can: rx-offload: can_rx_offload_add_manual(): add new initialization function

This patch adds a new initialization function:
can_rx_offload_add_manual()

It should be used to add support rx-offload to a driver, if the callback
mechanism should not be used. Use e.g. can_rx_offload_queue_sorted() to queue
skbs into rx-offload.

Link: https://lore.kernel.org/r/20200915223527.1417033-33-mkl@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: mscan: simplify clock enable/disable

All the NULL checks are pointless, clk_*() routines already deal with
NULL just fine.

Signed-off-by: Zhang Changzhong <zhangchangzhong@huawei.com>
Link: https://lore.kernel.org/r/1594972875-27631-1-git-send-email-zhangchangzhong@huawei.com
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: mscan: mpc5xxx_can: update contact email

The 'pengutronix' address is defunct for years. Use the proper contact
address.

Signed-off-by: Wolfram Sang <wsa@kernel.org>
Link: https://lore.kernel.org/r/20200502142657.19199-1-wsa@kernel.org
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: mcp251x: add support for half duplex controllers

Some SPI host controllers do not support full-duplex SPI and are
marked as such via the SPI_CONTROLLER_HALF_DUPLEX controller flag.

For such controllers use half duplex transactions but retain full
duplex transactions for the controllers that can handle those.

Signed-off-by: Tim Harvey <tharvey@gateworks.com>
Link: https://lore.kernel.org/r/1595516275-1179-1-git-send-email-tharvey@gateworks.com
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: mcp251x: Use readx_poll_timeout() helper

We may use special helper macro to poll IO till condition or timeout occurs.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Link: https://lore.kernel.org/r/20200217161038.25009-1-andriy.shevchenko@linux.intel.com
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: mcp251x: add GPIO support

The mcp251x variants feature 3 general purpose digital inputs and 2
outputs. With this patch they are accessible through the gpio framework.

Signed-off-by: Timo Schlüßler <schluessler@krause.de>
Tested-by: Timo Schlüßler <schluessler@krause.de>
Link: https://lore.kernel.org/r/20200915223527.1417033-28-mkl@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: mcp251x: sort include files alphabetically

This patch sorts the include files alphabetically.

Link: https://lore.kernel.org/r/20200915223527.1417033-27-mkl@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

dt-bindings: can: mcp251x: document GPIO support

The next patch adds gpio controller support to the mcp251x driver. This
patch updates the binding accordingly.

Cc: Timo Schlüßler <schluessler@krause.de>
Link: https://lore.kernel.org/r/20200915223527.1417033-26-mkl@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

dt-bindings: can: mcp251x: change example interrupt type to IRQ_TYPE_LEVEL_LOW

The MCP2515 datasheet clearly describes a level-triggered interrupt pin.
Change example bindings accordingly.

Link: https://lore.kernel.org/r/20200915223527.1417033-25-mkl@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: spi: Kconfig: remove unneeded dependencies form Kconfig symbols

Since commits

    653ee35ce6d5 ("can: hi311x: remove custom DMA mapped buffer")
    Fixes: df58525df395 ("can: mcp251x: remove custom DMA mapped buffer")

both the hi3111x and the mcp251x driver don't make use of the DMA API
any more. So we can safely remove the HAS_DMA dependency.

While we're here, remove the unneeded CAN_DEV and SPI dependencies from
the CAN_HI311X symbol, as the parent menus already have these
dependencies.

Link: https://lore.kernel.org/r/20200915223527.1417033-24-mkl@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: pcan_usb: add support of rxerr/txerr counters

This patch adds the support of the rx/tx errors CAN counters to the
driver of the PCAN-USB PC-CAN interface from PEAK-System GmbH.

The PCAN-USB is capable of giving back the values of the rx/tx errors
counters, to provide more details and statistics to the linux-can layer.
Getting these values allows the driver to better tune CAN_ERR_CRTL_TX_xxx
and CAN_ERR_CRTL_RX_xxx bits in case of the interface enters any
CAN_STATE_ERROR_xxx state.

Signed-off-by: Stephane Grosjean <s.grosjean@peak-system.com>
Link: https://lore.kernel.org/r/20191206153803.17725-3-s.grosjean@peak-system.com
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: pcan_usb: Document the commands sent to the device

This patch documents the functions and numbers of the commands sent by the
driver to the PCAN-USB CAN device from PEAK-System GmbH.

Signed-off-by: Stephane Grosjean <s.grosjean@peak-system.com>
Link: https://lore.kernel.org/r/20191206153803.17725-2-s.grosjean@peak-system.com
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: pch_can: use generic power management

Drivers using legacy power management .suspen()/.resume() callbacks have to
manage PCI states and device's PM states themselves. They also need to take
care of standard configuration registers.

Switch to generic power management framework using a single "struct dev_pm_ops"
variable to take the unnecessary load from the driver. This also avoids the
need for the driver to directly call most of the PCI helper functions and
device power state control functions, as through the generic framework PCI Core
takes care of the necessary operations, and drivers are required to do only
device-specific jobs.

Signed-off-by: Vaibhav Gupta <vaibhavgupta40@gmail.com>
Link: https://lore.kernel.org/r/20200728085757.888620-1-vaibhavgupta40@gmail.com
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: peak_canfd: Remove unused macros

CANFD_CLK_SEL_DIV_MASK and CANFD_OPTIONS_SET are
never used after they were introduced. Remove them.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wang Hai <wanghai38@huawei.com>
Link: https://lore.kernel.org/r/20200904131247.23021-1-wanghai38@huawei.com
[mkl: keep CANFD_CLK_SEL_DIV_MASK for documentation purpose]
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: peak_usb: convert to use le32_add_cpu()

Convert cpu_to_le32(le32_to_cpu(E1) + E2) to use le32_add_cpu().

Signed-off-by: Liu Shixin <liushixin2@huawei.com>
Link: https://lore.kernel.org/r/20200914041744.3701840-1-liushixin2@huawei.com
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: ti_hecc: convert to devm_platform_ioremap_resource_byname()

Use the function devm_platform_ioremap_resource_byname() to simplify
source code which calls the functions platform_get_resource_byname()
and devm_ioremap_resource(). Remove also a few error messages which
became unnecessary with this software refactoring.

Signed-off-by: Dejin Zheng <zhengdejin5@gmail.com>
Link: https://lore.kernel.org/r/20200420132207.8536-1-zhengdejin5@gmail.com
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: mscan: mark expected switch fall-through

As -Wimplicit-fallthrough is now enabled the mscan driver shows this warning,
which is fixed by this patch:

drivers/net/can/mscan/mscan.c: In function ‘mscan_start_xmit’:
drivers/net/can/mscan/mscan.c:211:3: warning: this statement may fall through [-Wimplicit-fallthrough=]
  211 |   netif_stop_queue(dev);
      |   ^~~~~~~~~~~~~~~~~~~~~
drivers/net/can/mscan/mscan.c:212:2: note: here
  212 |  case 2:
      |  ^~~~

Link: https://lore.kernel.org/r/20200915223527.1417033-17-mkl@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: mcba_usb: remove redundant initialization of variable err

The variable err is being initialized with a value that is never read
and it is being updated later with a new value.  The initialization is
redundant and can be removed.

Addresses-Coverity: ("Unused value")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Link: https://lore.kernel.org/r/20200706164240.518623-1-colin.king@canonical.com
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: c_can: Remove unused inline function

commit 524369e2391f ("can: c_can: remove obsolete STRICT_FRAME_ORDERING Kconfig option")
left behind this, remove it.

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Link: https://lore.kernel.org/r/20200505084149.23848-1-yuehaibing@huawei.com
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: dev: can_bus_off(): print scheduling of restart if activated

If a CAN device goes into bus-off and has automatic restart enabled, inform
user that a automatic restart is scheduled after the configured delay.

Link: https://lore.kernel.org/r/20200915223527.1417033-14-mkl@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: dev: can_change_state(): print human readable state change messages

In order to ease debugging let can_change_state() print the human readable
state change messages. Also print the old and new state.

Link: https://lore.kernel.org/r/20200915223527.1417033-13-mkl@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: dev: can_put_echo_skb(): propagate error in case of errors

The function can_put_echo_skb() can fail for several reasons. It may
fail due to OOM, but when it fails it's usually due to locking problems
in the driver.

In order to help developing and debugging of new drivers propagate error
value in case of errors.

Link: https://lore.kernel.org/r/20200915223527.1417033-12-mkl@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: dev: can_put_echo_skb(): print number of echo_skb that is occupied

This patch prints the number of the occupied echo_skb, to ease
implementing and debugging of new drivers.

Link: https://lore.kernel.org/r/20200915223527.1417033-11-mkl@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: remove "WITH Linux-syscall-note" from SPDX tag of C files

The "WITH Linux-syscall-note" exception is intended for UAPI headers.

See LICENSES/exceptions/Linux-syscall-note

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Link: https://lore.kernel.org/r/20200403073741.18352-1-masahiroy@kernel.org
Acked-by: Oliver Hartkopp <socketcan@hartkopp.net>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: softing: update dead link

BerliOS has not been operating for more than five years. linux-can moved to
GitHub.

Signed-off-by: Diego Elio Pettenò <flameeyes@flameeyes.com>
Link: https://lore.kernel.org/r/20200413170241.13207-1-flameeyes@flameeyes.com
[mkl: split into two patches - handle softing part here]
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: slcan: update dead link

BerliOS has not been operating for more than five years. linux-can moved to
GitHub.

Signed-off-by: Diego Elio Pettenò <flameeyes@flameeyes.com>
Link: https://lore.kernel.org/r/20200413170241.13207-1-flameeyes@flameeyes.com
[mkl: split into two patches - handle slcan part here]
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: raw: fix indention

This patch fixes the indention to follow kernel coding style.

Link: https://lore.kernel.org/r/20200915223527.1417033-7-mkl@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: drivers: fix spelling mistakes

This patch fixes spelling erros found by "codespell" in the
drivers/net/can subtree.

Link: https://lore.kernel.org/r/20200915223527.1417033-6-mkl@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: net: fix spelling mistakes

This patch fixes spelling erros found by "codespell" in the net/can
subtree.

Link: https://lore.kernel.org/r/20200915223527.1417033-5-mkl@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: include: fix spelling mistakes

This patch fixes spelling erros found by "codespell" in the
include/linux/can subtree.

Link: https://lore.kernel.org/r/20200915223527.1417033-4-mkl@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: flexcan: fix spelling mistake "reserverd" -> "reserved"

Fix a mistake in a register layout description.

Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Link: https://lore.kernel.org/r/20200214141751.21168-1-alexandre.belloni@bootlin.com
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

can: grcan: fix spelling mistake "buss" -> "bus"

There is a spelling mistake in a netdev_err error message. Fix it.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Link: https://lore.kernel.org/r/20200806105616.46790-1-colin.king@canonical.com
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

RDS: drop double zeroing

sg_init_table zeroes its first argument, so the allocation of that argument
doesn't have to.

the semantic patch that makes this change is as follows:
(http://coccinelle.lip6.fr/)

// <smpl>
@@
expression x,n,flags;
@@

x =
- kcalloc
+ kmalloc_array
  (n,sizeof(*x),flags)
...
sg_init_table(x,n)
// </smpl>

Signed-off-by: Julia Lawall <Julia.Lawall@inria.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>

dpaa2-eth: drop double zeroing

sg_init_table zeroes its first argument, so the allocation of that argument
doesn't have to.

the semantic patch that makes this change is as follows:
(http://coccinelle.lip6.fr/)

// <smpl>
@@
expression x,n,flags;
@@

x =
- kcalloc
+ kmalloc_array
  (n,sizeof(struct scatterlist),flags)
...
sg_init_table(x,n)
// </smpl>

Signed-off-by: Julia Lawall <Julia.Lawall@inria.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'DSA-with-VLAN-filtering-and-offloading-masters'

Vladimir Oltean says:

====================
DSA with VLAN filtering and offloading masters

This series attempts to make DSA VLANs work in the presence of a master
interface that is:
- filtering, so it drops VLANs that aren't explicitly added to its
  filter list
- offloading, so the old assumptions in the tagging code about there
  being a VLAN tag in the skb are not necessarily true anymore.

For more context:
https://lore.kernel.org/netdev/20200910150738.mwhh2i6j2qgacqev@skbuf/

This probably marks the beginning of a series of patches in which DSA
starts paying much more attention to its upper interfaces, not only for
VLAN purposes but also for address filtering and for management of the
CPU flooding domain. There was a comment from Florian on whether we
could factor some of the mlxsw logic into some common functionality, but
it doesn't look so. This seems bound to be open-coded, but frankly there
isn't a lot to it.

Changes in v2:
Applied Florian's cosmetic suggestion in patch 4/9.
====================

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

net: dsa: tag_sja1105: add compatibility with hwaccel VLAN tags

Check whether there is any hwaccel VLAN tag on RX, and if there is,
treat it as the tag_8021q header.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: tag_8021q: add VLANs to the master interface too

The whole purpose of tag_8021q is to send VLAN-tagged traffic to the
CPU, from which the driver can decode the source port and switch id.

Currently this only works if the VLAN filtering on the master is
disabled. Change that by explicitly adding code to tag_8021q.c to add
the VLANs corresponding to the tags to the filter of the master
interface.

Because we now need to call vlan_vid_add, then we also need to hold the
RTNL mutex. Propagate that requirement to the callers of dsa_8021q_setup
and modify the existing call sites as appropriate. Note that one call
path, sja1105_best_effort_vlan_filtering_set -> sja1105_vlan_filtering
-> sja1105_setup_8021q_tagging, was already holding this lock.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: install VLANs into the master's RX filter too

Most DSA switch tags shift the EtherType to the right, causing the
master to not parse the VLAN as VLAN.
However, not all switches do that (example: tail tags, tag_8021q etc),
and if the DSA master has "rx-vlan-filter: on" in ethtool -k, then we
have a problem.

Therefore, we could populate the VLAN table of the master, just in case
(for some switches it will not make a difference), so that network I/O
can work even with a VLAN filtering master.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: allow 8021q uppers while the bridge has vlan_filtering=0

When the bridge has VLAN awareness disabled there isn't any duplication
of functionality, since the bridge does not process VLAN. Don't deny
adding 8021q uppers to DSA switch ports in that case. The switch is
supposed to simply pass traffic leaving the VLAN tag as-is, and the
stack will happily strip the VLAN tag for all 8021q uppers that exist.

We need to ensure that there are no 8021q uppers when the user attempts
to enable bridge vlan_filtering.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: refuse configuration in prepare phase of dsa_port_vlan_filtering()

The current logic beats me a little bit. The comment that "bridge skips
-EOPNOTSUPP, so skip the prepare phase" was introduced in commit
fb2dabad69f0 ("net: dsa: support VLAN filtering switchdev attr").

I'm not sure:
(a) ok, the bridge skips -EOPNOTSUPP, but, so what, where are we
    returning -EOPNOTSUPP?
(b) even if we are, and I'm just not seeing it, what is the causality
    relationship between the bridge skipping -EOPNOTSUPP and DSA
    skipping the prepare phase, and just returning zero?

One thing is certain beyond doubt though, and that is that DSA currently
refuses VLAN filtering from the "commit" phase instead of "prepare", and
that this is not a good thing:

ip link add br0 type bridge
ip link add br1 type bridge vlan_filtering 1
ip link set swp2 master br0
ip link set swp3 master br1
[ 3790.379389] 001: sja1105 spi0.1: VLAN filtering is a global setting
[ 3790.379399] 001: ------------[ cut here ]------------
[ 3790.379403] 001: WARNING: CPU: 1 PID: 515 at net/switchdev/switchdev.c:157 switchdev_port_attr_set_now+0x9c/0xa4
[ 3790.379420] 001: swp3: Commit of attribute (id=6) failed.
[ 3790.379533] 001: [<c11ff588>] (switchdev_port_attr_set_now) from [<c11b62e4>] (nbp_vlan_init+0x84/0x148)
[ 3790.379544] 001: [<c11b62e4>] (nbp_vlan_init) from [<c11a2ff0>] (br_add_if+0x514/0x670)
[ 3790.379554] 001: [<c11a2ff0>] (br_add_if) from [<c1031b5c>] (do_setlink+0x38c/0xab0)
[ 3790.379565] 001: [<c1031b5c>] (do_setlink) from [<c1036fe8>] (__rtnl_newlink+0x44c/0x748)
[ 3790.379573] 001: [<c1036fe8>] (__rtnl_newlink) from [<c1037328>] (rtnl_newlink+0x44/0x60)
[ 3790.379580] 001: [<c1037328>] (rtnl_newlink) from [<c10315fc>] (rtnetlink_rcv_msg+0x124/0x2f8)
[ 3790.379590] 001: [<c10315fc>] (rtnetlink_rcv_msg) from [<c10926b8>] (netlink_rcv_skb+0xb8/0x110)
[ 3790.379806] 001: ---[ end trace 0000000000000002 ]---
[ 3790.379819] 001: sja1105 spi0.1 swp3: failed to initialize vlan filtering on this port

So move the current logic that may fail (except ds->ops->port_vlan_filtering,
that is way harder) into the prepare stage of the switchdev transaction.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: convert denying bridge VLAN with existing 8021q upper to PRECHANGEUPPER

This is checking for the following order of operations, and makes sure
to deny that configuration:

ip link add link swp2 name swp2.100 type vlan id 100
ip link add br0 type bridge vlan_filtering 1
ip link set swp2 master br0
bridge vlan add dev swp2 vid 100

Instead of using vlan_for_each(), which looks at the VLAN filters
installed with vlan_vid_add(), just track the 8021q uppers. This has the
advantage of freeing up the vlan_vid_add() call for actual VLAN
filtering.

There is another change in this patch. The check is moved in slave.c,
from switch.c. I don't think it makes sense to have this 8021q upper
check for each switch port that gets notified of that VLAN addition
(these include DSA links and CPU ports, we know those can't have 8021q
uppers because they don't have a net_device registered for them), so
just do it in slave.c, for that one slave interface.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: convert check for 802.1Q upper when bridged into PRECHANGEUPPER

DSA tries to prevent having a VLAN added by a bridge and by an 802.1Q
upper at the same time. It does that by checking the VID in
.ndo_vlan_rx_add_vid(), since that's something that the 8021q module
calls, via vlan_vid_add(). When a VLAN matches in both subsystems, this
check returns -EBUSY.

However the vlan_vid_add() function isn't specific to the 8021q module
in any way at all. It is simply the kernel's way to tell an interface to
add a VLAN to its RX filter and not drop that VLAN. So there's no reason
to return -EBUSY when somebody tries to call vlan_vid_add() for a VLAN
that was installed by the bridge. The proper behavior is to accept that
configuration.

So what's wrong is how DSA checks that it has an 8021q upper. It should
look at the actual uppers for that, not just assume that the 8021q
module was somewhere in the call stack of .ndo_vlan_rx_add_vid().

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: rename dsa_slave_upper_vlan_check to something more suggestive

We'll be adding a new check in the PRECHANGEUPPER notifier, where we'll
need to check some VLAN uppers. It is hard to do that when there is
already a function named dsa_slave_upper_vlan_check. So rename this one.

Not to mention that this function probably shouldn't have started with
"dsa_slave_" in the first place, since the struct net_device argument
isn't a DSA slave, but an 8021q upper of one.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: deny enslaving 802.1Q upper to VLAN-aware bridge from PRECHANGEUPPER

There doesn't seem to be any strong technical reason for doing it this
way, but we'll be adding more checks for invalid upper device
configurations, and it will be easier to have them all grouped under
PRECHANGEUPPER.

Tested that it still works:
ip link set br0 type bridge vlan_filtering 1
ip link add link swp2 name swp2.100 type vlan id 100
ip link set swp2.100 master br0
[   20.321312] br0: port 5(swp2.100) entered blocking state
[   20.326711] br0: port 5(swp2.100) entered disabled state
Error: dsa_core: Cannot enslave VLAN device into VLAN aware bridge.
[   20.346549] br0: port 5(swp2.100) entered blocking state
[   20.351957] br0: port 5(swp2.100) entered disabled state

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: mventa: remove unused variable 'dummy' in mvneta_mib_counters_clear()

Fixes the following W=1 kernel build warning(s):

drivers/net/ethernet/marvell/mvneta.c:754:6: warning:
 variable 'dummy' set but not used [-Wunused-but-set-variable]
  754 |  u32 dummy;
      |      ^~~~~

This variable is not used in function mvneta_mib_counters_clear(), so
remove it to avoid build warning.

Signed-off-by: Zhang Changzhong <zhangchangzhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: remove unnecessary NULL checking in napi_consume_skb()

When budget is non-zero, skb_unref() has already handled the
NULL checking.

When budget is zero, the dev_consume_skb_any() has handled NULL
checking in __dev_kfree_skb_irq(), or dev_kfree_skb() which also
ultimately call skb_unref().

So remove the unnecessary checking in napi_consume_skb().

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

net: mvneta: avoid copying shared_info frags in mvneta_swbm_build_skb

Avoid copying skb_shared_info frags array in mvneta_swbm_build_skb() since
__build_skb_around() does not overwrite it

Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: rtl8366: Skip PVID setting if not requested

We go to lengths to determine whether the PVID should be set
for this port or not, and then fail to take it into account.
Fix this oversight.

Fixes: d8652956cf37 ("net: dsa: realtek-smi: Add Realtek SMI driver")
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch '100base-Fx-link-modes'

Dan Murphy says:

====================
100base Fx link modes

As per patch https://lore.kernel.org/patchwork/patch/1300241/ the link
modes for 100base FX full and half duplex modes did not exist.  Adding
these link modes to the core and ethtool allow devices like the
DP83822, DP83869 and Broadcomm PHYs to properly advertise the correct
mode for Fiber 100Mbps.

Corresponding user land ethtool patches are available but rely on
these patches to be applied first.
====================

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

net: phy: dp83822: Update the fiber advertisement for speed

Update the fiber advertisement for speed and duplex modes with the
100base-FX full and half linkmode entries.

Signed-off-by: Dan Murphy <dmurphy@ti.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>

net: dp83869: Add ability to advertise Fiber connection

Add the ability to advertise the Fiber connection if the strap or the
op-mode is configured for 100Base-FX.

Auto negotiation is not supported on this PHY when in fiber mode.

Signed-off-by: Dan Murphy <dmurphy@ti.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>

ethtool: Add 100base-FX link mode entries

Add entries for the 100base-FX full and half duplex supported modes.

$ ethtool eth0
        Supported ports: [ FIBRE ]
        Supported link modes:  100baseFX/Half 100baseFX/Full
        Supported pause frame use: Symmetric Receive-only
        Supports auto-negotiation: No
        Supported FEC modes: Not reported
        Advertised link modes: 100baseFX/Half 100baseFX/Full
        Advertised pause frame use: No
        Advertised auto-negotiation: No
        Advertised FEC modes: Not reported
        Speed: 100Mb/s
        Duplex: Full
        Auto-negotiation: off
        Port: MII
        PHYAD: 1
        Transceiver: external
        Supports Wake-on: gs
        Wake-on: d
        SecureOn password: 00:00:00:00:00:00
        Current message level: 0x00000000 (0)

        Link detected: yes

Signed-off-by: Dan Murphy <dmurphy@ti.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>

net/packet: Fix a comment about network_header

skb->nh.raw has been renamed as skb->network_header in 2007, in
commit b0e380b1d8a8 ("[SK_BUFF]: unions of just one member don't get
                      anything done, kill them")

So here we change it to the new name.

Cc: Willem de Bruijn <willemdebruijn.kernel@gmail.com>
Signed-off-by: Xie He <xie.he.0141@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: ethernet: ti: cpsw: use true,false for bool variables

This addresses the following coccinelle warning:

drivers/net/ethernet/ti/cpsw.c:1599:2-17: WARNING: Assignment of 0/1 to
bool variable
drivers/net/ethernet/ti/cpsw.c:1300:2-17: WARNING: Assignment of 0/1 to
bool variable

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Jason Yan <yanaijie@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

8139too: use true,false for bool variables

This addresses the following coccinelle warning:

drivers/net/ethernet/realtek/8139too.c:981:2-8: WARNING: Assignment of
0/1 to bool variable

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Jason Yan <yanaijie@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: use true,false for bool variables

This addresses the following coccinelle warning:

drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c:15415:1-26: WARNING:
Assignment of 0/1 to bool variable
drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c:12393:2-17: WARNING:
Assignment of 0/1 to bool variable
drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c:15497:2-27: WARNING:
Assignment of 0/1 to bool variable

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Jason Yan <yanaijie@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: qed: use true,false for bool variables

This addresses the following coccinelle warning:

drivers/net/ethernet/qlogic/qed/qed_rdma.c:1465:2-13: WARNING:
Assignment of 0/1 to bool variable
drivers/net/ethernet/qlogic/qed/qed_rdma.c:1468:2-14: WARNING:
Assignment of 0/1 to bool variable
drivers/net/ethernet/qlogic/qed/qed_rdma.c:1471:2-13: WARNING:
Assignment of 0/1 to bool variable
drivers/net/ethernet/qlogic/qed/qed_rdma.c:1472:2-14: WARNING:
Assignment of 0/1 to bool variable

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Jason Yan <yanaijie@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: b44: use true,false for bool variables

This addresses the following coccinelle warning:

drivers/net/ethernet/broadcom/b44.c:2213:6-20: WARNING: Assignment of
0/1 to bool variable
drivers/net/ethernet/broadcom/b44.c:2218:2-16: WARNING: Assignment of
0/1 to bool variable
drivers/net/ethernet/broadcom/b44.c:2226:3-17: WARNING: Assignment of
0/1 to bool variable
drivers/net/ethernet/broadcom/b44.c:2230:3-17: WARNING: Assignment of
0/1 to bool variable

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Jason Yan <yanaijie@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: micrel: Remove set but not used variable

Fixes gcc '-Wunused-but-set-variable' warning:

drivers/net/ethernet/micrel/ksz884x.c: In function rx_proc:
drivers/net/ethernet/micrel/ksz884x.c:4981:6: warning: variable ‘rx_status’ set but not used [-Wunused-but-set-variable]

drivers/net/ethernet/micrel/ksz884x.c: In function netdev_get_ethtool_stats:
drivers/net/ethernet/micrel/ksz884x.c:6512:6: warning: variable ‘rc’ set but not used [-Wunused-but-set-variable]

these variable is never used, so remove it.

Signed-off-by: Zheng Yongjun <zhengyongjun3@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: e1000: Remove set but not used variable

Fixes gcc '-Wunused-but-set-variable' warning:

drivers/net/ethernet/intel/e1000/e1000_hw.c: In function e1000_phy_init_script:
drivers/net/ethernet/intel/e1000/e1000_hw.c:132:6: warning: variable ‘ret_val’ set but not used [-Wunused-but-set-variable]

`ret_val` is never used, so remove it.

Signed-off-by: Zheng Yongjun <zhengyongjun3@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: liquidio: Remove set but not used variable

Fixes gcc '-Wunused-but-set-variable' warning:

drivers/net/ethernet/cavium/liquidio/octeon_device.c: In function lio_pci_readq:
drivers/net/ethernet/cavium/liquidio/octeon_device.c:1327:6: warning: variable ‘val32’ set but not used [-Wunused-but-set-variable]

drivers/net/ethernet/cavium/liquidio/octeon_device.c: In function lio_pci_writeq:
drivers/net/ethernet/cavium/liquidio/octeon_device.c:1358:6: warning: variable ‘val32’ set but not used [-Wunused-but-set-variable]

these variable is never used, so remove it.

Signed-off-by: Zheng Yongjun <zhengyongjun3@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'mv88e6xxx-Add-devlink-regions-support'

Andrew Lunn says:

====================
mv88e6xxx: Add devlink regions support

Make use of devlink regions to allow read access to some of the
internal of the switches. Currently access to global1, global2 and the
ATU is provided.

The switch itself will never trigger a region snapshot, it is assumed
it is performed from user space as needed.

v2:
Remove left of debug print
Comment ATU format is generic to mv88e6xxx
Combine declaration and the assignment on a single line.

v3:
Drop support for port regions
Improve the devlink API with a priv member and passing the region to
the snapshot function
Make the helper to convert from devlink to ds an inline function

v4:
Add missing kerneldoc in ICE driver
Fix typo for global2 reading global1 registers
Make use of enum to make code more readable
====================

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

net: dsa: mv88e6xxx: Implement devlink info get callback

Return the driver name and the asic.id with the switch name.

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

net: dsa: wire up devlink info get

Allow the DSA drivers to implement the devlink call to get info info,
e.g. driver name, firmware version, ASIC ID, etc.

v2:
Combine declaration and the assignment on a single line.

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

net: dsa: mv88e6xxx: Add devlink regions

Allow the global registers, and the ATU to be snapshot via devlink
regions. It is later planned to add support for the port registers.

v2:
Remove left over debug prints
Comment ATU format is generic for mv88e6xxx, not wider

v3:
Make use of ops structure passed to snapshot function
Remove port regions

v4:
Make use of enum mv88e6xxx_region_id
Fix global2/global1 read typ0

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>

net: dsa: mv88e6xxx: Create helper for FIDs in use

Refactor the code in mv88e6xxx_atu_new() which builds a bitmaps of
FIDs in use into a helper function. This will be reused by the devlink
code when dumping the ATU.

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

net: dsa: mv88e6xxx: Move devlink code into its own file

There will soon be more devlink code. Move the existing code into a
file of its own, before we start adding this new code.

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

net: dsa: Add devlink regions support to DSA

Allow DSA drivers to make use of devlink regions, via simple wrappers.

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

net: dsa: Add helper to convert from devlink to ds

Given a devlink instance, return the dsa switch it is associated to.

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

net: devlink: region: Pass the region ops to the snapshot function

Pass the region to be snapshotted to the function performing the
snapshot. This allows one function to operate on numerous regions.

v4:
Add missing kerneldoc for ICE

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>

net: devlink: regions: Add a priv member to the regions ops struct

The driver may have multiple regions which can be dumped using one
function. However, for this to work, additional information is
needed. Add a priv member to the ops structure for the driver to use
however it likes.

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

Merge branch 'Felix-DSA-driver-cleanup-build-Seville-separately'

Vladimir Oltean says:

====================
Felix DSA driver cleanup: build Seville separately

When introducing the Seville switch support to the Felix driver, some
technical debt was created. Since both VSC9959 and VSC9953 are embedded
switches (one on an arm64 SoC and the other on a powerpc SoC), there is
no use case for having the code for both be present in the same module.

This was necessary at the time due to the common SERDES PCS code that
they were using, but that has been since refactored into
drivers/net/pcs/pcs-lynx.c.

This makes the Seville driver stop uselessly depending upon PCI and
FSL_ENETC_MDIO, which were only dependencies of Felix in fact.

Some whitespace/tab conversions are also present in this series as part
of the cleanup process.
====================

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

net: dsa: seville: build as separate module

Seville does not need to depend on PCI or on the ENETC MDIO controller.
There will also be other compile-time differences in the future.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: felix: move the PTP clock structure to felix_vsc9959.c

Not only does Sevile not have a PTP clock, but with separate modules,
this structure cannot even live in felix.c, due to the .owner =
THIS_MODULE assignment causing this link time error:

drivers/net/dsa/ocelot/felix.o:(.data+0x0): undefined reference to `__this_module'

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: mscc: ocelot: make ocelot_init_timestamp take a const struct ptp_clock_info

It is a good measure to ensure correctness if the structures that are
meant to remain constant are only processed by functions that thake
constant arguments.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: seville: duplicate vsc9959_mdio_bus_free

While we don't plan on making any changes to this function, currently
this is the only remaining dependency between felix and seville, after
the PCS has been refactored out into pcs-lynx.c.

Duplicate this function in seville to break the dependency completely.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: felix: replace tabs with spaces

Over the time, some patches have introduced structures aligned with
spaces, near structures aligned with tabs. Fix the inconsistencies.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: seville: reindent defines for MDIO controller

Reindent these definitions to be in line with the rest of the driver.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: seville: remove unused defines for the mdio controller

Some definitions were likely copied from
drivers/net/mdio/mdio-mscc-miim.c.

They are not necessary, remove them.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: ocelot: document why reset procedure is different for felix/seville

The overall idea (issue soft reset, enable memories, initialize
memories, enable core) is the same, so it would make sense that an
attempt is made to unify the procedures.

It is not immediately obvious that the fields are not part of the same
register targets, though. So add a comment.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: seville: first enable memories, then initialize them

As per documentation, proper startup sequence is:
* Enable memories
* Initialize memories
* Enable core

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: seville: don't write to MEM_ENA twice

There is another one of these right above the readx_poll_status.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: felix: use ocelot_field_{read,write} helpers consistently

Since these helpers for regmap fields are available, use them.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'ptp_qoriq-support-FIPER3'

Yangbo Lu says:

====================
ptp_qoriq: support FIPER3

The FIPER3 (fixed interval period pulse generator) is supported on
DPAA2 and ENETC network controller hardware. This patch-set is to
support it in ptp_qoriq driver and dt-binding.

Changes for v2:
- Some improvement in code.
- Added ACK from Vladimir.
====================

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

ptp_qoriq: support FIPER3

The FIPER3 (fixed interval period pulse generator) is supported on
DPAA2 and ENETC network controller hardware. This patch is to support
it in ptp_qoriq driver.

Signed-off-by: Yangbo Lu <yangbo.lu@nxp.com>
Acked-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

dt-binding: ptp_qoriq: support fsl,tmr-fiper3 property

Add fsl,tmr-fiper3 property definition which is supported only
on DPAA2 and ENETC network controller hardware.

Signed-off-by: Yangbo Lu <yangbo.lu@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'net-ipa-wake-up-system-on-RX-available'

Alex Elder says:

====================
net: ipa: wake up system on RX available

This series arranges for the IPA driver to wake up a suspended
system if the IPA hardware has a packet to deliver to the AP.

Version 2 replaced the first patch from version 1 with three
patches, in response to David Miller's feedback.  And based on
Bjorn Andersson's feedback on version 2, this version reworks
the tracking of IPA clock references.  As a result, we no
longer need a flag to determine whether a "don't' suspend" clock
reference is held (though an bit in a bitmask is still used for
a different purpose).

In summary:
    - A refcount_t is used to track IPA clock references where an
      atomic_t was previously used.  (This may go away soon as well,
      with upcoming work to implement runtime PM.)
    - We no longer track whether a special reference has been taken
      to avoid suspending IPA.
    - A bit in a bitmask is used to ensure we only trigger a system
      resume once per system suspend.
And from the original series:
    - Suspending endpoints only occurs when suspending the driver,
      not when dropping the last clock reference.  Resuming
      endpoints is also disconnected from starting the clock.
    - The IPA SUSPEND interrupt is now a wakeup interrupt.  If it
      fires, it schedules a system resume operation.
    - The GSI interrupt is no longer a wakeup interrupt.
====================

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

net: ipa: do not enable GSI interrupt for wakeup

We now trigger a system resume when we receive an IPA SUSPEND
interrupt.  We should *not* wake up on GSI interrupts.

Signed-off-by: Alex Elder <elder@linaro.org>
Reviewed-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: ipa: enable wakeup on IPA interrupt

Now that we handle wakeup interrupts properly, arrange for the IPA
interrupt to be treated as a wakeup interrupt.

Signed-off-by: Alex Elder <elder@linaro.org>
Reviewed-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: ipa: repurpose CLOCK_HELD flag

The previous patch causes a system resume to be triggered when a
packet is available for receipt on a suspended RX endpoint.

The CLOCK_HELD flag was previously used to indicate that an extra
clock reference was held, preventing suspend.  But we no longer need
such a flag:
  - We take an initial reference in ipa_config().
  - That reference is held until ipa_suspend() releases it.
  - A subsequent system resume leads to a reference getting
    re-acquired in ipa_resume().
  - This can repeat until ultimately the module is removed, where
    ipa_remove() releases the reference.
We no longer need a special flag to determine whether this extra
reference is held--it is, provided probe has completed successfully
and the driver is not suspended (or removed).

On the other hand, once suspended, it's possible for more than one
endpoint to trip the IPA SUSPEND interrupt, and we only want to
trigger the system resume once.  So repurpose the Boolean CLOCK_HELD
flag to record whether the IPA SUSPEND handler should initiate a
system resume.

The flag will be be cleared each time ipa_suspend() is called,
*before* any endpoints are suspended.  And it will be set inside the
IPA SUSPEND interrupt handler exactly once per suspend.

Rename the flag IPA_FLAG_RESUMED to reflect its new purpose.

Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: ipa: use device_init_wakeup()

The call to wakeup_source_register() in ipa_probe() does not do what
it was intended to do.  Call device_init_wakeup() in ipa_setup()
instead, to set the IPA device as wakeup-capable and to initially
enable wakeup capability.

When we receive a SUSPEND interrupt, call pm_wakeup_dev_event()
with a zero processing time, to simply call for a resume without
any other processing.  The ipa_resume() call will take care of
waking things up again, and will handle receiving the packet.

Note that this gets rid of a clock reference counting bug that
occurred when handling an IPA SUSPEND interrupt.  Specifically,
ipa_suspend_handler() took an IPA clock reference *in addition*
to the one taken by ipa_resume().  There is no need to back-port
this fix however, because it only affects code that was not
previously working (this patch is part of fixing that).

Signed-off-by: Alex Elder <elder@linaro.org>
Reviewed-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: ipa: manage endpoints separate from clock

Currently, when (before) the last IPA clock reference is dropped,
all endpoints are suspended.  And whenever the first IPA clock
reference is taken, all endpoints are resumed (or started).

In most cases there's no need to start endpoints when the clock
starts.  So move the calls to ipa_endpoint_suspend() and
ipa_endpoint_resume() out of ipa_clock_put() and ipa_clock_get(),
respectiely.  Instead, only suspend endpoints when handling a system
suspend, and only resume endpoints when handling a system resume.

Signed-off-by: Alex Elder <elder@linaro.org>
Reviewed-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: ipa: replace ipa->suspend_ref with a flag bit

We take a clock reference in ipa_config() in order to prevent the
the IPA clock from being shutdown until a power management suspend
request arrives.  An atomic field in the IPA structure records
whether that extra reference had been taken.

Rather than using an atomic to represent a Boolean value, define
a new flags bitmap, and define a "clock held" flag to represent
whether the extra clock reference has been taken.

Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: ipa: use refcount_t for IPA clock reference count

Take advantage of the checking provided by refcount_t, rather than
using a plain atomic to represent the IPA clock reference count.

Note that we need to *set* the value to 1 in ipa_clock_get() rather
than incrementing it from 0 (because doing that is considered an
error for a refcount_t).

Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: phy: bcm7xxx: request and manage GPHY clock

The internal Gigabit PHY on Broadcom STB chips has a digital clock which
drives its MDIO interface among other things, the driver now requests
and manage that clock during .probe() and .remove() accordingly.

Because the PHY driver can be probed with the clocks turned off we need
to apply the dummy BMSR workaround during the driver probe function to
ensure subsequent MDIO read or write towards the PHY will succeed.

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

net: tipc: Supply missing udp_media.h include file

If the header file containing a function's prototype isn't included by
the sourcefile containing the associated function, the build system
complains of missing prototypes.

Fixes the following W=1 kernel build warning(s):

net/tipc/udp_media.c:446:5: warning: no previous prototype for ‘tipc_udp_nl_dump_remoteip’ [-Wmissing-prototypes]
net/tipc/udp_media.c:532:5: warning: no previous prototype for ‘tipc_udp_nl_add_bearer_data’ [-Wmissing-prototypes]
net/tipc/udp_media.c:614:5: warning: no previous prototype for ‘tipc_udp_nl_bearer_add’ [-Wmissing-prototypes]

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

tipc: Remove unused macro CF_SERVER

It is no used any more, so can remove it.

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

net: hns3: Supply missing hclge_dcb.h include file

If the header file containing a function's prototype isn't included by
the sourcefile containing the associated function, the build system
complains of missing prototypes.

Fixes the following W=1 kernel build warning(s):

drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_dcb.c:453:6: warning: no previous prototype for ‘hclge_dcb_ops_set’ [-Wmissing-prototypes]

Signed-off-by: Wang Hai <wanghai38@huawei.com>
Reviewed-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/appletalk: Supply missing net/Space.h include file

If the header file containing a function's prototype isn't included by
the sourcefile containing the associated function, the build system
complains of missing prototypes.

Fixes the following W=1 kernel build warning(s):

drivers/net/appletalk/cops.c:213:28: warning: no previous prototype for ‘cops_probe’ [-Wmissing-prototypes]
drivers/net/appletalk/ltpc.c:1014:28: warning: no previous prototype for ‘ltpc_probe’ [-Wmissing-prototypes]

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

liquidio: Fix -Wmissing-prototypes warnings for liquidio

If the header file containing a function's prototype isn't included by
the sourcefile containing the associated function, the build system
complains of missing prototypes.

Fixes the following W=1 kernel build warning(s):

drivers/net/ethernet/cavium/liquidio/cn68xx_device.c:124:5: warning: no previous prototype for ‘lio_setup_cn68xx_octeon_device’ [-Wmissing-prototypes]
drivers/net/ethernet/cavium/liquidio/octeon_mem_ops.c:159:1: warning: no previous prototype for ‘octeon_pci_read_core_mem’ [-Wmissing-prototypes]
drivers/net/ethernet/cavium/liquidio/octeon_mem_ops.c:168:1: warning: no previous prototype for ‘octeon_pci_write_core_mem’ [-Wmissing-prototypes]
drivers/net/ethernet/cavium/liquidio/octeon_mem_ops.c:176:5: warning: no previous prototype for ‘octeon_read_device_mem64’ [-Wmissing-prototypes]
drivers/net/ethernet/cavium/liquidio/octeon_mem_ops.c:185:5: warning: no previous prototype for ‘octeon_read_device_mem32’ [-Wmissing-prototypes]
drivers/net/ethernet/cavium/liquidio/octeon_mem_ops.c:194:6: warning: no previous prototype for ‘octeon_write_device_mem32’ [-Wmissing-prototypes]

Signed-off-by: Wang Hai <wanghai38@huawei.com>
Reviewed-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>