Saeed Mahameed says:
====================
mlx5-updates-2021-10-25
Misc updates for mlx5 driver:
1) Misc updates and cleanups:
- Don't write directly to netdev->dev_addr, From Jakub Kicinski
- Remove unnecessary checks for slow path flag in tc module
- Fix unused function warning of mlx5i_flow_type_mask
- Bridge, support replacing existing FDB entry
2) Sub Functions, Reduction in memory usage:
- Reduce flow counters bulk query buffer size
- Implement max_macs devlink parameter
- Add devlink vendor params to control Event Queue sizes
- Added SF life cycle trace points by Parav/
3) From Aya, Firmware health buffer reporting improvements
- Print health buffer by log level and more missing information
- Periodic update of host time to firmware
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
RFC 5082 IP_MINTTL option is rarely used on hosts.
Add a static key to remove from TCP fast path useless code,
and potential cache line miss to fetch inet_sk(sk)->min_ttl
Note that once ip4_min_ttl static key has been enabled,
it stays enabled until next boot.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
RFC 5082 IPV6_MINHOPCOUNT is rarely used on hosts.
Add a static key to remove from TCP fast path useless code,
and potential cache line miss to fetch tcp_inet6_sk(sk)->min_hopcount
Note that once ip6_min_hopcount static key has been enabled,
it stays enabled until next boot.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
sk->sk_rx_queue_mapping can be modified locklessly,
add a couple of READ_ONCE()/WRITE_ONCE() to document this fact.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
sk_rx_queue_mapping is located in a cache line that should be kept read mostly.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
sk_napi_id is located in a cache line that can be kept read mostly.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Increase cache locality by moving rx_dst_coookie next to sk->sk_rx_dst
This removes one or two cache line misses in IPv6 early demux (TCP/UDP)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Increase cache locality by moving rx_dst_ifindex next to sk->sk_rx_dst
This is part of an effort to reduce cache line misses in TCP fast path.
This removes one cache line miss in early demux.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently, max_macs is taking 70Kbytes of memory per function. This
size is not needed in all use cases, and is critical with large scale.
Hence, allow user to configure the number of max_macs.
For example, to reduce the number of max_macs to 1, execute::
$ devlink dev param set pci/0000:00:0b.0 name max_macs value 1 \
cmode driverinit
$ devlink dev reload pci/0000:00:0b.0
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Parav Pandit <parav@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Event EQ is an EQ which received the notification of almost all the
events generated by the NIC.
Currently, each event EQ is taking 512KB of memory. This size is not
needed in most use cases, and is critical with large scale. Hence,
allow user to configure the size of the event EQ.
For example to reduce event EQ size to 64, execute::
$ devlink resource set pci/0000:00:0b.0 path /event_eq_size/ size 64
$ devlink dev reload pci/0000:00:0b.0
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Parav Pandit <parav@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Currently, each I/O EQ is taking 128KB of memory. This size
is not needed in all use cases, and is critical with large scale.
Hence, allow user to configure the size of I/O EQs.
For example, to reduce I/O EQ size to 64, execute:
$ devlink resource set pci/0000:00:0b.0 path /io_eq_size/ size 64
$ devlink dev reload pci/0000:00:0b.0
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Parav Pandit <parav@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Firmware logs its asserts also to non-volatile memory. In order to
reduce drift between the NIC and the host, the driver sets the host
epoch-time to the firmware every hour.
Signed-off-by: Aya Levin <ayal@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
tls already supports the SM4 GCM/CCM algorithms. It is also necessary
to add support for these two algorithms in tls_crypto_context to avoid
potential issues caused by forced type conversion.
Signed-off-by: Tianjia Zhang <tianjia.zhang@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add the constants for 2.5G fast retrain capability
in 10G AN control register, fast retrain status and
control register and THP bypass register into mdio.h.
Signed-off-by: Luo Jie <luoj@codeaurora.org>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now that the rtnl_mutex is going away for dsa_port_{host_,}fdb_{add,del},
no one is serializing access to the address lists that DSA keeps for the
purpose of reference counting on shared ports (CPU and cascade ports).
It can happen for one dsa_switch_do_fdb_del to do list_del on a dp->fdbs
element while another dsa_switch_do_fdb_{add,del} is traversing dp->fdbs.
We need to avoid that.
Currently dp->mdbs is not at risk, because dsa_switch_do_mdb_{add,del}
still runs under the rtnl_mutex. But it would be nice if it would not
depend on that being the case. So let's introduce a mutex per port (the
address lists are per port too) and share it between dp->mdbs and
dp->fdbs.
The place where we put the locking is interesting. It could be tempting
to put a DSA-level lock which still serializes calls to
.port_fdb_{add,del}, but it would still not avoid concurrency with other
driver code paths that are currently under rtnl_mutex (.port_fdb_dump,
.port_fast_age). So it would add a very false sense of security (and
adding a global switch-wide lock in DSA to resynchronize with the
rtnl_lock is also counterproductive and hard).
So the locking is intentionally done only where the dp->fdbs and dp->mdbs
lists are traversed. That means, from a driver perspective, that
.port_fdb_add will be called with the dp->addr_lists_lock mutex held on
the CPU port, but not held on user ports. This is done so that driver
writers are not encouraged to rely on any guarantee offered by
dp->addr_lists_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>
DSA would like to remove the rtnl_lock from its
SWITCHDEV_FDB_{ADD,DEL}_TO_DEVICE handlers, and the felix driver uses
the same MAC table functions as ocelot.
This means that the MAC table functions will no longer be implicitly
serialized with respect to each other by the rtnl_mutex, we need to add
a dedicated lock in ocelot for the non-atomic operations of selecting a
MAC table row, reading/writing what we want and polling for completion.
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>
Marc Kleine-Budde says:
====================
pull-request: can-next 2021-10-24
this is a pull request of 15 patches for net-next/master.
The first patch is by Thomas Gleixner and makes use of
hrtimer_forward_now() in the CAN broad cast manager (bcm).
The next patch is by me and changes the type of the variables used in
the CAN bit timing calculation can_fixup_bittiming() to unsigned int.
Vincent Mailhol provides 6 patches targeting the CAN device
infrastructure. The CAN-FD specific Transmitter Delay Compensation
(TDC) is updated and configuration via the CAN netlink interface is
added.
Qing Wang's patch updates the at91 and janz-ican3 drivers to use
sysfs_emit() instead of snprintf() in the sysfs show functions.
Geert Uytterhoeven's patch drops the unneeded ARM dependency from the
rar Kconfig.
Cai Huoqing's patch converts the mscan driver to make use of the
dev_err_probe() helper function.
A patch by me against the gsusb driver changes the printf format
strings to use %u to print unsigned values.
Stephane Grosjean's patch updates the peak_usb CAN-FD driver to use
the 64 bit timestamps provided by the hardware.
The last 2 patches target the xilinx_can driver. Michal Simek provides
a patch that removes repeated word from the kernel-doc and Dongliang
Mu's patch removes a redundant netif_napi_del() from the xcan_remove()
function.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
struct can_tdc::tdco represents the absolute offset from TDCV. Some
controllers use instead an offset relative to the Sample Point (SP)
such that:
| SSP = TDCV + absolute TDCO
| = TDCV + SP + relative TDCO
Consequently:
| relative TDCO = absolute TDCO - SP
The function can_tdc_get_relative_tdco() allow to retrieve this
relative TDCO value.
Link: https://lore.kernel.org/all/20210918095637.20108-7-mailhol.vincent@wanadoo.fr
CC: Stefan Mätje <Stefan.Maetje@esd.eu>
Signed-off-by: Vincent Mailhol <mailhol.vincent@wanadoo.fr>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Some CAN device can measure the TDCV (Transmission Delay Compensation
Value) automatically for each transmitted CAN frames.
A callback function do_get_auto_tdcv() is added to retrieve that
value. This function is used only if CAN_CTRLMODE_TDC_AUTO is enabled
(if CAN_CTRLMODE_TDC_MANUAL is selected, the TDCV value is provided by
the user).
If the device does not support reporting of TDCV, do_get_auto_tdcv()
should be set to NULL and TDCV will not be reported by the netlink
interface.
On success, do_get_auto_tdcv() shall return 0. If the value can not be
measured by the device, for example because network is down or because
no frames were transmitted yet, can_priv::do_get_auto_tdcv() shall
return a negative error code (e.g. -EINVAL) to signify that the value
is not yet available. In such cases, TDCV is not reported by the
netlink interface.
Link: https://lore.kernel.org/all/20210918095637.20108-6-mailhol.vincent@wanadoo.fr
CC: Stefan Mätje <stefan.maetje@esd.eu>
Signed-off-by: Vincent Mailhol <mailhol.vincent@wanadoo.fr>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Add the netlink interface for TDC parameters of struct can_tdc_const
and can_tdc.
Contrary to the can_bittiming(_const) structures for which there is
just a single IFLA_CAN(_DATA)_BITTMING(_CONST) entry per structure,
here, we create a nested entry IFLA_CAN_TDC. Within this nested entry,
additional IFLA_CAN_TDC_TDC* entries are added for each of the TDC
parameters of the newly introduced struct can_tdc_const and struct
can_tdc.
For struct can_tdc_const, these are:
IFLA_CAN_TDC_TDCV_MIN
IFLA_CAN_TDC_TDCV_MAX
IFLA_CAN_TDC_TDCO_MIN
IFLA_CAN_TDC_TDCO_MAX
IFLA_CAN_TDC_TDCF_MIN
IFLA_CAN_TDC_TDCF_MAX
For struct can_tdc, these are:
IFLA_CAN_TDC_TDCV
IFLA_CAN_TDC_TDCO
IFLA_CAN_TDC_TDCF
This is done so that changes can be applied in the future to the
structures without breaking the netlink interface.
The TDC netlink logic works as follow:
* CAN_CTRLMODE_FD is not provided:
- if any TDC parameters are provided: error.
- TDC parameters not provided: TDC parameters unchanged.
* CAN_CTRLMODE_FD is provided and is false:
- TDC is deactivated: both the structure and the
CAN_CTRLMODE_TDC_{AUTO,MANUAL} flags are flushed.
* CAN_CTRLMODE_FD provided and is true:
- CAN_CTRLMODE_TDC_{AUTO,MANUAL} and tdc{v,o,f} not provided: call
can_calc_tdco() to automatically decide whether TDC should be
activated and, if so, set CAN_CTRLMODE_TDC_AUTO and uses the
calculated tdco value.
- CAN_CTRLMODE_TDC_AUTO and tdco provided: set
CAN_CTRLMODE_TDC_AUTO and use the provided tdco value. Here,
tdcv is illegal and tdcf is optional.
- CAN_CTRLMODE_TDC_MANUAL and both of tdcv and tdco provided: set
CAN_CTRLMODE_TDC_MANUAL and use the provided tdcv and tdco
value. Here, tdcf is optional.
- CAN_CTRLMODE_TDC_{AUTO,MANUAL} are mutually exclusive. Whenever
one flag is turned on, the other will automatically be turned
off. Providing both returns an error.
- Combination other than the one listed above are illegal and will
return an error.
N.B. above rules mean that whenever CAN_CTRLMODE_FD is provided, the
previous TDC values will be overwritten. The only option to reuse
previous TDC value is to not provide CAN_CTRLMODE_FD.
All the new parameters are defined as u32. This arbitrary choice is
done to mimic the other bittiming values with are also all of type
u32. An u16 would have been sufficient to hold the TDC values.
This patch completes below series (c.f. [1]):
- commit 289ea9e4ae ("can: add new CAN FD bittiming parameters:
Transmitter Delay Compensation (TDC)")
- commit c25cc79932 ("can: bittiming: add calculation for CAN FD
Transmitter Delay Compensation (TDC)")
[1] https://lore.kernel.org/linux-can/20210224002008.4158-1-mailhol.vincent@wanadoo.fr/T/#t
Link: https://lore.kernel.org/all/20210918095637.20108-5-mailhol.vincent@wanadoo.fr
Signed-off-by: Vincent Mailhol <mailhol.vincent@wanadoo.fr>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
The function can_calc_tdco() directly retrieves can_priv from the
net_device and directly modifies it.
This is annoying for the upcoming patch. In
drivers/net/can/dev/netlink.c:can_changelink(), the data bittiming are
written to a temporary structure and memcpyed to can_priv only after
everything succeeded. In the next patch, where we will introduce the
netlink interface for TDC parameters, we will add a new TDC block
which can potentially fail. For this reason, the data bittiming
temporary structure has to be copied after that to-be-introduced TDC
block. However, TDC also needs to access data bittiming information.
We change the prototype so that the data bittiming structure is passed
to can_calc_tdco() as an argument instead of retrieving it from
priv. This way can_calc_tdco() can access the data bittiming before it
gets memcpyed to priv.
Link: https://lore.kernel.org/all/20210918095637.20108-4-mailhol.vincent@wanadoo.fr
Signed-off-by: Vincent Mailhol <mailhol.vincent@wanadoo.fr>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
In the current implementation, all Transmission Delay Compensation
(TDC) parameters are expressed in time quantum. However, ISO 11898-1
actually specifies that these should be expressed in *minimum* time
quantum.
Furthermore, the minimum time quantum is specified to be "one node
clock period long" (c.f. paragraph 11.3.1.1 "Bit time"). For sake of
simplicity, we prefer to use the "clock period" term instead of
"minimum time quantum" because we believe that it is more broadly
understood.
This patch fixes that discrepancy by updating the documentation and
the formula for TDCO calculation.
N.B. In can_calc_tdco(), the sample point (in time quantum) was
calculated using a division, thus introducing a risk of rounding and
truncation errors. On top of changing the unit to clock period, we
also modified the formula to use only additions.
Link: https://lore.kernel.org/all/20210918095637.20108-3-mailhol.vincent@wanadoo.fr
Suggested-by: Stefan Mätje <Stefan.Maetje@esd.eu>
Signed-off-by: Vincent Mailhol <mailhol.vincent@wanadoo.fr>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
ISO 11898-1 specifies in section 11.3.3 "Transmitter delay
compensation" that "the configuration range for [the] SSP position
shall be at least 0 to 63 minimum time quanta."
Because SSP = TDCV + TDCO, it means that we should allow both TDCV and
TDCO to hold zero value in order to honor SSP's minimum possible
value.
However, current implementation assigned special meaning to TDCV and
TDCO's zero values:
* TDCV = 0 -> TDCV is automatically measured by the transceiver.
* TDCO = 0 -> TDC is off.
In order to allow for those values to really be zero and to maintain
current features, we introduce two new flags:
* CAN_CTRLMODE_TDC_AUTO indicates that the controller support
automatic measurement of TDCV.
* CAN_CTRLMODE_TDC_MANUAL indicates that the controller support
manual configuration of TDCV. N.B.: current implementation failed
to provide an option for the driver to indicate that only manual
mode was supported.
TDC is disabled if both CAN_CTRLMODE_TDC_AUTO and
CAN_CTRLMODE_TDC_MANUAL flags are off, c.f. the helper function
can_tdc_is_enabled() which is also introduced in this patch.
Also, this patch adds three fields: tdcv_min, tdco_min and tdcf_min to
struct can_tdc_const. While we are not convinced that those three
fields could be anything else than zero, we can imagine that some
controllers might specify a lower bound on these. Thus, those minimums
are really added "just in case".
Comments of struct can_tdc and can_tdc_const are updated accordingly.
Finally, the changes are applied to the etas_es58x driver.
Link: https://lore.kernel.org/all/20210918095637.20108-2-mailhol.vincent@wanadoo.fr
Signed-off-by: Vincent Mailhol <mailhol.vincent@wanadoo.fr>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Now that the rtnl_mutex is going away for dsa_port_{host_,}fdb_{add,del},
no one is serializing access to the address lists that DSA keeps for the
purpose of reference counting on shared ports (CPU and cascade ports).
It can happen for one dsa_switch_do_fdb_del to do list_del on a dp->fdbs
element while another dsa_switch_do_fdb_{add,del} is traversing dp->fdbs.
We need to avoid that.
Currently dp->mdbs is not at risk, because dsa_switch_do_mdb_{add,del}
still runs under the rtnl_mutex. But it would be nice if it would not
depend on that being the case. So let's introduce a mutex per port (the
address lists are per port too) and share it between dp->mdbs and
dp->fdbs.
The place where we put the locking is interesting. It could be tempting
to put a DSA-level lock which still serializes calls to
.port_fdb_{add,del}, but it would still not avoid concurrency with other
driver code paths that are currently under rtnl_mutex (.port_fdb_dump,
.port_fast_age). So it would add a very false sense of security (and
adding a global switch-wide lock in DSA to resynchronize with the
rtnl_lock is also counterproductive and hard).
So the locking is intentionally done only where the dp->fdbs and dp->mdbs
lists are traversed. That means, from a driver perspective, that
.port_fdb_add will be called with the dp->addr_lists_lock mutex held on
the CPU port, but not held on user ports. This is done so that driver
writers are not encouraged to rely on any guarantee offered by
dp->addr_lists_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>
DSA would like to remove the rtnl_lock from its
SWITCHDEV_FDB_{ADD,DEL}_TO_DEVICE handlers, and the felix driver uses
the same MAC table functions as ocelot.
This means that the MAC table functions will no longer be implicitly
serialized with respect to each other by the rtnl_mutex, we need to add
a dedicated lock in ocelot for the non-atomic operations of selecting a
MAC table row, reading/writing what we want and polling for completion.
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>
7712 is a 16nm process SoC with a 10/100 integrated Ethernet PHY,
utilize the recently defined 16nm EPHY macro to configure that PHY.
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>
This adds some helpers for accessing non-phy MDIO devices. They are
analogous to phy_(read|write|modify), except that they take an mdio_device
and not a phy_device.
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
The change of devlink_register() to be last devlink command together
with delayed notification logic made the publish API to be obsolete.
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Johannes Berg says:
====================
Quite a few changes:
* the applicable eth_hw_addr_set() and const hw_addr changes
* various code cleanups/refactorings
* stack usage reductions across the wireless stack
* some unstructured find_ie() -> structured find_element()
changes
* a few more pieces of multi-BSSID support
* some 6 GHz regulatory support
* 6 GHz support in hwsim, for testing userspace code
* Light Communications (LC, 802.11bb) early band definitions
to be able to add a first driver soon
* tag 'mac80211-next-for-net-next-2021-10-21' of git://git.kernel.org/pub/scm/linux/kernel/git/jberg/mac80211-next: (35 commits)
cfg80211: fix kernel-doc for MBSSID EMA
mac80211: Prevent AP probing during suspend
nl80211: Add LC placeholder band definition to nl80211_band
...
====================
Link: https://lore.kernel.org/r/20211021154953.134849-1-johannes@sipsolutions.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pull ucounts fixes from Eric Biederman:
"There has been one very hard to track down bug in the ucount code that
we have been tracking since roughly v5.14 was released. Alex managed
to find a reliable reproducer a few days ago and then I was able to
instrument the code and figure out what the issue was.
It turns out the sigqueue_alloc single atomic operation optimization
did not play nicely with ucounts multiple level rlimits. It turned out
that either sigqueue_alloc or sigqueue_free could be operating on
multiple levels and trigger the conditions for the optimization on
more than one level at the same time.
To deal with that situation I have introduced inc_rlimit_get_ucounts
and dec_rlimit_put_ucounts that just focuses on the optimization and
the rlimit and ucount changes.
While looking into the big bug I found I couple of other little issues
so I am including those fixes here as well.
When I have time I would very much like to dig into process ownership
of the shared signal queue and see if we could pick a single owner for
the entire queue so that all of the rlimits can count to that owner.
That should entirely remove the need to call get_ucounts and
put_ucounts in sigqueue_alloc and sigqueue_free. It is difficult
because Linux unlike POSIX supports setuid that works on a single
thread"
* 'ucount-fixes-for-v5.15' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace:
ucounts: Move get_ucounts from cred_alloc_blank to key_change_session_keyring
ucounts: Proper error handling in set_cred_ucounts
ucounts: Pair inc_rlimit_ucounts with dec_rlimit_ucoutns in commit_creds
ucounts: Fix signal ucount refcounting
Pull networking fixes from Jakub Kicinski:
"Including fixes from netfilter, and can.
We'll have one more fix for a socket accounting regression, it's still
getting polished. Otherwise things look fine.
Current release - regressions:
- revert "vrf: reset skb conntrack connection on VRF rcv", there are
valid uses for previous behavior
- can: m_can: fix iomap_read_fifo() and iomap_write_fifo()
Current release - new code bugs:
- mlx5: e-switch, return correct error code on group creation failure
Previous releases - regressions:
- sctp: fix transport encap_port update in sctp_vtag_verify
- stmmac: fix E2E delay mechanism (in PTP timestamping)
Previous releases - always broken:
- netfilter: ip6t_rt: fix out-of-bounds read of ipv6_rt_hdr
- netfilter: xt_IDLETIMER: fix out-of-bound read caused by lack of
init
- netfilter: ipvs: make global sysctl read-only in non-init netns
- tcp: md5: fix selection between vrf and non-vrf keys
- ipv6: count rx stats on the orig netdev when forwarding
- bridge: mcast: use multicast_membership_interval for IGMPv3
- can:
- j1939: fix UAF for rx_kref of j1939_priv abort sessions on
receiving bad messages
- isotp: fix TX buffer concurrent access in isotp_sendmsg() fix
return error on FC timeout on TX path
- ice: fix re-init of RDMA Tx queues and crash if RDMA was not inited
- hns3: schedule the polling again when allocation fails, prevent
stalls
- drivers: add missing of_node_put() when aborting
for_each_available_child_of_node()
- ptp: fix possible memory leak and UAF in ptp_clock_register()
- e1000e: fix packet loss in burst mode on Tiger Lake and later
- mlx5e: ipsec: fix more checksum offload issues"
* tag 'net-5.15-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (75 commits)
usbnet: sanity check for maxpacket
net: enetc: make sure all traffic classes can send large frames
net: enetc: fix ethtool counter name for PM0_TERR
ptp: free 'vclock_index' in ptp_clock_release()
sfc: Don't use netif_info before net_device setup
sfc: Export fibre-specific supported link modes
net/mlx5e: IPsec: Fix work queue entry ethernet segment checksum flags
net/mlx5e: IPsec: Fix a misuse of the software parser's fields
net/mlx5e: Fix vlan data lost during suspend flow
net/mlx5: E-switch, Return correct error code on group creation failure
net/mlx5: Lag, change multipath and bonding to be mutually exclusive
ice: Add missing E810 device ids
igc: Update I226_K device ID
e1000e: Fix packet loss on Tiger Lake and later
e1000e: Separate TGP board type from SPT
ptp: Fix possible memory leak in ptp_clock_register()
net: stmmac: Fix E2E delay mechanism
nfc: st95hf: Make spi remove() callback return zero
net: hns3: disable sriov before unload hclge layer
net: hns3: fix vf reset workqueue cannot exit
...
Pass a single argument to dsa_8021q_rx_vid and dsa_8021q_tx_vid that
contains the necessary information from the two arguments that are
currently provided: the switch and the port number.
Also rename those functions so that they have a dsa_port_* prefix, since
they operate on a struct dsa_port *.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Ever since Vivien's conversion of the ds->ports array into a dst->ports
list, and the introduction of dsa_to_port, iterations through the ports
of a switch became quadratic whenever dsa_to_port was needed.
dsa_to_port can either be called directly, or indirectly through the
dsa_is_{user,cpu,dsa,unused}_port helpers.
Use the newly introduced dsa_switch_for_each_port() iteration macro
that works with the iterator variable being a struct dsa_port *dp
directly, and not an int i. It is an expensive variable to go from i to
dp, but cheap to go from dp to i.
This macro iterates through the entire ds->dst->ports list and filters
by the ports belonging just to the switch provided as argument.
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>
Since the DSA conversion from the ds->ports array into the dst->ports
list, the DSA API has encouraged driver writers, as well as the core
itself, to write inefficient code.
Currently, code that wants to filter by a specific type of port when
iterating, like {!unused, user, cpu, dsa}, uses the dsa_is_*_port helper.
Under the hood, this uses dsa_to_port which iterates again through
dst->ports. But the driver iterates through the port list already, so
the complexity is quadratic for the typical case of a single-switch
tree.
This patch introduces some iteration helpers where the iterator is
already a struct dsa_port *dp, so that the other variant of the
filtering functions, dsa_port_is_{unused,user,cpu_dsa}, can be used
directly on the iterator. This eliminates the second lookup.
These functions can be used both by the core and by drivers.
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>
Now that we have a list of struct ocelot_bridge_vlan entries, we can
rewrite the pvid logic to simply point to one of those structures,
instead of having a separate structure with a "bool valid".
The NULL pointer will represent the lack of a bridge pvid (not to be
confused with the lack of a hardware pvid on the port, that is present
at all times).
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
At present, the ocelot driver accepts a single egress-untagged bridge
VLAN, meaning that this sequence of operations:
ip link add br0 type bridge vlan_filtering 1
ip link set swp0 master br0
bridge vlan add dev swp0 vid 2 pvid untagged
fails because the bridge automatically installs VID 1 as a pvid & untagged
VLAN, and vid 2 would be the second untagged VLAN on this port. It is
necessary to delete VID 1 before proceeding to add VID 2.
This limitation comes from the fact that we operate the port tag, when
it has an egress-untagged VID, in the OCELOT_PORT_TAG_NATIVE mode.
The ocelot switches do not have full flexibility and can either have one
single VID as egress-untagged, or all of them.
There are use cases for having all VLANs as egress-untagged as well, and
this patch adds support for that.
The change rewrites ocelot_port_set_native_vlan() into a more generic
ocelot_port_manage_port_tag() function. Because the software bridge's
state, transmitted to us via switchdev, can become very complex, we
don't attempt to track all possible state transitions, but instead take
a more declarative approach and just make ocelot_port_manage_port_tag()
figure out which more to operate in:
- port is VLAN-unaware: the classified VLAN (internal, unrelated to the
802.1Q header) is not inserted into packets on egress
- port is VLAN-aware:
- port has tagged VLANs:
-> port has no untagged VLAN: set up as pure trunk
-> port has one untagged VLAN: set up as trunk port + native VLAN
-> port has more than one untagged VLAN: this is an invalid config
which is rejected by ocelot_vlan_prepare
- port has no tagged VLANs
-> set up as pure egress-untagged port
We don't keep the number of tagged and untagged VLANs, we just count the
structures we keep.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
First and foremost, the driver currently allocates a constant sized
4K * u32 (16KB memory) array for the VLAN masks. However, a typical
application might not need so many VLANs, so if we dynamically allocate
the memory as needed, we might actually save some space.
Secondly, we'll need to keep more advanced bookkeeping of the VLANs we
have, notably we'll have to check how many untagged and how many tagged
VLANs we have. This will have to stay in a structure, and allocating
another 16 KB array for that is again a bit too much.
So refactor the bridge VLANs in a linked list of structures.
The hook points inside the driver are ocelot_vlan_member_add() and
ocelot_vlan_member_del(), which previously used to operate on the
ocelot->vlan_mask[vid] array element.
ocelot_vlan_member_add() and ocelot_vlan_member_del() used to call
ocelot_vlan_member_set() to commit to the ocelot->vlan_mask.
Additionally, we had two calls to ocelot_vlan_member_set() from outside
those callers, and those were directly from ocelot_vlan_init().
Those calls do not set up bridging service VLANs, instead they:
- clear the VLAN table on reset
- set the port pvid to the value used by this driver for VLAN-unaware
standalone port operation (VID 0)
So now, when we have a structure which represents actual bridge VLANs,
VID 0 doesn't belong in that structure, since it is not part of the
bridging layer.
So delete the middle man, ocelot_vlan_member_set(), and let
ocelot_vlan_init() call directly ocelot_vlant_set_mask() which forgoes
any data structure and writes directly to hardware, which is all that we
need.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is a cosmetic patch which clarifies what are the port tagging
options for Ocelot switches.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit e72aeb9ee0 ("fq_codel: implement L4S style ce_threshold_ect1
marking") expanded the ce_threshold feature of FQ-CoDel so it can
be applied to a subset of the traffic, using the ECT(1) bit of the ECN
field as the classifier. However, hard-coding ECT(1) as the only
classifier for this feature seems limiting, so let's expand it to be more
general.
To this end, change the parameter from a ce_threshold_ect1 boolean, to a
one-byte selector/mask pair (ce_threshold_{selector,mask}) which is applied
to the whole diffserv/ECN field in the IP header. This makes it possible to
classify packets by any value in either the ECN field or the diffserv
field. In particular, setting a selector of INET_ECN_ECT_1 and a mask of
INET_ECN_MASK corresponds to the functionality before this patch, and a
mask of ~INET_ECN_MASK allows using the selector as a straight-forward
match against a diffserv code point:
# apply ce_threshold to ECT(1) traffic
tc qdisc replace dev eth0 root fq_codel ce_threshold 1ms ce_threshold_selector 0x1/0x3
# apply ce_threshold to ECN-capable traffic marked as diffserv AF22
tc qdisc replace dev eth0 root fq_codel ce_threshold 1ms ce_threshold_selector 0x50/0xfc
Regardless of the selector chosen, the normal rules for ECN-marking of
packets still apply, i.e., the flow must still declare itself ECN-capable
by setting one of the bits in the ECN field to get marked at all.
v2:
- Add tc usage examples to patch description
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20211019174709.69081-1-toke@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Both multipath and bonding events are changing the HW LAG state
independently.
Handling one of the features events while the other is already
enabled can cause unwanted behavior, for example handling
bonding event while multipath enabled will disable the lag and
cause multipath to stop working.
Fix it by ignoring bonding event while in multipath and ignoring FIB
events while in bonding mode.
Fixes: 544fe7c2e6 ("net/mlx5e: Activate HW multipath and handle port affinity based on FIB events")
Signed-off-by: Maor Dickman <maord@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Pull tracing fix from Steven Rostedt:
"Recursion fix for tracing.
While cleaning up some of the tracing recursion protection logic, I
discovered a scenario that the current design would miss, and would
allow an infinite recursion. Removing an optimization trick that
opened the hole fixes the issue and cleans up the code as well"
* tag 'trace-v5.15-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
tracing: Have all levels of checks prevent recursion
