Kuniyuki Iwashima says:

====================
af_unix: Introduce per-netns socket hash table.

This series replaces unix_socket_table with a per-netns hash table and
reduces lock contention and time on iterating over the list.

Note the 3rd-6th patches can be a single patch, but for ease of review,
they are split into small changes without breakage.

Changes:
  v3:
    6th:
      * Remove unix_table_locks from comments.
      * Remove missed spin_unlock(&unix_table_locks) in
        unix_lookup_by_ino() (kernel test robot)

  v2: https://lore.kernel.org/netdev/20220620185151.65294-1-kuniyu@amazon.com/
    3rd:
      * Update changelog
      * Remove holes from per-netns hash table structure
      * Use kvmalloc_array() instead of kmalloc() (Eric Dumazet)
      * Remove unnecessary parts in af_unix_init() (Eric Dumazet)
      * Move `err_sysctl` label into ifdef block (kernel test robot)
      * Remove struct netns_unix from struct net if CONFIG_UNIX is disabled
    4th:
      * Use spin_lock_nested() (kernel test robot)

  v1: https://lore.kernel.org/netdev/20220616234714.4291-1-kuniyu@amazon.com/
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

unix_table_locks are to protect the global hash table, unix_socket_table.
The previous commit removed it, so let's clean up the unnecessary locks.

Here is a test result on EC2 c5.9xlarge where 10 processes run concurrently
in different netns and bind 100,000 sockets for each.

  without this series : 1m 38s
  with this series    :    11s

It is ~10x faster because the global hash table is split into 10 netns in
this case.
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This commit replaces the global hash table with a per-netns one and removes
the global one.

We now link a socket in each netns's hash table so we can save some netns
comparisons when iterating through a hash bucket.
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This commit adds extra spin_lock/spin_unlock() for a per-netns
hash table inside the existing ones for unix_table_locks.

As of this commit, sockets are still linked in the global hash
table.  After putting sockets in a per-netns hash table and
removing the old one in the next patch, we remove the global
locks in the last patch.
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This commit adds a per netns hash table for AF_UNIX, which size is fixed
as UNIX_HASH_SIZE for now.

The first implementation defines a per-netns hash table as a single array
of lock and list:

	struct unix_hashbucket {
		spinlock_t		lock;
		struct hlist_head	head;
	};

	struct netns_unix {
		struct unix_hashbucket	*hash;
		...
	};

But, Eric pointed out memory cost that the structure has holes because of
sizeof(spinlock_t), which is 4 (or more if LOCKDEP is enabled). [0]  It
could be expensive on a host with thousands of netns and few AF_UNIX
sockets.  For this reason, a per-netns hash table uses two dense arrays.

	struct unix_table {
		spinlock_t		*locks;
		struct hlist_head	*buckets;
	};

	struct netns_unix {
		struct unix_table	table;
		...
	};

Note the length of the list has a significant impact rather than lock
contention, so having shared locks can be an option.  But, per-netns
locks and lists still perform better than the global locks and per-netns
lists. [1]

Also, this patch adds a change so that struct netns_unix disappears from
struct net if CONFIG_UNIX is disabled.

[0]: https://lore.kernel.org/netdev/CANn89iLVxO5aqx16azNU7p7Z-nz5NrnM5QTqOzueVxEnkVTxyg@mail.gmail.com/
[1]: https://lore.kernel.org/netdev/20220617175215.1769-1-kuniyu@amazon.com/Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Currently, the size of AF_UNIX hash table is UNIX_HASH_SIZE * 2,
the first half for bind()ed sockets and the second half for unbound
ones.  UNIX_HASH_SIZE * 2 is used to define the table and iterate
over it.

In some places, we use ARRAY_SIZE(unix_socket_table) instead of
UNIX_HASH_SIZE * 2.  However, we cannot use it anymore because we
will allocate the hash table dynamically.  Then, we would have to
add UNIX_HASH_SIZE * 2 in many places, which would be troublesome.

This patch adapts the UNIX_HASH_SIZE definition to include bound
and unbound sockets and defines a new UNIX_HASH_MOD macro to ease
calculations.
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Some functions define a net pointer only for one-shot use.  Others call
sock_net() redundantly even when a net pointer is available.  Let's fix
these and make the code simpler.
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Ido Schimmel says:

====================
mlxsw: Unified bridge conversion - part 2/6

This is the second part of the conversion of mlxsw to the unified bridge
model. Part 1 was merged in commit 4336487e ("Merge branch
'mlxsw-unified-bridge-conversion-part-1'") which includes details about
the new model and the motivation behind the conversion.

This patchset does not begin the conversion, but rather prepares the code
base for it.

Patchset overview:

Patch #1 removes an unnecessary field from one of the FID families.

Patches #2-#7 make various improvements in the layer 2 multicast code,
making it more receptive towards upcoming changes.

Patches #8-#10 prepare the CONFIG_PROFILE command for the unified bridge
model. This command will be used to enable the new model in the last
patchset.

Patches #11-#13 perform small changes in the FID code, preparing it for
upcoming changes.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

rFID and dummy FID do not support FID->VNI mapping. Currently, these
families do not implement the vni_{set, clear}() operations. Instead, there
is a check if these functions are implemented.

Similarly, 'SFMR.nve_tunnel_flood_ptr' is not relevant for rFID and dummy
FID, therefore, these families do not implement
nve_flood_index_{set, clear}().

Align the behavior to other unsupported operations, implement the functions
and just return an error or warn. Then, checks like '!ops->vni_set' can be
removed.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The previous patch added 'fid_offset' field to FID structure. Now, this
field can be used when VNI is set using SFMR register. Currently
'fid_offset' is set to zero, instead, use the new field which is now set
to zero and in the future will be changed.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

SFMR register contains a 'fid_offset' field which is used when flooding
tables of type FID offset are used.

Currently, the driver sets this field to zero, as flooding tables of type
FID are used.

Using unified bridge model, the driver will use FID offset flooding
tables. As preparation, add 'fid_offset' to 'struct mlxsw_sp_fid'. Then,
use this field instead of passing zero to the function that configures
SFMR.

Set the new field as part of 'ops->setup()', for that, implement this
function for dummy FID and rFID.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Currently, as part of mlxsw_pci_init(), resources are queried from firmware
before issuing the 'CONFIG_PROFILE' command.

There are resources whose size depend on the enablement of the unified
bridge model that is performed via 'CONFIG_PROFILE' command. As a
preparation for unified bridge model, add an additional query after issuing
this command. Both queries are required as KVD sizes are read from
firmware and then are configured as part of 'CONFIG_PROFILE' command.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Currently, the length of 'config_profile.flood_mode' is defined as 2
bits, while the correct length is 3 bits.

As preparation for unified bridge model, which will use the whole field
length, fix it and increase the field to the correct size.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Currently magic constant is used for setting 'flood_mode' as part of
config profile.

As preparation for unified bridge model, which will require different
'flood_mode', add a dedicated enumerator for this field and use it as
part of 'struct mlxsw_config_profile'.

Add the relevant value for unified bridge model.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The above mentioned function calls two functions which return values, but
it ignores them.

Check if the return value is an error, handle it in such case and return
an error back to the caller.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The function returns 'false' upon failure and 'true' upon success.
Convert it to the usual scheme of returning integer error codes and
align the callers.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The above mentioned function just returns an error in case that
mlxsw_sp_bridge_ports_flood_table_set() fails. That means that the previous
configurations are not cleaned.

Fix it by adding error path to clean the configurations in case of error.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The above mentioned function is called for each port in bridge when the
multicast behavior is changed. Currently, it updates all the relevant
MID entries only for the first call, but the update of flooding per port
is done only for the port that it was called for it.

To simplify this behavior, it is possible to handle flooding for all the
ports in the first call, then, the other calls will do nothing. For
that, new functions are required to set flooding for all ports, no
'struct mlxsw_sp_port' is required.

This issue was found while extending this function for unified bridge
model, the above mentioned change will ease the extending later.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The function mlxsw_sp_port_mc_disabled_set() sets
'bridge_device->multicast_enabled' twice. Remove the unnecessary setting.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Currently the above mentioned function gets 'struct mlxsw_sp_port',
while the only use of this structure is to get 'mlxsw_sp_port->mlxsw_sp'.

Instead, pass 'struct mlxsw_sp'.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Currently 'struct mlxsw_sp_fid_family' has a field which indicates if
'lag_vid' is valid for use in SFD register.

This is a leftover from using .1Q FIDs instead of emulating them using
.1D FIDs.

Currently when .1Q FIDs are emulated using .1D FIDs, this field is true
for both families, so there is no reason to maintain it.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Convert MSE (mean-square error) values to SNR and split it SQI (Signal Quality
Indicator) ranges. The used ranges are taken from "OPEN ALLIANCE - Advanced
diagnostic features for 100BASE-T1 automotive Ethernet PHYs"
specification.
Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20220620115601.2035452-1-o.rempel@pengutronix.deSigned-off-by: Jakub Kicinski <kuba@kernel.org>

Since commit 4c0d2e96 ("net: phy: consider that suspend2ram may cut
off PHY power"), phy_init_hw() invokes both, the ->config_init() and
->config_intr() callbacks.

In the SMSC PHY driver, the latter acknowledges stale interrupts, hence
there's no longer a need to acknowledge them in the former as well.

There are no other callers of ->config_init() besides phy_init_hw().

Drop the redundant code.
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Link: https://lore.kernel.org/r/0254edf48bddc96c6248c4414043a3699e94614a.1655716767.git.lukas@wunner.deSigned-off-by: Jakub Kicinski <kuba@kernel.org>

Arun Ramadoss says:

====================
net: dsa: microchip: common spi probe for the ksz series switches - part 1

This patch series aims to refactor the ksz_switch_register routine to have the
common flow for the ksz series switch. At present ksz8795.c & ksz9477.c have
its own dsa_switch_ops and switch detect functionality.
In ksz_switch_register, ksz_dev_ops is assigned based on the function parameter
passed by the individual ksz8/ksz9477 switch register function. And then switch
detect is performed based on the ksz_dev_ops.detect hook.  This patch modifies
the ksz_switch_register such a way that switch detect is performed first, based
on the chip ksz_dev_ops is assigned to ksz_device structure. It ensures the
common flow for the existing as well as LAN937x switches.
In the next series of patch, it will move ksz_dsa_ops and dsa_switch_ops
from ksz8795.c and ksz9477.c to ksz_common.c and have the common spi
probe all the ksz based switches.

Changes in v1
- Splitted the patch series into two.
- Replaced all occurrence of REG_PORT_STATUS_0 and PORT_FIBER_MODE to
  KSZ8_PORT_STATUS_0 and KSZ8_PORT_FIBER_MODE.
- Separated the tag protocol and phy read/write patch into two.
- Assigned the DSA_TAG_PROTO_NONE as the default value for get_tag_protocol hook.
- Reduced the indentation level by using the if(!dev->dev_ops->mirror_add).
- Added the stp_ctrl_reg as a member in ksz_chip_data and removed the member
  in ksz_dev_ops.
- Removed the r_dyn_mac_table, r_sta_mac_table and w_sta_mac_table from the
  ksz_dev_ops since it is used only in the ksz8795.c.

Changes in RFC v2
- Fixed the compilation issue.
- Reduced the patch set to 15.
====================
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Link: https://lore.kernel.org/r/20220617084255.19376-1-arun.ramadoss@microchip.comSigned-off-by: Paolo Abeni <pabeni@redhat.com>

This patch assigns the get_phy_flags & mtu  hook of ksz8795 and ksz9477
in dsa_switch_ops to ksz_common.  For get_phy_flags hooks,checks whether
the chip is ksz8863/kss8793 then it returns error for port1.
Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

This patch makes the dsa_switch_hook for fdbs to use ksz_common.c file.
And from ksz_common, individual switches fdb functions are called using
the dev->dev_ops. And removed the r_dyn_mac_table, r_sta_mac_table and
w_sta_mac_table from ksz_dev_ops as it is used only in ksz8795.c
Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

ksz_mdb_add/del in ksz_common.c is specific for the ksz8795.c file. The
ksz9477 has its separate ksz9477_port_mdb_add/del functions.  This patch
moves the ksz8795 specific mdb functionality from ksz_common to ksz8795.
And this dsa_switch_ops hooks for ksz8795/ksz9477 are invoked through
the ksz_port_mdb_add/del.
Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

This patch assigns the phylink_get_caps in ksz8795 and ksz9477 to
ksz_phylink_get_caps. And update their mac_capabilities in the
respective ksz_dev_ops.
Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

At present, P_STP_CTRL register value is passed as parameter to
ksz_port_stp_state from the individual dsa_switch_ops hooks. This patch
update the function to retrieve the register value through the
ksz_chip_data member.
And add the static to ksz_update_port_member since it is not called
outside the ksz_common.
Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

This patch updates the common port mirror add/del dsa_switch_ops in
ksz_common.c. The individual switches implementation is executed based
on the ksz_dev_ops function pointers.
Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

This patch moves the vlan dsa_switch_ops such as vlan_add, vlan_del and
vlan_filtering from the individual files ksz8795.c, ksz9477.c to
ksz_common.c file.
Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

ksz8795 and ksz9477 implementation on phy read/write hooks are
different. This patch modifies the ksz9477 implementation same as
ksz8795 by updating the ksz9477_dev_ops structure.
Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

This patch move the dsa hook get_tag_protocol to ksz_common file. And
the tag_protocol is returned based on the dev->chip_id.
Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

KSZ87xx and KSZ88xx have chip_id representation at reg location 0. And
KSZ9477 compatible switch and LAN937x switch have same chip_id detection
at location 0x01 and 0x02. To have the common switch detect
functionality for ksz switches, ksz_switch_detect function is
introduced.
Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

The ksz9477_switch_detect performs the detecting the chip id from the
location 0x00 and also check gigabit compatibility check & number of
ports based on the register global_options0. To prepare the common ksz
switch detect function, routine other than chip id read is moved to
ksz9477_switch_init.
Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

During initialization of the NFP driver, a file name for loading
application firmware is composed using the NIC's AMDA information and port
type (count and speed). E.g.: "nic_AMDA0145-1012_2x10.nffw".

In practice there may be many variants for each NIC type, and many of the
variants relate to assembly components which do not concern the driver and
application firmware implementation. Yet the current scheme leads to a
different application firmware file name for each variant, because they
have different AMDA information.

To reduce proliferation of content-duplicated application firmware images
or symlinks, the NIC's management firmware will only expose differences
between variants that need different application firmware via a newly
introduced hwinfo, "nffw.partno".

Use of the existing hwinfo, "assembly.partno", is maintained in order to
support for NICs with management firmware that does not expose
"nffw.partno".
Signed-off-by: Yu Xiao <yu.xiao@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Link: https://lore.kernel.org/r/20220620103912.46164-1-simon.horman@corigine.comSigned-off-by: Paolo Abeni <pabeni@redhat.com>

raw_diag_dump() can use rcu_read_lock() instead of read_lock()

Now the hashinfo lock is only used from process context,
in write mode only, we can convert it to a spinlock,
and we do not need to block BH anymore.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20220620100509.3493504-1-eric.dumazet@gmail.comSigned-off-by: Paolo Abeni <pabeni@redhat.com>

Make sure skb_mac_header(), skb_mac_offset() and skb_mac_header_len() uses
are not fooled if the mac header has not been set.

These checks are enabled if CONFIG_DEBUG_NET=y

This commit will likely expose existing bugs in linux networking stacks.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20220620093017.3366713-1-eric.dumazet@gmail.comSigned-off-by: Paolo Abeni <pabeni@redhat.com>

Ido Schimmel says:

====================
mlxsw: Unified bridge conversion - part 1/6

This set starts converting mlxsw to the unified bridge model and mainly
adds new device registers and extends existing ones that will be used in
follow-up patchsets.

High-level summary
==================

The unified bridge model is a new way of managing low-level device
objects such as filtering identifiers (FIDs). The conversion moves a lot
of logic out of the device's firmware towards the driver, but its main
selling point is that it allows to overcome various scalability issues
related to the amount of entries that need to be programmed to the
device.

The only (intended) user visible changes of the conversion are
improvement in resource utilization and ability to support more router
interfaces (RIFs) in Spectrum-{2,3}.

Details
=======

Commit 50853808 ("Merge branch
'mlxsw-Prepare-for-VLAN-aware-bridge-w-VxLAN'") converted mlxsw to
emulate 802.1Q FIDs (represent VLANs in a VLAN-aware bridge) using
802.1D FIDs (represent VLAN-unaware bridges). This was necessary because
at that time VNI could not be assigned to 802.1Q FIDs, which effectively
meant that mlxsw could not support VXLAN with VLAN-aware bridges.

The downside of this approach is that multiple {Port,VID}->FID entries
are required in order to classify incoming traffic to a FID, as opposed
to a single VID->FID entry that can be used with actual 802.1Q FIDs.

For example, if 10 ports are members in the same VLAN-aware bridge and
the same 100 VLANs are configured on each port, then only 100 VID->FID
entries are required with 802.1Q FIDs, whereas 1000 {Port,VID}->FID
entries are required with emulated 802.1Q FIDs.

The above limitation is the result of various assumptions that were made
in the design of the API that was exposed to software. In the unified
bridge model the API is much more "raw" and therefore avoids these
assumptions, allowing software to configure the device in a more
efficient manner.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Router interfaces (RIFs) constructed on top of VLAN-aware bridges are of
"VLAN" type, whereas RIFs constructed on top of VLAN-unaware bridges of
"FID" type.

In other words, the RIF type is derived from the underlying FID type.
VLAN RIFs are used on top of 802.1Q FIDs, whereas FID RIFs are used on
top of 802.1D FIDs.

Currently 802.1Q FIDs are emulated using 802.1D FIDs, and therefore VLAN
RIFs are emulated using FID RIFs.

As part of converting the driver to use unified bridge, 802.1Q FIDs and
VLAN RIFs will be used.

Add the relevant fields to RITR register, add pack() function for VLAN
RIF and rename one field to fit the internal name.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>