULE_LICENSE to tag drivers
Andrew Lunn [Sun, 28 Apr 2019 17:37:13 +0000 (19:37 +0200)]
dsa: Add MODULE_LICENSE to tag drivers

All the tag drivers are some variant of GPL. Add a MODULE_LICENSE()
indicating this, so the drivers can later be compiled as modules.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodsa: Add MODULE_ALIAS to taggers in preparation to become modules
Andrew Lunn [Sun, 28 Apr 2019 17:37:12 +0000 (19:37 +0200)]
dsa: Add MODULE_ALIAS to taggers in preparation to become modules

When the tag drivers become modules, we will need to dynamically load
them based on what the switch drivers need. Add aliases to map between
the TAG protocol and the driver.

In order to do this, we need the tag protocol number as something
which the C pre-processor can stringinfy. Only the compiler knows the
value of an enum, CPP cannot use them. So add #defines.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodsa: Move tagger name into its ops structure
Andrew Lunn [Sun, 28 Apr 2019 17:37:11 +0000 (19:37 +0200)]
dsa: Move tagger name into its ops structure

Rather than keep a list to map a tagger ops to a name, place the name
into the ops structure. This removes the hard coded list, a step
towards making the taggers more dynamic.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
v2:
Move name to end of structure, keeping the hot entries at the beginning.

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodsa: Add SPDX header to tag drivers.
Andrew Lunn [Sun, 28 Apr 2019 17:37:10 +0000 (19:37 +0200)]
dsa: Add SPDX header to tag drivers.

Add an SPDX header, and remove the license boilerplate text.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next
David S. Miller [Sun, 28 Apr 2019 12:42:41 +0000 (08:42 -0400)]
Merge git://git./linux/kernel/git/bpf/bpf-next

Daniel Borkmann says:

====================
pull-request: bpf-next 2019-04-28

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

The main changes are:

1) Introduce BPF socket local storage map so that BPF programs can store
   private data they associate with a socket (instead of e.g. separate hash
   table), from Martin.

2) Add support for bpftool to dump BTF types. This is done through a new
   `bpftool btf dump` sub-command, from Andrii.

3) Enable BPF-based flow dissector for skb-less eth_get_headlen() calls which
   was currently not supported since skb was used to lookup netns, from Stanislav.

4) Add an opt-in interface for tracepoints to expose a writable context
   for attached BPF programs, used here for NBD sockets, from Matt.

5) BPF xadd related arm64 JIT fixes and scalability improvements, from Daniel.

6) Change the skb->protocol for bpf_skb_adjust_room() helper in order to
   support tunnels such as sit. Add selftests as well, from Willem.

7) Various smaller misc fixes.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agocxgb4: Delete all hash and TCAM filters before resource cleanup
Vishal Kulkarni [Fri, 26 Apr 2019 08:28:48 +0000 (13:58 +0530)]
cxgb4: Delete all hash and TCAM filters before resource cleanup

During driver unload, hash/TCAM filter deletion doesn't wait for
completion.This patch deletes all the filters with completion before
clearing the resources.

Signed-off-by: Vishal Kulkarni <vishal@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: mv88e6xxx: Remove legacy probe support
Andrew Lunn [Sat, 27 Apr 2019 17:19:10 +0000 (19:19 +0200)]
net: dsa: mv88e6xxx: Remove legacy probe support

Remove the legacy method of probing the mv88e6xxx driver, now that all
the mainline boards have been converted to use mdio based probing for
a number of cycles.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'mv88e6060-cleanups'
David S. Miller [Sun, 28 Apr 2019 00:23:05 +0000 (20:23 -0400)]
Merge branch 'mv88e6060-cleanups'

Andrew Lunn says:

====================
mv88e6060 cleanups

This patchset performs some cleanups of the mv88e6060 DSA driver, as a
step towards making it an MDIO device, rather than use the old probing
method. The changes here are all pretty mechanical and only compile
tested.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: mv88e6060: Replace REG_READ macro
Andrew Lunn [Sat, 27 Apr 2019 17:32:59 +0000 (19:32 +0200)]
net: dsa: mv88e6060: Replace REG_READ macro

The REG_READ macro contains a return statement, making it not very
safe. Remove it by inlining the code.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: mv88e6060: Replace REG_WRITE macro
Andrew Lunn [Sat, 27 Apr 2019 17:32:58 +0000 (19:32 +0200)]
net: dsa: mv88e6060: Replace REG_WRITE macro

The REG_WRITE macro contains a return statement, making it not very
safe. Remove it by inlining the code.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: mv88e6060: Replace ds with priv
Andrew Lunn [Sat, 27 Apr 2019 17:32:57 +0000 (19:32 +0200)]
net: dsa: mv88e6060: Replace ds with priv

Pass around priv, not ds. This will help with changing to an mdio
driver, and makes this driver more like mv88e6xxx.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: mv88e6060: Add SPDX header
Andrew Lunn [Sat, 27 Apr 2019 17:32:56 +0000 (19:32 +0200)]
net: dsa: mv88e6060: Add SPDX header

Add an SPDX header, and remove the license text.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoibmvnic: Add device identification to requested IRQs
Murilo Fossa Vicentini [Thu, 25 Apr 2019 14:02:33 +0000 (11:02 -0300)]
ibmvnic: Add device identification to requested IRQs

The ibmvnic driver currently uses the same fixed name when using
request_irq, this makes it hard to parse when multiple VNIC devices are
available at the same time. This patch adds the unit_address as the device
identification along with an id for each queue.

The original idea was to use the interface name as an identifier, but it
is not feasible given these requests happen at adapter probe, and at this
point netdev is not yet registered so it doesn't have the proper name
assigned to it.

Signed-off-by: Murilo Fossa Vicentini <muvic@linux.ibm.com>
Reviewed-by: Mauro S. M. Rodrigues <maurosr@linux.vnet.ibm.com>
Reviewed-by: Thomas Falcon <tlfalcon@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agocpsw: Put back cpsw_ndo_poll_controller()
David S. Miller [Sun, 28 Apr 2019 00:08:25 +0000 (20:08 -0400)]
cpsw: Put back cpsw_ndo_poll_controller()

To fix the build.

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'net-ethernet-ti-clean-up-and-optimizations'
David S. Miller [Sat, 27 Apr 2019 21:11:49 +0000 (17:11 -0400)]
Merge branch 'net-ethernet-ti-clean-up-and-optimizations'

Grygorii Strashko says:

====================
net: ethernet: ti: clean up and optimizations

This is a preparation series for introducing new switchbase TI CPSW driver which
was originally introduced [1][2] by Ilias Apalodimas <ilias.apalodimas@linaro.org>
and also discussed in private mails and at Netdev x13 confernce.

Following discussions and suggestions (mostly by Andrew and Ivan) we going
to introduce the new driver which is operating in dual-emac mode
by default, thus working as 2 individual network interfaces.
When both interfaces joined the bridge - CPSW driver will enter a switch
mode and discard dual_mac configuration. The CPSW will be switched back
to dual_mac mode if any port leaves the bridge. All configuration is going to be
implemented via switchdev API.

Hence overall change is already very big I'm sending prerequisite patches which
are mostly minor fixes/clean ups and code refactoring to separate common parts
to be reused by both drivers.
Probably the most serious change from functional point of view is Patch 11.

These patches were NFS boot tetested on TI AM335x/AM437x/AM5xx boards.

These patches can be found at:
 git@git.ti.com:~gragst/ti-linux-kernel/gragsts-ti-linux-kernel.git
 branch: lkml-5.1-cpsw-clean-up-v2

changes in v2:
- added new patch 16 to get rid of force type conversation
- other chages metioned in patches
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ethernet: ti: cpsw: move ethtool func in separate file
Grygorii Strashko [Fri, 26 Apr 2019 17:12:42 +0000 (20:12 +0300)]
net: ethernet: ti: cpsw: move ethtool func in separate file

As a preparatory patch to add support for a switchdev based cpsw driver,
move common ethtool functions to separate cpsw-ethtool.c file so that they
can be used across both drivers. It will simplify CPSW driver code
maintenance also.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ethernet: ti: cpsw: switch to use mac sl api
Grygorii Strashko [Fri, 26 Apr 2019 17:12:41 +0000 (20:12 +0300)]
net: ethernet: ti: cpsw: switch to use mac sl api

Switch CPSW driver to use the new MAC SL API.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ethernet: ti: cpsw: introduce mac sl module api
Grygorii Strashko [Fri, 26 Apr 2019 17:12:40 +0000 (20:12 +0300)]
net: ethernet: ti: cpsw: introduce mac sl module api

The MAC SL submodule has a lot of common functions between many of TI SoCs
AM335x/AM437x/DRA7(AM57xx), Keystone 2 66AK2HK/E/L/G and K3 AM654, but
there are also differences especially in registers offsets and sets of
supported functions.

This patch introduces the MAC SL submodule API which is intended to provide
a common way to access the MAC SL submodule and hide HW integrations
details.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: Sekhar Nori <nsekhar@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ethernet: ti: cpsw: move common hw init code in separate func
Grygorii Strashko [Fri, 26 Apr 2019 17:12:39 +0000 (20:12 +0300)]
net: ethernet: ti: cpsw: move common hw init code in separate func

move common hw init code in separate function as preparation for adding new
switchdev driver.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ethernet: ti: davinci_cpdma: use dma_addr_t for desc_mem_phys and desc_hw_addr
Grygorii Strashko [Fri, 26 Apr 2019 17:12:38 +0000 (20:12 +0300)]
net: ethernet: ti: davinci_cpdma: use dma_addr_t for desc_mem_phys and desc_hw_addr

Use dma_addr_t for desc_mem_phys and desc_hw_addr to avoid types
conversions.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ethernet: ti: cpsw: move cpsw definitions in priv header
Grygorii Strashko [Fri, 26 Apr 2019 17:12:37 +0000 (20:12 +0300)]
net: ethernet: ti: cpsw: move cpsw definitions in priv header

As a preparatory patch to add a switchdev based cpsw driver move the common
header definitions to cpsw_priv.h. The plan is to develop a new driver on
switchdev driver model and obsolete the current cpsw driver after all
required functions are added to the new driver. This patch allows the same
header file to be re-used on both drivers during the transition period.

Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ethernet: ti: cpsw: refactor probe to group common hw initialization
Grygorii Strashko [Fri, 26 Apr 2019 17:12:36 +0000 (20:12 +0300)]
net: ethernet: ti: cpsw: refactor probe to group common hw initialization

Rework probe to group common hw initialization:
- group resources request at the beginning of the probe
- move net device initialization and registration at the end of the probe
- drop cpsw_slave_init
as preparation of refactoring of common hw initialization code to
separate function.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ethernet: ti: davinci_mdio: use devm_ioremap()
Grygorii Strashko [Fri, 26 Apr 2019 17:12:35 +0000 (20:12 +0300)]
net: ethernet: ti: davinci_mdio: use devm_ioremap()

The Davinci MDIO in most of the case implemented as module inside of TI
CPSW subsystem and fully depends on CPSW to be enabled, but historically
it's implemented as separate Platform device/driver and defined in DT files
in two ways:
- as standalone node
- as child node of CPSW subsystem.

In later case it's required to split CPSW subsystem "reg" property to
exclude MDIO I/O range which is not useful.

Hence, replace devm_ioremap_resource() with devm_ioremap() to allow define
full I/O range in parent CPSW subsystem without spliting.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ethernet: ti: ale: do not auto delete mcast super entries
Grygorii Strashko [Fri, 26 Apr 2019 17:12:34 +0000 (20:12 +0300)]
net: ethernet: ti: ale: do not auto delete mcast super entries

Do not delete multicast supervisory packet's (SUPER) entries while flushing
multicast addresses from ALE table cpsw_ale_flush_multicast(). Those
entries have to be added/removed only explicitly.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ethernet: ti: cpsw: fix allmulti cfg in dual_mac mode
Grygorii Strashko [Fri, 26 Apr 2019 17:12:33 +0000 (20:12 +0300)]
net: ethernet: ti: cpsw: fix allmulti cfg in dual_mac mode

Now CPSW ALE will set/clean Host port bit in Unregistered Multicast Flood
Mask (UNREG_MCAST_FLOOD_MASK) for every VLAN without checking if this port
belongs to VLAN or not when ALLMULTI mode flag is set for nedev. This is
working in non dual_mac mode, but in dual_mac - it causes
enabling/disabling ALLMULTI flag for both ports.

Hence fix it by adding additional parameter to cpsw_ale_set_allmulti() to
specify ALE port number for which ALLMULTI has to be enabled and check if
port belongs to VLAN before modifying UNREG_MCAST_FLOOD_MASK.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ethernet: ti: ale: use define for host port in cpsw_ale_set_allmulti()
Grygorii Strashko [Fri, 26 Apr 2019 17:12:32 +0000 (20:12 +0300)]
net: ethernet: ti: ale: use define for host port in cpsw_ale_set_allmulti()

Use ALE_PORT_HOST define for host port in cpsw_ale_set_allmulti() instead
of constants.

Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ethernet: ti: ale: fix mcast super setting
Grygorii Strashko [Fri, 26 Apr 2019 17:12:31 +0000 (20:12 +0300)]
net: ethernet: ti: ale: fix mcast super setting

Use correct define ALE_SUPER for ALE Multicast Address Table Entry
Supervisory Packet (SUPER) bit setting instead of ALE_BLOCKED. No issues
were observed till now as it have never been set, but it's going to be used
by new CPSW switch driver.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ethernet: ti: cpsw: drop cpsw_tx_packet_submit()
Grygorii Strashko [Fri, 26 Apr 2019 17:12:30 +0000 (20:12 +0300)]
net: ethernet: ti: cpsw: drop cpsw_tx_packet_submit()

Drop unnecessary wrapper function cpsw_tx_packet_submit() which is used
only in one place.

Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ethernet: ti: cpsw: use devm_alloc_etherdev_mqs()
Grygorii Strashko [Fri, 26 Apr 2019 17:12:29 +0000 (20:12 +0300)]
net: ethernet: ti: cpsw: use devm_alloc_etherdev_mqs()

Use devm_alloc_etherdev_mqs() and simplify code.

Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ethernet: ti: cpsw: drop pinctrl_pm_select_default_state call
Grygorii Strashko [Fri, 26 Apr 2019 17:12:28 +0000 (20:12 +0300)]
net: ethernet: ti: cpsw: drop pinctrl_pm_select_default_state call

Drop pinctrl_pm_select_default_state call from probe as default
pinctrl state is set by DD core.

Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ethernet: ti: cpsw: use local var dev in probe
Grygorii Strashko [Fri, 26 Apr 2019 17:12:27 +0000 (20:12 +0300)]
net: ethernet: ti: cpsw: use local var dev in probe

Use local variable struct device *dev in probe to simplify code.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ethernet: ti: cpsw: update cpsw_split_res() to accept cpsw_common
Grygorii Strashko [Fri, 26 Apr 2019 17:12:26 +0000 (20:12 +0300)]
net: ethernet: ti: cpsw: update cpsw_split_res() to accept cpsw_common

Update cpsw_split_res() to accept struct cpsw_common instead of
struct net_device to simplify code.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ethernet: ti: cpsw: drop CONFIG_TI_CPSW_ALE config option
Grygorii Strashko [Fri, 26 Apr 2019 17:12:25 +0000 (20:12 +0300)]
net: ethernet: ti: cpsw: drop CONFIG_TI_CPSW_ALE config option

All TI drivers CPSW/NETCP can't work without ALE, hence simplify
build of those drivers by always linking cpsw_ale and drop
CONFIG_TI_CPSW_ALE config option.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ethernet: ti: cpsw: drop TI_DAVINCI_CPDMA config option
Grygorii Strashko [Fri, 26 Apr 2019 17:12:24 +0000 (20:12 +0300)]
net: ethernet: ti: cpsw: drop TI_DAVINCI_CPDMA config option

Both drivers CPSW and EMAC can't work without CPDMA, hence simplify build
of those drivers by always linking davinci_cpdma and drop TI_DAVINCI_CPDMA
config option.
Note. the davinci_emac driver module was changed to "ti_davinci_emac" to
make build work.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ethernet: ti: convert to SPDX license identifiers
Grygorii Strashko [Fri, 26 Apr 2019 17:12:23 +0000 (20:12 +0300)]
net: ethernet: ti: convert to SPDX license identifiers

Replace textual license with SPDX-License-Identifier.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'strict-netlink-validation'
David S. Miller [Sat, 27 Apr 2019 21:07:22 +0000 (17:07 -0400)]
Merge branch 'strict-netlink-validation'

Johannes Berg says:

====================
strict netlink validation

Here's a respin, with the following changes:
 * change message when rejecting unknown attribute types (David Ahern)
 * drop nl80211 patch - I'll apply it separately
 * remove NL_VALIDATE_POLICY - we have a lot of calls to nla_parse()
   that really should be without a policy as it has previously been
   validated - need to find a good way to handle this later
 * include the correct generic netlink change (d'oh, sorry)
====================

Acked-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agogenetlink: optionally validate strictly/dumps
Johannes Berg [Fri, 26 Apr 2019 12:07:31 +0000 (14:07 +0200)]
genetlink: optionally validate strictly/dumps

Add options to strictly validate messages and dump messages,
sometimes perhaps validating dump messages non-strictly may
be required, so add an option for that as well.

Since none of this can really be applied to existing commands,
set the options everwhere using the following spatch:

    @@
    identifier ops;
    expression X;
    @@
    struct genl_ops ops[] = {
    ...,
     {
            .cmd = X,
    +       .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
            ...
     },
    ...
    };

For new commands one should just not copy the .validate 'opt-out'
flags and thus get strict validation.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonetlink: add strict parsing for future attributes
Johannes Berg [Fri, 26 Apr 2019 12:07:30 +0000 (14:07 +0200)]
netlink: add strict parsing for future attributes

Unfortunately, we cannot add strict parsing for all attributes, as
that would break existing userspace. We currently warn about it, but
that's about all we can do.

For new attributes, however, the story is better: nobody is using
them, so we can reject bad sizes.

Also, for new attributes, we need not accept them when the policy
doesn't declare their usage.

David Ahern and I went back and forth on how to best encode this, and
the best way we found was to have a "boundary type", from which point
on new attributes have all possible validation applied, and NLA_UNSPEC
is rejected.

As we didn't want to add another argument to all functions that get a
netlink policy, the workaround is to encode that boundary in the first
entry of the policy array (which is for type 0 and thus probably not
really valid anyway). I put it into the validation union for the rare
possibility that somebody is actually using attribute 0, which would
continue to work fine unless they tried to use the extended validation,
which isn't likely. We also didn't find any in-tree users with type 0.

The reason for setting the "start strict here" attribute is that we
never really need to start strict from 0, which is invalid anyway (or
in legacy families where that isn't true, it cannot be set to strict),
so we can thus reserve the value 0 for "don't do this check" and don't
have to add the tag to all policies right now.

Thus, policies can now opt in to this validation, which we should do
for all existing policies, at least when adding new attributes.

Note that entirely *new* policies won't need to set it, as the use
of that should be using nla_parse()/nlmsg_parse() etc. which anyway
do fully strict validation now, regardless of this.

So in effect, this patch only covers the "existing command with new
attribute" case.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonetlink: re-add parse/validate functions in strict mode
Johannes Berg [Fri, 26 Apr 2019 12:07:29 +0000 (14:07 +0200)]
netlink: re-add parse/validate functions in strict mode

This re-adds the parse and validate functions like nla_parse()
that are now actually strict after the previous rename and were
just split out to make sure everything is converted (and if not
compilation of the previous patch would fail.)

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonetlink: make validation more configurable for future strictness
Johannes Berg [Fri, 26 Apr 2019 12:07:28 +0000 (14:07 +0200)]
netlink: make validation more configurable for future strictness

We currently have two levels of strict validation:

 1) liberal (default)
     - undefined (type >= max) & NLA_UNSPEC attributes accepted
     - attribute length >= expected accepted
     - garbage at end of message accepted
 2) strict (opt-in)
     - NLA_UNSPEC attributes accepted
     - attribute length >= expected accepted

Split out parsing strictness into four different options:
 * TRAILING     - check that there's no trailing data after parsing
                  attributes (in message or nested)
 * MAXTYPE      - reject attrs > max known type
 * UNSPEC       - reject attributes with NLA_UNSPEC policy entries
 * STRICT_ATTRS - strictly validate attribute size

The default for future things should be *everything*.
The current *_strict() is a combination of TRAILING and MAXTYPE,
and is renamed to _deprecated_strict().
The current regular parsing has none of this, and is renamed to
*_parse_deprecated().

Additionally it allows us to selectively set one of the new flags
even on old policies. Notably, the UNSPEC flag could be useful in
this case, since it can be arranged (by filling in the policy) to
not be an incompatible userspace ABI change, but would then going
forward prevent forgetting attribute entries. Similar can apply
to the POLICY flag.

We end up with the following renames:
 * nla_parse           -> nla_parse_deprecated
 * nla_parse_strict    -> nla_parse_deprecated_strict
 * nlmsg_parse         -> nlmsg_parse_deprecated
 * nlmsg_parse_strict  -> nlmsg_parse_deprecated_strict
 * nla_parse_nested    -> nla_parse_nested_deprecated
 * nla_validate_nested -> nla_validate_nested_deprecated

Using spatch, of course:
    @@
    expression TB, MAX, HEAD, LEN, POL, EXT;
    @@
    -nla_parse(TB, MAX, HEAD, LEN, POL, EXT)
    +nla_parse_deprecated(TB, MAX, HEAD, LEN, POL, EXT)

    @@
    expression NLH, HDRLEN, TB, MAX, POL, EXT;
    @@
    -nlmsg_parse(NLH, HDRLEN, TB, MAX, POL, EXT)
    +nlmsg_parse_deprecated(NLH, HDRLEN, TB, MAX, POL, EXT)

    @@
    expression NLH, HDRLEN, TB, MAX, POL, EXT;
    @@
    -nlmsg_parse_strict(NLH, HDRLEN, TB, MAX, POL, EXT)
    +nlmsg_parse_deprecated_strict(NLH, HDRLEN, TB, MAX, POL, EXT)

    @@
    expression TB, MAX, NLA, POL, EXT;
    @@
    -nla_parse_nested(TB, MAX, NLA, POL, EXT)
    +nla_parse_nested_deprecated(TB, MAX, NLA, POL, EXT)

    @@
    expression START, MAX, POL, EXT;
    @@
    -nla_validate_nested(START, MAX, POL, EXT)
    +nla_validate_nested_deprecated(START, MAX, POL, EXT)

    @@
    expression NLH, HDRLEN, MAX, POL, EXT;
    @@
    -nlmsg_validate(NLH, HDRLEN, MAX, POL, EXT)
    +nlmsg_validate_deprecated(NLH, HDRLEN, MAX, POL, EXT)

For this patch, don't actually add the strict, non-renamed versions
yet so that it breaks compile if I get it wrong.

Also, while at it, make nla_validate and nla_parse go down to a
common __nla_validate_parse() function to avoid code duplication.

Ultimately, this allows us to have very strict validation for every
new caller of nla_parse()/nlmsg_parse() etc as re-introduced in the
next patch, while existing things will continue to work as is.

In effect then, this adds fully strict validation for any new command.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonetlink: add NLA_MIN_LEN
Johannes Berg [Fri, 26 Apr 2019 12:07:27 +0000 (14:07 +0200)]
netlink: add NLA_MIN_LEN

Rather than using NLA_UNSPEC for this type of thing, use NLA_MIN_LEN
so we can make NLA_UNSPEC be NLA_REJECT under certain conditions for
future attributes.

While at it, also use NLA_EXACT_LEN for the struct example.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'nla_nest_start'
David S. Miller [Sat, 27 Apr 2019 21:03:44 +0000 (17:03 -0400)]
Merge branch 'nla_nest_start'

Michal Kubecek says:

====================
make nla_nest_start() add NLA_F_NESTED flag

One of the comments in recent review of the ethtool netlink series pointed
out that proposed ethnl_nest_start() helper which adds NLA_F_NESTED to
second argument of nla_nest_start() is not really specific to ethtool
netlink code. That is hard to argue with as closer inspection revealed that
exactly the same helper already exists in ipset code (except it's a macro
rather than an inline function).

Another observation was that even if NLA_F_NESTED flag was introduced in
2007, only few netlink based interfaces set it in kernel generated messages
and even many recently added APIs omit it. That is unfortunate as without
the flag, message parsers not familiar with attribute semantics cannot
recognize nested attributes and do not see message structure; this affects
e.g. wireshark dissector or mnl_nlmsg_fprintf() from libmnl.

This is why I'm suggesting to rename existing nla_nest_start() to different
name (nla_nest_start_noflag) and reintroduce nla_nest_start() as a wrapper
adding NLA_F_NESTED flag. This is implemented in first patch which is
mostly generated by spatch. Second patch drops ipset helper macros which
lose their purpose. Third patch cleans up minor coding style issues found
by checkpatch.pl in first patch.

We could leave nla_nest_start() untouched and simply add a wrapper adding
NLA_F_NESTED but that would probably preserve the state when even most new
code doesn't set the flag.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: fix two coding style issues
Michal Kubecek [Fri, 26 Apr 2019 09:13:12 +0000 (11:13 +0200)]
net: fix two coding style issues

This is a simple cleanup addressing two coding style issues found by
checkpatch.pl in an earlier patch. It's submitted as a separate patch to
keep the original patch as it was generated by spatch.

Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoipset: drop ipset_nest_start() and ipset_nest_end()
Michal Kubecek [Fri, 26 Apr 2019 09:13:09 +0000 (11:13 +0200)]
ipset: drop ipset_nest_start() and ipset_nest_end()

After the previous commit, both ipset_nest_start() and ipset_nest_end() are
just aliases for nla_nest_start() and nla_nest_end() so that there is no
need to keep them.

Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Acked-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonetlink: make nla_nest_start() add NLA_F_NESTED flag
Michal Kubecek [Fri, 26 Apr 2019 09:13:06 +0000 (11:13 +0200)]
netlink: make nla_nest_start() add NLA_F_NESTED flag

Even if the NLA_F_NESTED flag was introduced more than 11 years ago, most
netlink based interfaces (including recently added ones) are still not
setting it in kernel generated messages. Without the flag, message parsers
not aware of attribute semantics (e.g. wireshark dissector or libmnl's
mnl_nlmsg_fprintf()) cannot recognize nested attributes and won't display
the structure of their contents.

Unfortunately we cannot just add the flag everywhere as there may be
userspace applications which check nlattr::nla_type directly rather than
through a helper masking out the flags. Therefore the patch renames
nla_nest_start() to nla_nest_start_noflag() and introduces nla_nest_start()
as a wrapper adding NLA_F_NESTED. The calls which add NLA_F_NESTED manually
are rewritten to use nla_nest_start().

Except for changes in include/net/netlink.h, the patch was generated using
this semantic patch:

@@ expression E1, E2; @@
-nla_nest_start(E1, E2)
+nla_nest_start_noflag(E1, E2)

@@ expression E1, E2; @@
-nla_nest_start_noflag(E1, E2 | NLA_F_NESTED)
+nla_nest_start(E1, E2)

Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Acked-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'net-tls-small-code-cleanup'
David S. Miller [Sat, 27 Apr 2019 20:52:22 +0000 (16:52 -0400)]
Merge branch 'net-tls-small-code-cleanup'

Jakub Kicinski says:

====================
net/tls: small code cleanup

This small patch set cleans up tls (mostly offload parts).
Other than avoiding unnecessary error messages - no functional
changes here.

v2 (Saeed):
 - fix up Review tags;
 - remove the warning on failure completely.
====================

Reviewed-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/tls: byte swap device req TCP seq no upon setting
Jakub Kicinski [Thu, 25 Apr 2019 19:32:04 +0000 (12:32 -0700)]
net/tls: byte swap device req TCP seq no upon setting

To avoid a sparse warning byteswap the be32 sequence number
before it's stored in the atomic value.  While at it drop
unnecessary brackets and use kernel's u64 type.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/tls: move definition of tls ops into net/tls.h
Jakub Kicinski [Thu, 25 Apr 2019 19:32:03 +0000 (12:32 -0700)]
net/tls: move definition of tls ops into net/tls.h

There seems to be no reason for tls_ops to be defined in netdevice.h
which is included in a lot of places.  Don't wrap the struct/enum
declaration in ifdefs, it trickles down unnecessary ifdefs into
driver code.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/tls: remove old exports of sk_destruct functions
Jakub Kicinski [Thu, 25 Apr 2019 19:32:02 +0000 (12:32 -0700)]
net/tls: remove old exports of sk_destruct functions

tls_device_sk_destruct being set on a socket used to indicate
that socket is a kTLS device one.  That is no longer true -
now we use sk_validate_xmit_skb pointer for that purpose.
Remove the export.  tls_device_attach() needs to be moved.

While at it, remove the dead declaration of tls_sk_destruct().

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/tls: don't log errors every time offload can't proceed
Jakub Kicinski [Thu, 25 Apr 2019 19:32:01 +0000 (12:32 -0700)]
net/tls: don't log errors every time offload can't proceed

Currently when CONFIG_TLS_DEVICE is set each time kTLS
connection is opened and the offload is not successful
(either because the underlying device doesn't support
it or e.g. it's tables are full) a rate limited error
will be printed to the logs.

There is nothing wrong with failing TLS offload.  SW
path will process the packets just fine, drop the
noisy messages.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'sk-local-storage'
Alexei Starovoitov [Sat, 27 Apr 2019 16:07:49 +0000 (09:07 -0700)]
Merge branch 'sk-local-storage'

Martin KaFai Lau says:

====================
v4:
- Move checks to map_alloc_check in patch 1 (Stanislav Fomichev)
- Refactor BTF encoding macros to test_btf.h at
  a new patch 4 (Stanislav Fomichev)
- Refactor getenv and add print PASS message at the
  end of the test in patch 6 (Yonghong Song)

v3:
- Replace spinlock_types.h with spinlock.h in patch 1
  (kbuild test robot <lkp@intel.com>)

v2:
- Add the "test_maps.h" file in patch 5

This series introduces the BPF sk local storage.  The
details is in the patch 1 commit message.
====================

Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agobpf: Add ene-to-end test for bpf_sk_storage_* helpers
Martin KaFai Lau [Fri, 26 Apr 2019 23:39:54 +0000 (16:39 -0700)]
bpf: Add ene-to-end test for bpf_sk_storage_* helpers

This patch rides on an existing BPF_PROG_TYPE_CGROUP_SKB test
(test_sock_fields.c) to do a TCP end-to-end test on the new
bpf_sk_storage_* helpers.

Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agobpf: Add BPF_MAP_TYPE_SK_STORAGE test to test_maps
Martin KaFai Lau [Fri, 26 Apr 2019 23:39:52 +0000 (16:39 -0700)]
bpf: Add BPF_MAP_TYPE_SK_STORAGE test to test_maps

This patch adds BPF_MAP_TYPE_SK_STORAGE test to test_maps.
The src file is rather long, so it is put into another dir map_tests/
and compile like the current prog_tests/ does.  Other existing
tests in test_maps can also be re-factored into map_tests/ in the
future.

Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agobpf: Add verifier tests for the bpf_sk_storage
Martin KaFai Lau [Fri, 26 Apr 2019 23:39:49 +0000 (16:39 -0700)]
bpf: Add verifier tests for the bpf_sk_storage

This patch adds verifier tests for the bpf_sk_storage:
1. ARG_PTR_TO_MAP_VALUE_OR_NULL
2. Map and helper compatibility (e.g. disallow bpf_map_loookup_elem)

It also takes this chance to remove the unused struct btf_raw_data
and uses the BTF encoding macros from "test_btf.h".

Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agobpf: Refactor BTF encoding macro to test_btf.h
Martin KaFai Lau [Fri, 26 Apr 2019 23:39:46 +0000 (16:39 -0700)]
bpf: Refactor BTF encoding macro to test_btf.h

Refactor common BTF encoding macros for other tests to use.
The libbpf may reuse some of them in the future  which requires
some more thoughts before publishing as a libbpf API.

Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agobpf: Support BPF_MAP_TYPE_SK_STORAGE in bpf map probing
Martin KaFai Lau [Fri, 26 Apr 2019 23:39:44 +0000 (16:39 -0700)]
bpf: Support BPF_MAP_TYPE_SK_STORAGE in bpf map probing

This patch supports probing for the new BPF_MAP_TYPE_SK_STORAGE.
BPF_MAP_TYPE_SK_STORAGE enforces BTF usage, so the new probe
requires to create and load a BTF also.

Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agobpf: Sync bpf.h to tools
Martin KaFai Lau [Fri, 26 Apr 2019 23:39:42 +0000 (16:39 -0700)]
bpf: Sync bpf.h to tools

This patch sync the bpf.h to tools/.

Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agobpf: Introduce bpf sk local storage
Martin KaFai Lau [Fri, 26 Apr 2019 23:39:39 +0000 (16:39 -0700)]
bpf: Introduce bpf sk local storage

After allowing a bpf prog to
- directly read the skb->sk ptr
- get the fullsock bpf_sock by "bpf_sk_fullsock()"
- get the bpf_tcp_sock by "bpf_tcp_sock()"
- get the listener sock by "bpf_get_listener_sock()"
- avoid duplicating the fields of "(bpf_)sock" and "(bpf_)tcp_sock"
  into different bpf running context.

this patch is another effort to make bpf's network programming
more intuitive to do (together with memory and performance benefit).

When bpf prog needs to store data for a sk, the current practice is to
define a map with the usual 4-tuples (src/dst ip/port) as the key.
If multiple bpf progs require to store different sk data, multiple maps
have to be defined.  Hence, wasting memory to store the duplicated
keys (i.e. 4 tuples here) in each of the bpf map.
[ The smallest key could be the sk pointer itself which requires
  some enhancement in the verifier and it is a separate topic. ]

Also, the bpf prog needs to clean up the elem when sk is freed.
Otherwise, the bpf map will become full and un-usable quickly.
The sk-free tracking currently could be done during sk state
transition (e.g. BPF_SOCK_OPS_STATE_CB).

The size of the map needs to be predefined which then usually ended-up
with an over-provisioned map in production.  Even the map was re-sizable,
while the sk naturally come and go away already, this potential re-size
operation is arguably redundant if the data can be directly connected
to the sk itself instead of proxy-ing through a bpf map.

This patch introduces sk->sk_bpf_storage to provide local storage space
at sk for bpf prog to use.  The space will be allocated when the first bpf
prog has created data for this particular sk.

The design optimizes the bpf prog's lookup (and then optionally followed by
an inline update).  bpf_spin_lock should be used if the inline update needs
to be protected.

BPF_MAP_TYPE_SK_STORAGE:
-----------------------
To define a bpf "sk-local-storage", a BPF_MAP_TYPE_SK_STORAGE map (new in
this patch) needs to be created.  Multiple BPF_MAP_TYPE_SK_STORAGE maps can
be created to fit different bpf progs' needs.  The map enforces
BTF to allow printing the sk-local-storage during a system-wise
sk dump (e.g. "ss -ta") in the future.

The purpose of a BPF_MAP_TYPE_SK_STORAGE map is not for lookup/update/delete
a "sk-local-storage" data from a particular sk.
Think of the map as a meta-data (or "type") of a "sk-local-storage".  This
particular "type" of "sk-local-storage" data can then be stored in any sk.

The main purposes of this map are mostly:
1. Define the size of a "sk-local-storage" type.
2. Provide a similar syscall userspace API as the map (e.g. lookup/update,
   map-id, map-btf...etc.)
3. Keep track of all sk's storages of this "type" and clean them up
   when the map is freed.

sk->sk_bpf_storage:
------------------
The main lookup/update/delete is done on sk->sk_bpf_storage (which
is a "struct bpf_sk_storage").  When doing a lookup,
the "map" pointer is now used as the "key" to search on the
sk_storage->list.  The "map" pointer is actually serving
as the "type" of the "sk-local-storage" that is being
requested.

To allow very fast lookup, it should be as fast as looking up an
array at a stable-offset.  At the same time, it is not ideal to
set a hard limit on the number of sk-local-storage "type" that the
system can have.  Hence, this patch takes a cache approach.
The last search result from sk_storage->list is cached in
sk_storage->cache[] which is a stable sized array.  Each
"sk-local-storage" type has a stable offset to the cache[] array.
In the future, a map's flag could be introduced to do cache
opt-out/enforcement if it became necessary.

The cache size is 16 (i.e. 16 types of "sk-local-storage").
Programs can share map.  On the program side, having a few bpf_progs
running in the networking hotpath is already a lot.  The bpf_prog
should have already consolidated the existing sock-key-ed map usage
to minimize the map lookup penalty.  16 has enough runway to grow.

All sk-local-storage data will be removed from sk->sk_bpf_storage
during sk destruction.

bpf_sk_storage_get() and bpf_sk_storage_delete():
------------------------------------------------
Instead of using bpf_map_(lookup|update|delete)_elem(),
the bpf prog needs to use the new helper bpf_sk_storage_get() and
bpf_sk_storage_delete().  The verifier can then enforce the
ARG_PTR_TO_SOCKET argument.  The bpf_sk_storage_get() also allows to
"create" new elem if one does not exist in the sk.  It is done by
the new BPF_SK_STORAGE_GET_F_CREATE flag.  An optional value can also be
provided as the initial value during BPF_SK_STORAGE_GET_F_CREATE.
The BPF_MAP_TYPE_SK_STORAGE also supports bpf_spin_lock.  Together,
it has eliminated the potential use cases for an equivalent
bpf_map_update_elem() API (for bpf_prog) in this patch.

Misc notes:
----------
1. map_get_next_key is not supported.  From the userspace syscall
   perspective,  the map has the socket fd as the key while the map
   can be shared by pinned-file or map-id.

   Since btf is enforced, the existing "ss" could be enhanced to pretty
   print the local-storage.

   Supporting a kernel defined btf with 4 tuples as the return key could
   be explored later also.

2. The sk->sk_lock cannot be acquired.  Atomic operations is used instead.
   e.g. cmpxchg is done on the sk->sk_bpf_storage ptr.
   Please refer to the source code comments for the details in
   synchronization cases and considerations.

3. The mem is charged to the sk->sk_omem_alloc as the sk filter does.

Benchmark:
---------
Here is the benchmark data collected by turning on
the "kernel.bpf_stats_enabled" sysctl.
Two bpf progs are tested:

One bpf prog with the usual bpf hashmap (max_entries = 8192) with the
sk ptr as the key. (verifier is modified to support sk ptr as the key
That should have shortened the key lookup time.)

Another bpf prog is with the new BPF_MAP_TYPE_SK_STORAGE.

Both are storing a "u32 cnt", do a lookup on "egress_skb/cgroup" for
each egress skb and then bump the cnt.  netperf is used to drive
data with 4096 connected UDP sockets.

BPF_MAP_TYPE_HASH with a modifier verifier (152ns per bpf run)
27: cgroup_skb  name egress_sk_map  tag 74f56e832918070b run_time_ns 58280107540 run_cnt 381347633
    loaded_at 2019-04-15T13:46:39-0700  uid 0
    xlated 344B  jited 258B  memlock 4096B  map_ids 16
    btf_id 5

BPF_MAP_TYPE_SK_STORAGE in this patch (66ns per bpf run)
30: cgroup_skb  name egress_sk_stora  tag d4aa70984cc7bbf6 run_time_ns 25617093319 run_cnt 390989739
    loaded_at 2019-04-15T13:47:54-0700  uid 0
    xlated 168B  jited 156B  memlock 4096B  map_ids 17
    btf_id 6

Here is a high-level picture on how are the objects organized:

       sk
    ┌──────┐
    │      │
    │      │
    │      │
    │*sk_bpf_storage─────▶ bpf_sk_storage
    └──────┘                 ┌───────┐
                 ┌───────────┤ list  │
                 │           │       │
                 │           │       │
                 │           │       │
                 │           └───────┘
                 │
                 │     elem
                 │  ┌────────┐
                 ├─▶│ snode  │
                 │  ├────────┤
                 │  │  data  │          bpf_map
                 │  ├────────┤        ┌─────────┐
                 │  │map_node│◀─┬─────┤  list   │
                 │  └────────┘  │     │         │
                 │              │     │         │
                 │     elem     │     │         │
                 │  ┌────────┐  │     └─────────┘
                 └─▶│ snode  │  │
                    ├────────┤  │
   bpf_map          │  data  │  │
 ┌─────────┐        ├────────┤  │
 │  list   ├───────▶│map_node│  │
 │         │        └────────┘  │
 │         │                    │
 │         │           elem     │
 └─────────┘        ┌────────┐  │
                 ┌─▶│ snode  │  │
                 │  ├────────┤  │
                 │  │  data  │  │
                 │  ├────────┤  │
                 │  │map_node│◀─┘
                 │  └────────┘
                 │
                 │
                 │          ┌───────┐
     sk          └──────────│ list  │
  ┌──────┐                  │       │
  │      │                  │       │
  │      │                  │       │
  │      │                  └───────┘
  │*sk_bpf_storage───────▶bpf_sk_storage
  └──────┘

Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agoMerge branch 'writeable-bpf-tracepoints'
Alexei Starovoitov [Sat, 27 Apr 2019 02:05:44 +0000 (19:05 -0700)]
Merge branch 'writeable-bpf-tracepoints'

Matt Mullins says:

====================
This adds an opt-in interface for tracepoints to expose a writable context to
BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE programs that are attached, while
supporting read-only access from existing BPF_PROG_TYPE_RAW_TRACEPOINT
programs, as well as from non-BPF-based tracepoints.

The initial motivation is to support tracing that can be observed from the
remote end of an NBD socket, e.g. by adding flags to the struct nbd_request
header.  Earlier attempts included adding an NBD-specific tracepoint fd, but in
code review, I was recommended to implement it more generically -- as a result,
this patchset is far simpler than my initial try.

v4->v5:
  * rebased onto bpf-next/master and fixed merge conflicts
  * "tools: sync bpf.h" also syncs comments that have previously changed
    in bpf-next

v3->v4:
  * fixed a silly copy/paste typo in include/trace/events/bpf_test_run.h
    (_TRACE_NBD_H -> _TRACE_BPF_TEST_RUN_H)
  * fixed incorrect/misleading wording in patch 1's commit message,
    since the pointer cannot be directly dereferenced in a
    BPF_PROG_TYPE_RAW_TRACEPOINT
  * cleaned up the error message wording if the prog_tests fail
  * Addressed feedback from Yonghong
    * reject non-pointer-sized accesses to the buffer pointer
    * use sizeof(struct nbd_request) as one-byte-past-the-end in
      raw_tp_writable_reject_nbd_invalid.c
    * use BPF_MOV64_IMM instead of BPF_LD_IMM64

v2->v3:
  * Andrew addressed Josef's comments:
    * C-style commenting in nbd.c
    * Collapsed identical events into a single DECLARE_EVENT_CLASS.
      This saves about 2kB of kernel text

v1->v2:
  * add selftests
    * sync tools/include/uapi/linux/bpf.h
  * reject variable offset into the buffer
  * add string representation of PTR_TO_TP_BUFFER to reg_type_str
====================

Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agoselftests: bpf: test writable buffers in raw tps
Matt Mullins [Fri, 26 Apr 2019 18:49:51 +0000 (11:49 -0700)]
selftests: bpf: test writable buffers in raw tps

This tests that:
  * a BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE cannot be attached if it
    uses either:
    * a variable offset to the tracepoint buffer, or
    * an offset beyond the size of the tracepoint buffer
  * a tracer can modify the buffer provided when attached to a writable
    tracepoint in bpf_prog_test_run

Signed-off-by: Matt Mullins <mmullins@fb.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agotools: sync bpf.h
Matt Mullins [Fri, 26 Apr 2019 18:49:50 +0000 (11:49 -0700)]
tools: sync bpf.h

This adds BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, and fixes up the

error: enumeration value ‘BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE’ not handled in switch [-Werror=switch-enum]

build errors it would otherwise cause in libbpf.

Signed-off-by: Matt Mullins <mmullins@fb.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agonbd: add tracepoints for send/receive timing
Andrew Hall [Fri, 26 Apr 2019 18:49:49 +0000 (11:49 -0700)]
nbd: add tracepoints for send/receive timing

This adds four tracepoints to nbd, enabling separate tracing of payload
and header sending/receipt.

In the send path for headers that have already been sent, we also
explicitly initialize the handle so it can be referenced by the later
tracepoint.

Signed-off-by: Andrew Hall <hall@fb.com>
Signed-off-by: Matt Mullins <mmullins@fb.com>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agonbd: trace sending nbd requests
Matt Mullins [Fri, 26 Apr 2019 18:49:48 +0000 (11:49 -0700)]
nbd: trace sending nbd requests

This adds a tracepoint that can both observe the nbd request being sent
to the server, as well as modify that request , e.g., setting a flag in
the request that will cause the server to collect detailed tracing data.

The struct request * being handled is included to permit correlation
with the block tracepoints.

Signed-off-by: Matt Mullins <mmullins@fb.com>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agobpf: add writable context for raw tracepoints
Matt Mullins [Fri, 26 Apr 2019 18:49:47 +0000 (11:49 -0700)]
bpf: add writable context for raw tracepoints

This is an opt-in interface that allows a tracepoint to provide a safe
buffer that can be written from a BPF_PROG_TYPE_RAW_TRACEPOINT program.
The size of the buffer must be a compile-time constant, and is checked
before allowing a BPF program to attach to a tracepoint that uses this
feature.

The pointer to this buffer will be the first argument of tracepoints
that opt in; the pointer is valid and can be bpf_probe_read() by both
BPF_PROG_TYPE_RAW_TRACEPOINT and BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE
programs that attach to such a tracepoint, but the buffer to which it
points may only be written by the latter.

Signed-off-by: Matt Mullins <mmullins@fb.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agobpf, arm64: use more scalable stadd over ldxr / stxr loop in xadd
Daniel Borkmann [Fri, 26 Apr 2019 19:48:22 +0000 (21:48 +0200)]
bpf, arm64: use more scalable stadd over ldxr / stxr loop in xadd

Since ARMv8.1 supplement introduced LSE atomic instructions back in 2016,
lets add support for STADD and use that in favor of LDXR / STXR loop for
the XADD mapping if available. STADD is encoded as an alias for LDADD with
XZR as the destination register, therefore add LDADD to the instruction
encoder along with STADD as special case and use it in the JIT for CPUs
that advertise LSE atomics in CPUID register. If immediate offset in the
BPF XADD insn is 0, then use dst register directly instead of temporary
one.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jean-Philippe Brucker <jean-philippe.brucker@arm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agobpf, arm64: remove prefetch insn in xadd mapping
Daniel Borkmann [Fri, 26 Apr 2019 19:48:21 +0000 (21:48 +0200)]
bpf, arm64: remove prefetch insn in xadd mapping

Prefetch-with-intent-to-write is currently part of the XADD mapping in
the AArch64 JIT and follows the kernel's implementation of atomic_add.
This may interfere with other threads executing the LDXR/STXR loop,
leading to potential starvation and fairness issues. Drop the optional
prefetch instruction.

Fixes: 85f68fe89832 ("bpf, arm64: implement jiting of BPF_XADD")
Reported-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jean-Philippe Brucker <jean-philippe.brucker@arm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
5 years agoMerge tag 'mac80211-next-for-davem-2019-04-26' of git://git.kernel.org/pub/scm/linux...
David S. Miller [Fri, 26 Apr 2019 20:05:52 +0000 (16:05 -0400)]
Merge tag 'mac80211-next-for-davem-2019-04-26' of git://git./linux/kernel/git/jberg/mac80211-next

Johannes Berg says:

====================
Various updates, notably:
 * extended key ID support (from 802.11-2016)
 * per-STA TX power control support
 * mac80211 TX performance improvements
 * HE (802.11ax) updates
 * mesh link probing support
 * enhancements of multi-BSSID support (also related to HE)
 * OWE userspace processing support
====================

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