linux-toradex.git/net/dsa/master.c, branch v5.16

dev_ioctl: split out ndo_eth_ioctl

2021-07-27T19:11:45+00:00

Most users of ndo_do_ioctl are ethernet drivers that implement
the MII commands SIOCGMIIPHY/SIOCGMIIREG/SIOCSMIIREG, or hardware
timestamping with SIOCSHWTSTAMP/SIOCGHWTSTAMP.

Separate these from the few drivers that use ndo_do_ioctl to
implement SIOCBOND, SIOCBR and SIOCWANDEV commands.

This is a purely cosmetic change intended to help readers find
their way through the implementation.

Cc: Doug Ledford 
Cc: Jason Gunthorpe 
Cc: Jay Vosburgh 
Cc: Veaceslav Falico 
Cc: Andy Gospodarek 
Cc: Andrew Lunn 
Cc: Vivien Didelot 
Cc: Florian Fainelli 
Cc: Vladimir Oltean 
Cc: Leon Romanovsky 
Cc: linux-rdma@vger.kernel.org
Signed-off-by: Arnd Bergmann 
Acked-by: Jason Gunthorpe 
Signed-off-by: David S. Miller

net: dsa: generalize overhead for taggers that use both headers and trailers

2021-06-11T19:45:38+00:00

Some really really weird switches just couldn't decide whether to use a
normal or a tail tagger, so they just did both.

This creates problems for DSA, because we only have the concept of an
'overhead' which can be applied to the headroom or to the tailroom of
the skb (like for example during the central TX reallocation procedure),
depending on the value of bool tail_tag, but not to both.

We need to generalize DSA to cater for these odd switches by
transforming the 'overhead / tail_tag' pair into 'needed_headroom /
needed_tailroom'.

The DSA master's MTU is increased to account for both.

The flow dissector code is modified such that it only calls the DSA
adjustment callback if the tagger has a non-zero header length.

Taggers are trivially modified to declare either needed_headroom or
needed_tailroom, based on the tail_tag value that they currently
declare.

Signed-off-by: Vladimir Oltean 
Signed-off-by: David S. Miller

net: dsa: fix a crash if ->get_sset_count() fails

2021-05-10T21:31:27+00:00

If ds->ops->get_sset_count() fails then it "count" is a negative error
code such as -EOPNOTSUPP.  Because "i" is an unsigned int, the negative
error code is type promoted to a very high value and the loop will
corrupt memory until the system crashes.

Fix this by checking for error codes and changing the type of "i" to
just int.

Fixes: badf3ada60ab ("net: dsa: Provide CPU port statistics to master netdev")
Signed-off-by: Dan Carpenter 
Reviewed-by: Andrew Lunn 
Reviewed-by: Florian Fainelli 
Reviewed-by: Vladimir Oltean 
Signed-off-by: David S. Miller

net: dsa: allow changing the tag protocol via the "tagging" device attribute

2021-01-30T05:24:39+00:00

Currently DSA exposes the following sysfs:
$ cat /sys/class/net/eno2/dsa/tagging
ocelot

which is a read-only device attribute, introduced in the kernel as
commit 98cdb4807123 ("net: dsa: Expose tagging protocol to user-space"),
and used by libpcap since its commit 993db3800d7d ("Add support for DSA
link-layer types").

It would be nice if we could extend this device attribute by making it
writable:
$ echo ocelot-8021q > /sys/class/net/eno2/dsa/tagging

This is useful with DSA switches that can make use of more than one
tagging protocol. It may be useful in dsa_loop in the future too, to
perform offline testing of various taggers, or for changing between dsa
and edsa on Marvell switches, if that is desirable.

In terms of implementation, drivers can support this feature by
implementing .change_tag_protocol, which should always leave the switch
in a consistent state: either with the new protocol if things went well,
or with the old one if something failed. Teardown of the old protocol,
if necessary, must be handled by the driver.

Some things remain as before:
- The .get_tag_protocol is currently only called at probe time, to load
  the initial tagging protocol driver. Nonetheless, new drivers should
  report the tagging protocol in current use now.
- The driver should manage by itself the initial setup of tagging
  protocol, no later than the .setup() method, as well as destroying
  resources used by the last tagger in use, no earlier than the
  .teardown() method.

For multi-switch DSA trees, error handling is a bit more complicated,
since e.g. the 5th out of 7 switches may fail to change the tag
protocol. When that happens, a revert to the original tag protocol is
attempted, but that may fail too, leaving the tree in an inconsistent
state despite each individual switch implementing .change_tag_protocol
transactionally. Since the intersection between drivers that implement
.change_tag_protocol and drivers that support D in DSA is currently the
empty set, the possibility for this error to happen is ignored for now.

Testing:

$ insmod mscc_felix.ko
[   79.549784] mscc_felix 0000:00:00.5: Adding to iommu group 14
[   79.565712] mscc_felix 0000:00:00.5: Failed to register DSA switch: -517
$ insmod tag_ocelot.ko
$ rmmod mscc_felix.ko
$ insmod mscc_felix.ko
[   97.261724] libphy: VSC9959 internal MDIO bus: probed
[   97.267363] mscc_felix 0000:00:00.5: Found PCS at internal MDIO address 0
[   97.274998] mscc_felix 0000:00:00.5: Found PCS at internal MDIO address 1
[   97.282561] mscc_felix 0000:00:00.5: Found PCS at internal MDIO address 2
[   97.289700] mscc_felix 0000:00:00.5: Found PCS at internal MDIO address 3
[   97.599163] mscc_felix 0000:00:00.5 swp0 (uninitialized): PHY [0000:00:00.3:10] driver [Microsemi GE VSC8514 SyncE] (irq=POLL)
[   97.862034] mscc_felix 0000:00:00.5 swp1 (uninitialized): PHY [0000:00:00.3:11] driver [Microsemi GE VSC8514 SyncE] (irq=POLL)
[   97.950731] mscc_felix 0000:00:00.5 swp0: configuring for inband/qsgmii link mode
[   97.964278] 8021q: adding VLAN 0 to HW filter on device swp0
[   98.146161] mscc_felix 0000:00:00.5 swp2 (uninitialized): PHY [0000:00:00.3:12] driver [Microsemi GE VSC8514 SyncE] (irq=POLL)
[   98.238649] mscc_felix 0000:00:00.5 swp1: configuring for inband/qsgmii link mode
[   98.251845] 8021q: adding VLAN 0 to HW filter on device swp1
[   98.433916] mscc_felix 0000:00:00.5 swp3 (uninitialized): PHY [0000:00:00.3:13] driver [Microsemi GE VSC8514 SyncE] (irq=POLL)
[   98.485542] mscc_felix 0000:00:00.5: configuring for fixed/internal link mode
[   98.503584] mscc_felix 0000:00:00.5: Link is Up - 2.5Gbps/Full - flow control rx/tx
[   98.527948] device eno2 entered promiscuous mode
[   98.544755] DSA: tree 0 setup

$ ping 10.0.0.1
PING 10.0.0.1 (10.0.0.1): 56 data bytes
64 bytes from 10.0.0.1: seq=0 ttl=64 time=2.337 ms
64 bytes from 10.0.0.1: seq=1 ttl=64 time=0.754 ms
^C
 -  10.0.0.1 ping statistics  -
2 packets transmitted, 2 packets received, 0% packet loss
round-trip min/avg/max = 0.754/1.545/2.337 ms

$ cat /sys/class/net/eno2/dsa/tagging
ocelot
$ cat ./test_ocelot_8021q.sh
        #!/bin/bash

        ip link set swp0 down
        ip link set swp1 down
        ip link set swp2 down
        ip link set swp3 down
        ip link set swp5 down
        ip link set eno2 down
        echo ocelot-8021q > /sys/class/net/eno2/dsa/tagging
        ip link set eno2 up
        ip link set swp0 up
        ip link set swp1 up
        ip link set swp2 up
        ip link set swp3 up
        ip link set swp5 up
$ ./test_ocelot_8021q.sh
./test_ocelot_8021q.sh: line 9: echo: write error: Protocol not available
$ rmmod tag_ocelot.ko
rmmod: can't unload module 'tag_ocelot': Resource temporarily unavailable
$ insmod tag_ocelot_8021q.ko
$ ./test_ocelot_8021q.sh
$ cat /sys/class/net/eno2/dsa/tagging
ocelot-8021q
$ rmmod tag_ocelot.ko
$ rmmod tag_ocelot_8021q.ko
rmmod: can't unload module 'tag_ocelot_8021q': Resource temporarily unavailable
$ ping 10.0.0.1
PING 10.0.0.1 (10.0.0.1): 56 data bytes
64 bytes from 10.0.0.1: seq=0 ttl=64 time=0.953 ms
64 bytes from 10.0.0.1: seq=1 ttl=64 time=0.787 ms
64 bytes from 10.0.0.1: seq=2 ttl=64 time=0.771 ms
$ rmmod mscc_felix.ko
[  645.544426] mscc_felix 0000:00:00.5: Link is Down
[  645.838608] DSA: tree 0 torn down
$ rmmod tag_ocelot_8021q.ko

Signed-off-by: Vladimir Oltean 
Signed-off-by: Jakub Kicinski

net: dsa: unbind all switches from tree when DSA master unbinds

2021-01-13T02:48:40+00:00

Currently the following happens when a DSA master driver unbinds while
there are DSA switches attached to it:

$ echo 0000:00:00.5 > /sys/bus/pci/drivers/mscc_felix/unbind
------------[ cut here ]------------
WARNING: CPU: 0 PID: 392 at net/core/dev.c:9507
Call trace:
 rollback_registered_many+0x5fc/0x688
 unregister_netdevice_queue+0x98/0x120
 dsa_slave_destroy+0x4c/0x88
 dsa_port_teardown.part.16+0x78/0xb0
 dsa_tree_teardown_switches+0x58/0xc0
 dsa_unregister_switch+0x104/0x1b8
 felix_pci_remove+0x24/0x48
 pci_device_remove+0x48/0xf0
 device_release_driver_internal+0x118/0x1e8
 device_driver_detach+0x28/0x38
 unbind_store+0xd0/0x100

Located at the above location is this WARN_ON:

	/* Notifier chain MUST detach us all upper devices. */
	WARN_ON(netdev_has_any_upper_dev(dev));

Other stacked interfaces, like VLAN, do indeed listen for
NETDEV_UNREGISTER on the real_dev and also unregister themselves at that
time, which is clearly the behavior that rollback_registered_many
expects. But DSA interfaces are not VLAN. They have backing hardware
(platform devices, PCI devices, MDIO, SPI etc) which have a life cycle
of their own and we can't just trigger an unregister from the DSA
framework when we receive a netdev notifier that the master unregisters.

Luckily, there is something we can do, and that is to inform the driver
core that we have a runtime dependency to the DSA master interface's
device, and create a device link where that is the supplier and we are
the consumer. Having this device link will make the DSA switch unbind
before the DSA master unbinds, which is enough to avoid the WARN_ON from
rollback_registered_many.

Note that even before the blamed commit, DSA did nothing intelligent
when the master interface got unregistered either. See the discussion
here:
https://lore.kernel.org/netdev/20200505210253.20311-1-f.fainelli@gmail.com/
But this time, at least the WARN_ON is loud enough that the
upper_dev_link commit can be blamed.

The advantage with this approach vs dev_hold(master) in the attached
link is that the latter is not meant for long term reference counting.
With dev_hold, the only thing that will happen is that when the user
attempts an unbind of the DSA master, netdev_wait_allrefs will keep
waiting and waiting, due to DSA keeping the refcount forever. DSA would
not access freed memory corresponding to the master interface, but the
unbind would still result in a freeze. Whereas with device links,
graceful teardown is ensured. It even works with cascaded DSA trees.

$ echo 0000:00:00.2 > /sys/bus/pci/drivers/fsl_enetc/unbind
[ 1818.797546] device swp0 left promiscuous mode
[ 1819.301112] sja1105 spi2.0: Link is Down
[ 1819.307981] DSA: tree 1 torn down
[ 1819.312408] device eno2 left promiscuous mode
[ 1819.656803] mscc_felix 0000:00:00.5: Link is Down
[ 1819.667194] DSA: tree 0 torn down
[ 1819.711557] fsl_enetc 0000:00:00.2 eno2: Link is Down

This approach allows us to keep the DSA framework absolutely unchanged,
and the driver core will just know to unbind us first when the master
goes away - as opposed to the large (and probably impossible) rework
required if attempting to listen for NETDEV_UNREGISTER.

As per the documentation at Documentation/driver-api/device_link.rst,
specifying the DL_FLAG_AUTOREMOVE_CONSUMER flag causes the device link
to be automatically purged when the consumer fails to probe or later
unbinds. So we don't need to keep the consumer_link variable in struct
dsa_switch.

Fixes: 2f1e8ea726e9 ("net: dsa: link interfaces with the DSA master to get rid of lockdep warnings")
Signed-off-by: Vladimir Oltean 
Reviewed-by: Florian Fainelli 
Tested-by: Florian Fainelli 
Link: https://lore.kernel.org/r/20210111230943.3701806-1-olteanv@gmail.com
Signed-off-by: Jakub Kicinski

net: dsa: print the MTU value that could not be set

2020-12-08T19:24:07+00:00

These warnings become somewhat more informative when they include the
MTU value that could not be set and not just the errno.

Signed-off-by: Rasmus Villemoes 
Link: https://lore.kernel.org/r/20201205133944.10182-1-rasmus.villemoes@prevas.dk
Signed-off-by: Jakub Kicinski

net: dsa: allow drivers to request promiscuous mode on master

2020-09-26T21:17:58+00:00

Currently DSA assumes that taggers don't mess with the destination MAC
address of the frames on RX. That is not always the case. Some DSA
headers are placed before the Ethernet header (ocelot), and others
simply mangle random bytes from the destination MAC address (sja1105
with its incl_srcpt option).

Currently the DSA master goes to promiscuous mode automatically when the
slave devices go too (such as when enslaved to a bridge), but in
standalone mode this is a problem that needs to be dealt with.

So give drivers the possibility to signal that their tagging protocol
will get randomly dropped otherwise, and let DSA deal with fixing that.

Signed-off-by: Vladimir Oltean 
Reviewed-by: Andrew Lunn 
Signed-off-by: David S. Miller

net: dsa: stop overriding master's ndo_get_phys_port_name

2020-07-23T22:14:58+00:00

The purpose of this override is to give the user an indication of what
the number of the CPU port is (in DSA, the CPU port is a hardware
implementation detail and not a network interface capable of traffic).

However, it has always failed (by design) at providing this information
to the user in a reliable fashion.

Prior to commit 3369afba1e46 ("net: Call into DSA netdevice_ops
wrappers"), the behavior was to only override this callback if it was
not provided by the DSA master.

That was its first failure: if the DSA master itself was a DSA port or a
switchdev, then the user would not see the number of the CPU port in
/sys/class/net/eth0/phys_port_name, but the number of the DSA master
port within its respective physical switch.

But that was actually ok in a way. The commit mentioned above changed
that behavior, and now overrides the master's ndo_get_phys_port_name
unconditionally. That comes with problems of its own, which are worse in
a way.

The idea is that it's typical for switchdev users to have udev rules for
consistent interface naming. These are based, among other things, on
the phys_port_name attribute. If we let the DSA switch at the bottom
to start randomly overriding ndo_get_phys_port_name with its own CPU
port, we basically lose any predictability in interface naming, or even
uniqueness, for that matter.

So, there are reasons to let DSA override the master's callback (to
provide a consistent interface, a number which has a clear meaning and
must not be interpreted according to context), and there are reasons to
not let DSA override it (it breaks udev matching for the DSA master).

But, there is an alternative method for users to retrieve the number of
the CPU port of each DSA switch in the system:

  $ devlink port
  pci/0000:00:00.5/0: type eth netdev swp0 flavour physical port 0
  pci/0000:00:00.5/2: type eth netdev swp2 flavour physical port 2
  pci/0000:00:00.5/4: type notset flavour cpu port 4
  spi/spi2.0/0: type eth netdev sw0p0 flavour physical port 0
  spi/spi2.0/1: type eth netdev sw0p1 flavour physical port 1
  spi/spi2.0/2: type eth netdev sw0p2 flavour physical port 2
  spi/spi2.0/4: type notset flavour cpu port 4
  spi/spi2.1/0: type eth netdev sw1p0 flavour physical port 0
  spi/spi2.1/1: type eth netdev sw1p1 flavour physical port 1
  spi/spi2.1/2: type eth netdev sw1p2 flavour physical port 2
  spi/spi2.1/3: type eth netdev sw1p3 flavour physical port 3
  spi/spi2.1/4: type notset flavour cpu port 4

So remove this duplicated, unreliable and troublesome method. From this
patch on, the phys_port_name attribute of the DSA master will only
contain information about itself (if at all). If the users need reliable
information about the CPU port they're probably using devlink anyway.

Signed-off-by: Vladimir Oltean 
Acked-by: florian Fainelli 
Signed-off-by: David S. Miller

net: dsa: Setup dsa_netdev_ops

2020-07-20T23:48:22+00:00

Now that we have all the infrastructure in place for calling into the
dsa_ptr->netdev_ops function pointers, install them when we configure
the DSA CPU/management interface and tear them down. The flow is
unchanged from before, but now we preserve equality of tests when
network device drivers do tests like dev->netdev_ops == &foo_ops which
was not the case before since we were allocating an entirely new
structure.

Signed-off-by: Florian Fainelli 
Signed-off-by: David S. Miller

net: change addr_list_lock back to static key

2020-06-09T19:59:45+00:00

The dynamic key update for addr_list_lock still causes troubles,
for example the following race condition still exists:

CPU 0:				CPU 1:
(RCU read lock)			(RTNL lock)
dev_mc_seq_show()		netdev_update_lockdep_key()
				  -> lockdep_unregister_key()
 -> netif_addr_lock_bh()

because lockdep doesn't provide an API to update it atomically.
Therefore, we have to move it back to static keys and use subclass
for nest locking like before.

In commit 1a33e10e4a95 ("net: partially revert dynamic lockdep key
changes"), I already reverted most parts of commit ab92d68fc22f
("net: core: add generic lockdep keys").

This patch reverts the rest and also part of commit f3b0a18bb6cb
("net: remove unnecessary variables and callback"). After this
patch, addr_list_lock changes back to using static keys and
subclasses to satisfy lockdep. Thanks to dev->lower_level, we do
not have to change back to ->ndo_get_lock_subclass().

And hopefully this reduces some syzbot lockdep noises too.

Reported-by: syzbot+f3a0e80c34b3fc28ac5e@syzkaller.appspotmail.com
Cc: Taehee Yoo 
Cc: Dmitry Vyukov 
Signed-off-by: Cong Wang 
Signed-off-by: David S. Miller