call_switchdev_notifiers() doesn't require holding the RTNL lock since
commit ff5cf100 ("net: switchdev: Change notifier chain to be atomic").

We still need it for the "lost event" slow path, to avoid racing against
a concurrent .ndo_bridge_setlink().
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Reviewed-by: Alexandra Winter <wintera@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

In rtl_tx() the released descriptors are zero'ed by
rtl8169_unmap_tx_skb(). And in the beginning of rtl8169_start_xmit()
we check that enough descriptors are free, therefore there's no way
the DescOwn bit can be set here.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Link: https://lore.kernel.org/r/6965d665-6c50-90c5-70e6-0bb335d4ea47@gmail.comSigned-off-by: Jakub Kicinski <kuba@kernel.org>

Struct br_vlan_stats duplicates pcpu_sw_netstats (apart from
br_vlan_stats not defining an alignment requirement), therefore
switch to using the latter one.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Link: https://lore.kernel.org/r/04d25c3d-c5f6-3611-6d37-c2f40243dae2@gmail.comSigned-off-by: Jakub Kicinski <kuba@kernel.org>

Sebastian Andrzej Siewior says:

====================
atm: Replace in_interrupt usage

this mini series contains the removal of in_interrupt() in drivers/atm
====================

Link: https://lore.kernel.org/r/20201116162117.387191-1-bigeasy@linutronix.deSigned-off-by: Jakub Kicinski <kuba@kernel.org>

lanai_shutdown_tx_vci() uses in_interrupt() to issue a warning message
if the function was used in context in which it is not safe to sleep.

The usage of in_interrupt() in driver code is deprecated as it can not always
detect all states where it is not allowed to sleep.

msleep() has debug code which will trigger a warning if used in bad
context.

Remove in_interrupt().

Cc: Chas Williams <3chas3@gmail.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

push_scqe() uses in_interrupt() to figure out if it is allowed to sleep.

The usage of in_interrupt() in drivers is phased out and Linus clearly
requested that code which changes behaviour depending on context should
either be separated or the context be conveyed in an argument passed by the
caller, which usually knows the context.

Aside of that in_interrupt() is not correct as it does not catch preempt
disabled regions which neither can sleep.

ns_send() (the only caller of push_scqe()) has the following callers:

- vcc_sendmsg() used as proto_ops::sendmsg is expected to be invoked in
  preemtible context.
  -> vcc->dev->ops->send() (ns_send())

- atm_vcc::send via atmdev_ops::send either directly (pointer copied by
  atm_init_aal34() or atm_init_aal5()) or via atm_send_aal0().
  This is invoked by drivers (like br2684, clip, pppoatm, ...) which are
  called from net_device_ops::ndo_start_xmit with BH disabled.

Add atmdev_ops::send_bh which is used by callers from BH context
(atm_send_aal*()) and if this callback missing then ::send is used
instead.
Implement this callback in nicstar and use it to replace in_interrupt().

Cc: Chas Williams <3chas3@gmail.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Alex Elder says:

====================
net: ipa: IPA register cleanup

This series consists of cleanup patches, almost entirely related to
the definitions for IPA registers.  Some comments are updated or
added to provide better information about defined IPA registers.
Other cleanups ensure symbol names and their assigned values are
defined consistently.  Some essentially duplicate definitions get
consolidated for simplicity.  In a few cases some minor bugs
(missing definitions) are fixed.  With these changes, all IPA
register offsets and associated field masks should be correct for
IPA versions 3.5.1, 4.0, 4.1, and 4.2.
====================

Link: https://lore.kernel.org/r/20201116233805.13775-1-elder@linaro.orgSigned-off-by: Jakub Kicinski <kuba@kernel.org>

Some last cleanups for the existing IPA register definitions:
  - Remove the definition of IPA_REG_ENABLED_PIPES_OFFSET, because
    it is not used.
  - Use "IPA_" instead of "BAM_" as the prefix on fields associated
    with the FLAVOR_0 register.  We use GSI (not BAM), but the
    fields apply to both GSI and BAM.
  - Get rid of the definition of IPA_CS_RSVD; it is never used.
  - Add two missing field mask definitions for the INIT_DEAGGR
    endpoint register.
  - Eliminate a few of the defined sequencer types, because they
    are unused.  We can add them back when needed.
  - Add a field mask to indicate which bit causes an interrupt on
    the microcontroller.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Move the definition of the ipa_irq_id enumerated type out of
"ipa_interrupt.h" and into "ipa_reg.h", and flesh out its set of
defined values.  Each interrupt id indicates a particular type of
IPA interrupt that can be signaled.  Their numeric values define bit
positions in the IPA_IRQ_* registers, so should their definitions
should accompany the definition of those register offsets.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Move a few things around in "ipa_reg.h":
  - Move the definition of ipa_reg_state_aggr_active_offset() down
    a bit in the file so definitions are ordered by offset (for
    the lowest supported IPA version) like all other definitions.
  - Move the definition TIMER_FREQUENCY to be immediately above
    the definition of ipa_aggr_granularity_val() where it's used.
  - Move each register field value enumerated type definition to
    immediately follow the definitions of the register and field
    it is associated with.
No code functionality is modified by this patch.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Revise or add comments in "ipa_reg.h" for to provide more
information, and to improve clarity and consistency.
  - Always provide a comment to define when a register or field is
    supported (or not) for certain versions of IPA hardware.
  - Try to be specific about *which* or *how many* definitions
    a comment refers to.
  - Move comments stating that ipa->available defines the valid
    bits in various registers *above* the register offset
    definition, to avoid some checkpatch.pl warnings.
No code is changed by this patch.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Consistently define numeric values for enumerated type members using
hexidecimal (rather than decimal) format values.  Align the values
assigned in the same column in each file.

Only assign values where they really matter, for example don't
assign IPA_ENDPOINT_AP_MODEM_TX the value 0.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

The backward compatibility register field masks are defined using
single-bit masks defined with BIT(x) rather than GENMASK(x, x).
Change this one set of definitions to follow the GENMASK() pattern
used everywhere else.  Add a few missing field definitions for this
register as well.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Several IPA register field masks are defined without the "_FMASK"
suffix naming convention.  Rename these, so all field masks are
consistently named.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Rename two suspend IRQ registers so they follow the IPA_REG_IRQ_xxx
naming convention used elsewhere.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

IPA version 3.5.1 represents the timer used in avoiding head-of-line
blocking with a simple tick count.  IPA v4.2 changes that, instead
splitting the timer field into two parts (base and scale) to
represent the ticks in the timer period.

IPA v4.0 and IPA v4.1 use the same method as IPA v3.5.1.  Change the
test in ipa_reg_init_hol_block_timer_val() so the result is correct
for those versions as well.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

For IPA v3.5.1, the IPA filter/routing hash enable register actually
does exist, but it is at offset 0x8c into the IPA register space.
For newer versions of IPA it is at offset 0x148.

Define a new inline function ipa_reg_filt_rout_hash_en_offset() to
return the appropriate value for a given version of IPA hardware.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

The IPA filter/routing hash enable register and filter/routing hash
flush register each have four single-bit fields representing the
four hashed tables to be enabled or flushed.  The field positions
are identical, so just use a single set of field masks to represent
the fields for both registers.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

commit f7365919 ("net: wan: Delete the DLCI / SDLA drivers")
deleted "Documentation/networking/framerelay.rst". However, it is still
referenced in "Documentation/networking/index.rst". We need to remove the
reference, too.
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Xie He <xie.he.0141@gmail.com>
Link: https://lore.kernel.org/r/20201118124226.15588-1-xie.he.0141@gmail.comSigned-off-by: Jakub Kicinski <kuba@kernel.org>

Ido Schimmel says:

====================
mlxsw: Preparations for nexthop objects support - part 2/2

This patch set contains the second round of preparations towards nexthop
objects support in mlxsw. Follow up patches can be found here [1].

The patches are mostly small and trivial and contain non-functional
changes aimed at making it easier to integrate nexthop objects with
mlxsw.

Patch #1 is a fix for an issue introduced in previous submission. Found
by Coverity.

[1] https://github.com/idosch/linux/tree/submit/nexthop_objects
====================

Link: https://lore.kernel.org/r/20201117174704.291990-1-idosch@idosch.orgSigned-off-by: Jakub Kicinski <kuba@kernel.org>

The function is responsible for allocating the adjacency entries used by
the nexthop group and populating them with the adjacency information
such as egress RIF and MAC address.

Allow the function to return an error when it encounters a problem and
have the relevant call sites check it.
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Currently, a nexthop group is destroyed when the last FIB entry is
detached from it.

When nexthop objects are supported, this can no longer be the case, as
the group is a separate object whose lifetime is managed by user space.

Add an indication if a nexthop group can be destroyed and always set it
to true for the existing IPv4 and IPv6 nexthop groups.
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

When the IPv6 FIB info has a nexthop object, the nexthop offload
indication is set on the nexthop object and not on the FIB info itself.

Therefore, do not try to clear the offload indication from the FIB info
when it has a nexthop object.
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Attach the FIB entry to the nexthop group after setting the offload flag
on the IPv6 FIB info (i.e., 'struct fib6_info'). The second operation is
not needed when the nexthop group is a nexthop object. This will allow
us to have a common exit path from the function, regardless of the
nexthop group's type.
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

The previous patch associated a nexthop group with the FIB entry before
the entry's type is determined.

Make use of the nexthop group when determining the entry's type instead
of relying on helpers that assume that the nexthop info is not a nexthop
object (i.e., 'struct nexthop').
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Each FIB entry has a type (e.g., remote, local) that determines how the
entry is programmed to the device. In order to determine if the entry is
local (directly connected) or remote (has a gateway) the relevant FIB
info structures (e.g., 'struct fib_info') are checked.

When entries that use nexthop objects are supported, these checks will
need to be changed to take into account 'struct nexthop'.

Instead, first associate the entry with a nexthop group so that the next
patch could determine the entry's type based on the associated nexthop
group's type.
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

The sole caller of the function will soon only have the ifindex
available, instead of the pointer itself.

Therefore, change the function to take the ifindex as input and have it
get the pointer.
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

The ifindex of the nexthop device was never set for IPv4 nexthops,
unlike IPv6 nexthops. This went unnoticed since only IPv6 nexthops use
it.

Set the ifindex for IPv4 nexthops in order to be consistent with IPv6
and also because it will be used by a later patch.
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

The function allocates 'nhgi', not 'nh_grp', so it needs to free the
former in its error path.

Fixes: 7f7a417e ("mlxsw: spectrum_router: Split nexthop group configuration to a different struct")
Addresses-Coverity: ("Memory - corruptions  (USE_AFTER_FREE)")
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

It's arguable most people interested in configuring a PPS signal
want it as external output, not as kernel input. PTP_CLK_REQ_PPS
is for input though. Add documentation to nudge readers into
the correct direction.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Link: https://lore.kernel.org/r/20201117213826.18235-1-a.fatoum@pengutronix.deSigned-off-by: Jakub Kicinski <kuba@kernel.org>

Fix the following unreachable code issue:

   drivers/net/ethernet/netronome/nfp/crypto/tls.c: In function 'nfp_net_tls_add':
   include/linux/compiler_attributes.h:208:41: warning: statement will never be executed [-Wswitch-unreachable]
     208 | # define fallthrough                    __attribute__((__fallthrough__))
         |                                         ^~~~~~~~~~~~~
   drivers/net/ethernet/netronome/nfp/crypto/tls.c:299:3: note: in expansion of macro 'fallthrough'
     299 |   fallthrough;
         |   ^~~~~~~~~~~
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Link: https://lore.kernel.org/r/20201117171347.GA27231@embeddedorSigned-off-by: Jakub Kicinski <kuba@kernel.org>

Mauro Carvalho Chehab says:

====================
Fix several bad kernel-doc markups

Kernel-doc has always be limited to a probably bad documented
rule:

The kernel-doc markups should appear *imediatelly before* the
function or data structure that it documents.

On other words, if a C file would contain something like this:

	/**
	 * foo - function foo
	 * @args: foo args
	 */
	static inline void bar(int args);

	/**
	 * bar - function bar
	 * @args: foo args
	 */
	static inline void foo(void *args);

The output (in ReST format) will be:

	.. c:function:: void bar (int args)

	   function foo

	**Parameters**

	``int args``
	  foo args

	.. c:function:: void foo (void *args)

	   function bar

	**Parameters**

	``void *args``
	  foo args

Which is clearly a wrong result.  Before this changeset,
not even a warning is produced on such cases.

As placing such markups just before the documented
data is a common practice, on most cases this is fine.

However, as patches touch things, identifiers may be
renamed, and people may forget to update the kernel-doc
markups to follow such changes.

This has been happening for quite a while, as there are
lots of files with kernel-doc problems.

This series address those issues and add a file at the
end that will enforce that the identifier will match the
kernel-doc markup, avoiding this problem from
keep happening as time goes by.
====================

Link: https://lore.kernel.org/r/cover.1605521731.git.mchehab+huawei@kernel.orgSigned-off-by: Jakub Kicinski <kuba@kernel.org>

Some identifiers have different names between their prototypes
and the kernel-doc markup.

In the specific case of netif_subqueue_stopped(), keep the
current markup for __netif_subqueue_stopped(), adding a
new one for netif_subqueue_stopped().
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Some identifiers have different names between their prototypes
and the kernel-doc markup.
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Reviewed-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Some functions have different names between their prototypes
and the kernel-doc markup.
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Naveen Mamindlapalli says:

====================
Add ethtool ntuple filters support

This patch series adds support for ethtool ntuple filters, unicast
address filtering, VLAN offload and SR-IOV ndo handlers. All of the
above features are based on the Admin Function(AF) driver support to
install and delete the low level MCAM entries. Each MCAM entry is
programmed with the packet fields to match and what actions to take
if the match succeeds. The PF driver requests AF driver to allocate
set of MCAM entries to be used to install the flows by that PF. The
entries will be freed when the PF driver is unloaded.

* The patches 1 to 4 adds AF driver infrastructure to install and
  delete the low level MCAM flow entries.
* Patch 5 adds ethtool ntuple filter support.
* Patch 6 adds unicast MAC address filtering.
* Patch 7 adds support for dumping the MCAM entries via debugfs.
* Patches 8 to 10 adds support for VLAN offload.
* Patch 10 to 11 adds support for SR-IOV ndo handlers.
* Patch 12 adds support to read the MCAM entries.

Misc:
* Removed redundant mailbox NIX_RXVLAN_ALLOC.
====================

Link: https://lore.kernel.org/r/20201114195303.25967-1-naveenm@marvell.comSigned-off-by: Jakub Kicinski <kuba@kernel.org>

Since mailbox message for installing flows is in place,
remove the RXVLAN_ALLOC mbox message which is redundant.
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: Naveen Mamindlapalli <naveenm@marvell.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

This patch introduces new mailbox mesages to retrieve a given
MCAM entry or base flow steering rule of a VF installed by its
parent PF. This helps while updating the existing MCAM rules
with out re-framing the whole mailbox request again. The INSTALL
FLOW mailbox consumer can read-modify-write the existing entry.
Similarly while installing new flow rules for a VF, the base
flow steering rule match creteria is copied to the new flow rule
and the deltas are appended to the new rule.
Signed-off-by: Naveen Mamindlapalli <naveenm@marvell.com>
Co-developed-by: Vamsi Attunuru <vattunuru@marvell.com>
Signed-off-by: Vamsi Attunuru <vattunuru@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

This patch handles the VF mac address changes as given below.
    1. mac addr configrued by VF will be retained until VF module unload.
    2. mac addr configred by PF for VF will be retained until power cycle.
    3. mac addr confgired by PF for its VF can't be overwritten by VF.
Signed-off-by: Hariprasad Kelam <hkelam@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: Naveen Mamindlapalli <naveenm@marvell.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

This patch adds support for ndo_set_vf_mac, ndo_set_vf_vlan
and ndo_get_vf_config handlers. The traffic redirection
based on the VF mac address or vlan id is done by installing
MCAM rules. Reserved RX_VTAG_TYPE7 in each NIXLF for VF VLAN
which strips the VLAN tag from ingress VLAN traffic. The NIX PF
allocates two MCAM entries for VF VLAN feature, one used for
ingress VTAG strip and another entry for egress VTAG insertion.

This patch also updates the MAC address in PF installed VF VLAN
rule upon receiving nix_lf_start_rx mbox request for VF since
Administrative Function driver will assign a valid MAC addr
in nix_lf_start_rx function.
Signed-off-by: Naveen Mamindlapalli <naveenm@marvell.com>
Co-developed-by: Tomasz Duszynski <tduszynski@marvell.com>
Signed-off-by: Tomasz Duszynski <tduszynski@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: Hariprasad Kelam <hkelam@marvell.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>