s390/qeth: Split memcpy() of struct qeth_ipacmd_addr_change flexible array
To work around a misbehavior of the compiler's ability to see into
composite flexible array structs (as detailed in the coming memcpy()
hardening series[1]), split the memcpy() of the header and the payload
so no false positive run-time overflow warning will be generated.
Pavel Begunkov [Mon, 26 Sep 2022 10:35:36 +0000 (11:35 +0100)]
selftests/net: enable io_uring sendzc testing
d8b6171bd58a5 ("selftests/io_uring: test zerocopy send") added io_uring
zerocopy tests but forgot to enable it in make runs. Add missing
io_uring_zerocopy_tx.sh into TEST_PROGS.
====================
devlink: fix order of port and netdev register in drivers
Some of the drivers use wrong order in registering devlink port and
netdev, registering netdev first. That was not intended as the devlink
port is some sort of parent for the netdev. Fix the ordering.
Note that the follow-up patchset is going to make this ordering
mandatory.
====================
NFC: hci: Split memcpy() of struct hcp_message flexible array
To work around a misbehavior of the compiler's ability to see into
composite flexible array structs (as detailed in the coming memcpy()
hardening series[1]), split the memcpy() of the header and the payload
so no false positive run-time overflow warning will be generated. This
split already existed for the "firstfrag" case, so just generalize the
logic further.
Cc: Eric Dumazet <edumazet@google.com> Cc: Paolo Abeni <pabeni@redhat.com> Reported-by: "Gustavo A. R. Silva" <gustavoars@kernel.org> Signed-off-by: Kees Cook <keescook@chromium.org> Reviewed-by: Gustavo A. R. Silva <gustavoars@kernel.org> Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> Link: https://lore.kernel.org/r/20220924040835.3364912-1-keescook@chromium.org Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Daniel Golle [Sun, 25 Sep 2022 14:48:43 +0000 (15:48 +0100)]
net: ethernet: mtk_eth_soc: fix usage of foe_entry_size
As sizeof(hwe->data) can now longer be used as the actual size depends
on foe_entry_size, in commit 9d8cb4c096ab02
("net: ethernet: mtk_eth_soc: add foe_entry_size to mtk_eth_soc") the
use of sizeof(hwe->data) is hence replaced.
However, replacing it with ppe->eth->soc->foe_entry_size is wrong as
foe_entry_size represents the size of the whole descriptor and not just
the 'data' field.
Fix this by subtracing the size of the only other field in the struct
'ib1', so we actually end up with the correct size to be copied to the
data field.
Reported-by: Chen Minqiang <ptpt52@gmail.com> Fixes: 9d8cb4c096ab02 ("net: ethernet: mtk_eth_soc: add foe_entry_size to mtk_eth_soc") Signed-off-by: Daniel Golle <daniel@makrotopia.org> Acked-by: Lorenzo Bianconi <lorenzo@kernel.org> Link: https://lore.kernel.org/r/YzBqPIgQR2gLrPoK@makrotopia.org Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Daniel Golle [Sun, 25 Sep 2022 14:47:20 +0000 (15:47 +0100)]
net: ethernet: mtk_eth_soc: fix wrong use of new helper function
In function mtk_foe_entry_set_vlan() the call to field accessor macro
FIELD_GET(MTK_FOE_IB1_BIND_VLAN_LAYER, entry->ib1)
has been wrongly replaced by
mtk_prep_ib1_vlan_layer(eth, entry->ib1)
Use correct helper function mtk_get_ib1_vlan_layer instead.
Reported-by: Chen Minqiang <ptpt52@gmail.com> Fixes: 03a3180e5c09e1 ("net: ethernet: mtk_eth_soc: introduce flow offloading support for mt7986") Signed-off-by: Daniel Golle <daniel@makrotopia.org> Acked-by: Lorenzo Bianconi <lorenzo@kernel.org> Link: https://lore.kernel.org/r/YzBp+Kk04CFDys4L@makrotopia.org Signed-off-by: Jakub Kicinski <kuba@kernel.org>
====================
net: openvswitch: metering and conntrack in userns
Currently using openvswitch in a non-initial user namespace, e.g., an
unprivileged container, is possible but without metering and conntrack
support. This is due to the restriction of the corresponding Netlink
interfaces to the global CAP_NET_ADMIN.
This simple patches switch from GENL_ADMIN_PERM to GENL_UNS_ADMIN_PERM
in several cases to allow this also for the unprivileged container
use case.
We tested this for unprivileged containers created by the container
manager of GyroidOS (gyroidos.github.io). However, for other container
managers such as LXC or systemd which provide unprivileged containers
this should be apply equally.
====================
Michael Weiß [Fri, 23 Sep 2022 13:38:20 +0000 (15:38 +0200)]
net: openvswitch: allow conntrack in non-initial user namespace
Similar to the previous commit, the Netlink interface of the OVS
conntrack module was restricted to global CAP_NET_ADMIN by using
GENL_ADMIN_PERM. This is changed to GENL_UNS_ADMIN_PERM to support
unprivileged containers in non-initial user namespace.
Signed-off-by: Michael Weiß <michael.weiss@aisec.fraunhofer.de> Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Michael Weiß [Fri, 23 Sep 2022 13:38:19 +0000 (15:38 +0200)]
net: openvswitch: allow metering in non-initial user namespace
The Netlink interface for metering was restricted to global CAP_NET_ADMIN
by using GENL_ADMIN_PERM. To allow metring in a non-inital user namespace,
e.g., a container, this is changed to GENL_UNS_ADMIN_PERM.
Signed-off-by: Michael Weiß <michael.weiss@aisec.fraunhofer.de> Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Tony Lu [Thu, 22 Sep 2022 12:19:07 +0000 (20:19 +0800)]
net/smc: Support SO_REUSEPORT
This enables SO_REUSEPORT [1] for clcsock when it is set on smc socket,
so that some applications which uses it can be transparently replaced
with SMC. Also, this helps improve load distribution.
Here is a simple test of NGINX + wrk with SMC. The CPU usage is collected
on NGINX (server) side as below.
====================
net: sunhme: Cleanups and logging improvements
This series is a continuation of [1] with a focus on logging improvements (in
the style of commit b11e5f6a3a5c ("net: sunhme: output link status with a single
print.")). I have included several of Rolf's patches in the series where
appropriate (with slight modifications). After this series is applied, many more
messages from this driver will come with driver/device information.
Additionally, most messages (especially debug messages) have been condensed onto
one line (as KERN_CONT messages get split).
Sean Anderson [Sat, 24 Sep 2022 01:53:38 +0000 (21:53 -0400)]
sunhme: Use vdbg for spam-y prints
The SXD, TXD, and RXD macros are used only once (or twice). Just use the
vdbg print, which seems to have been devised for these sorts of very
verbose messages.
Signed-off-by: Sean Anderson <seanga2@gmail.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Sean Anderson [Sat, 24 Sep 2022 01:53:37 +0000 (21:53 -0400)]
sunhme: Combine continued messages
This driver seems to have been written under the assumption that messages
can be continued arbitrarily. I'm not when this changed (if ever), but such
ad-hoc continuations are liable to be rudely interrupted. Convert all such
instances to single prints. This loses a bit of timing information (such as
when a line was constructed piecemeal as the function executed), but it's
easy to add a few prints if necessary. This also adds newlines to the ends
of any prints without them.
Since (almost every) debug print included the name of the function, include
it automatically.
Signed-off-by: Sean Anderson <seanga2@gmail.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Sean Anderson [Sat, 24 Sep 2022 01:53:36 +0000 (21:53 -0400)]
sunhme: Use (net)dev_foo wherever possible
Wherever possible, use the associated netdev (or device) when printing
errors or other messages. This makes it immediately clear what device
caused the error, and provides more information than just the device name.
Signed-off-by: Sean Anderson <seanga2@gmail.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Sean Anderson [Sat, 24 Sep 2022 01:53:35 +0000 (21:53 -0400)]
sunhme: Convert printk(KERN_FOO ...) to pr_foo(...)
This is a mostly-mechanical translation of the existing printks into
pr_foos. In several places, I have pasted messages which were broken over
several lines to allow for easier grepping.
Signed-off-by: Sean Anderson <seanga2@gmail.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Sean Anderson [Sat, 24 Sep 2022 01:53:34 +0000 (21:53 -0400)]
sunhme: Clean up debug infrastructure
Remove all the single-use debug conditionals, and just collect the debug
defines at the top of the file. HMD seems like it is used for general debug
info, so just redefine it as pr_debug. Additionally, instead of using the
default loglevel, use the debug loglevel for debugging.
Signed-off-by: Sean Anderson <seanga2@gmail.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Sean Anderson [Sat, 24 Sep 2022 01:53:30 +0000 (21:53 -0400)]
sunhme: Return an ERR_PTR from quattro_pci_find
In order to differentiate between a missing bridge and an OOM condition,
return ERR_PTRs from quattro_pci_find. This also does some general linting
in the area.
Signed-off-by: Sean Anderson <seanga2@gmail.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Sean Anderson [Sat, 24 Sep 2022 01:53:28 +0000 (21:53 -0400)]
sunhme: Remove version
Module versions are not very useful:
> The basic problem is, the version string does not identify the sources
> with enough accuracy. It says nothing about back ported fixes in
> stable kernels. It tells you nothing about vendor patches to the
> network core, etc.
This patchset https://lore.kernel.org/all/20220921140524.3831101-8-yangyingliang@huawei.com/T/
removed all set_drvdata(NULL) in driver remove function.
i2c_set_clientdata() is another wrapper of set drvdata function, to follow
the same convention, remove i2c_set_clientdata() called in driver remove
function in drivers/net/dsa/.
====================
xdp: Adjust xdp_frame layout to avoid using bitfields
Practical experience (and advice from Alexei) tell us that bitfields in
structs lead to un-optimized assembly code. I've verified this change
does lead to better x86_64 assembly, both via objdump and playing with
code snippets in godbolt.org.
Using scripts/bloat-o-meter shows the code size is reduced with 24
bytes for xdp_convert_buff_to_frame() that gets inlined e.g. in
i40e_xmit_xdp_tx_ring() which were used for microbenchmarking.
Microbenchmarking results do show improvements, but very small and
varying between 0.5 to 2 nanosec improvement per packet.
The member @metasize is changed from u8 to u32. Future users of this
area could split this into two u16 fields. I've also benchmarked with
two u16 fields showing equal performance gains and code size reduction.
The moved member @frame_sz doesn't change sizeof struct due to existing
padding. Like xdp_buff member @frame_sz is placed next to @flags, which
allows compiler to optimize assignment of these.
====================
Improve tsn_lib selftests for future distributed tasks
Some of the boards I am working with are limited in the number of ports
that they offer, and as more TSN related selftests are added, it is
important to be able to distribute the work among multiple boards.
A large part of implementing that is ensuring network-wide
synchronization, but also permitting more streams of data to flow
through the network. There is the more important aspect of also
coordinating the timing characteristics of those streams, and that is
also something that is tackled, although not in this modest patch set.
The goal here is not to introduce new selftests yet, but just to lay a
better foundation for them. These patches are a part of the cleanup work
I've done while working on selftests for frame preemption. They are
regression-tested with psfp.sh.
====================
Vladimir Oltean [Fri, 23 Sep 2022 21:00:15 +0000 (00:00 +0300)]
selftests: net: tsn_lib: run phc2sys in automatic mode
We can make the phc2sys helper not only synchronize a PHC to
CLOCK_REALTIME, which is what it currently does, but also CLOCK_REALTIME
to a PHC, which is going to be needed in distributed TSN tests.
Instead of making the complexity of the arguments passed to
phc2sys_start() explode, we can let it figure out the sync direction
automatically, based on ptp4l's port states.
Towards that goal, pass just the path to the desired ptp4l instance's
UNIX domain socket, and remove the $if_name argument (from which it
derives the PHC). Also adapt the one caller from the ocelot psfp.sh
test. In the case of psfp.sh, phc2sys_start is able to properly figure
out that CLOCK_REALTIME is the source clock and swp1's PHC is the
destination, because of the way in which ptp4l_start for the
UDS_ADDRESS_SWP1 was called: with slave_only=false, so it will always
win the BMCA and always become the sync master between itself and $h1.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Reviewed-by: Kurt Kanzenbach <kurt@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Move the PID variable for the isochron receiver into a separate
namespace per stats port, to allow multiple receivers (and/or
orchestration daemons) to be instantiated by the same script.
Preserve the existing behavior by making isochron_do() use the default
stats TCP port of 5000.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Reviewed-by: Kurt Kanzenbach <kurt@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Vladimir Oltean [Fri, 23 Sep 2022 21:00:13 +0000 (00:00 +0300)]
selftests: net: tsn_lib: allow running ptp4l on multiple interfaces
Switch ports will want to act as Boundary Clocks, which are configured
using ptp4l by specifying the "-i" argument multiple times.
Since we track a log file and a pid file for each ptp4l instance, and we
want to be compatible with the existing single-port callers of
ptp4l_start and ptp4l_stop, pass the interface list as a single string
of space-separated values. Based on this, we create a label for each
ptp4l instance, where the spaces are replaced with underscores
(ptp4l_start "eth0 eth1" generates "ptp4l_pid_eth0_eth1").
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Reviewed-by: Kurt Kanzenbach <kurt@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Vladimir Oltean [Fri, 23 Sep 2022 21:00:12 +0000 (00:00 +0300)]
selftests: net: tsn_lib: don't overwrite isochron receiver extra args with UDS
The extra_args argument ($3) of isochron_recv_start is overwritten with
uds ($2), if that argument exists.
This is currently not a problem, because the only TSN selftest
(ocelot/psfp.sh) omits remote sync so it does not specify to the
receiver a UNIX domain socket for ptp4l. So $uds is currently an empty
string.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Reviewed-by: Kurt Kanzenbach <kurt@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This patch adds missing MODULE_DEVICE_TABLE definition which generates
correct modalias for automatic loading of this driver when it is built
as an external module.
net: vertexcom: mse102x: Silence no spi_device_id warnings
SPI devices use the spi_device_id for module autoloading even on
systems using device tree, after commit 5fa6863ba692 ("spi: Check
we have a spi_device_id for each DT compatible"), kernel warns as
follows since the spi_device_id is missing:
SPI driver mse102x has no spi_device_id for vertexcom,mse1021
SPI driver mse102x has no spi_device_id for vertexcom,mse1022
Add spi_device_id entries to silence the warnings, and ensure driver
module autoloading works.
net: ethernet: adi: Fix return value check in adin1110_probe_netdevs()
In case of error, the function get_phy_device() returns ERR_PTR()
and never returns NULL. The NULL test in the return value check
should be replaced with IS_ERR().
====================
net: dsa: microchip: ksz9477: enable interrupt for internal phy link detection
This patch series implements the common interrupt handling for ksz9477 based
switches and lan937x. The ksz9477 and lan937x has similar interrupt registers
except ksz9477 has 4 port based interrupts whereas lan937x has 6 interrupts.
The patch moves the phy interrupt hanler implemented in lan937x_main.c to
ksz_common.c, along with the mdio_register functionality.
====================
net: dsa: microchip: use common irq routines for girq and pirq
The global port interrupt routines and individual ports interrupt
routines has similar implementation except the mask & status register
and number of nested irqs in them. The mask & status register and
pointer to ksz_device is added to ksz_irq and uses the ksz_irq as
irq_chip_data.
Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
net: dsa: microchip: move interrupt handling logic from lan937x to ksz_common
To support the phy link detection through interrupt method for ksz9477
based switch, the interrupt handling routines are moved from
lan937x_main.c to ksz_common.c. The only changes made are functions
names are prefixed with ksz_ instead of lan937x_.
Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
net: dsa: microchip: lan937x: return zero if mdio node not present
Currently, if the mdio node is not present in the dts file then
lan937x_mdio_register return -ENODEV and entire probing process fails.
To make the mdio_register generic for all ksz series switches and to
maintain back-compatibility with existing dts file, return -ENODEV is
replaced with return 0.
Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
net: dsa: microchip: enable phy interrupts only if interrupt enabled in dts
In the lan937x_mdio_register function, phy interrupts are enabled
irrespective of irq is enabled in the switch. Now, the check is added to
enable the phy interrupt only if the irq is enabled in the switch.
Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
net: dsa: microchip: determine number of port irq based on switch type
Currently the number of port irqs is hard coded for the lan937x switch
as 6. In order to make the generic interrupt handler for ksz switches,
number of port irq supported by the switch is added to the
ksz_chip_data. It is 4 for ksz9477, 2 for ksz9897 and 3 for ksz9567.
Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Vladimir Oltean [Fri, 23 Sep 2022 14:59:21 +0000 (17:59 +0300)]
net/sched: taprio: simplify list iteration in taprio_dev_notifier()
taprio_dev_notifier() subscribes to netdev state changes in order to
determine whether interfaces which have a taprio root qdisc have changed
their link speed, so the internal calculations can be adapted properly.
The 'qdev' temporary variable serves no purpose, because we just use it
only once, and can just as well use qdisc_dev(q->root) directly (or the
"dev" that comes from the netdev notifier; this is because qdev is only
interesting if it was the subject of the state change, _and_ its root
qdisc belongs in the taprio list).
The 'found' variable also doesn't really serve too much of a purpose
either; we can just call taprio_set_picos_per_byte() within the loop,
and exit immediately afterwards.
Vladimir Oltean [Wed, 21 Sep 2022 18:54:28 +0000 (21:54 +0300)]
net: dsa: make user ports return to init_net on netns deletion
As pointed out during review, currently the following set of commands
crashes the kernel:
$ ip netns add ns0
$ ip link set swp0 netns ns0
$ ip netns del ns0
WARNING: CPU: 1 PID: 27 at net/core/dev.c:10884 unregister_netdevice_many+0xaa4/0xaec
Workqueue: netns cleanup_net
pstate: 20000005 (nzCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
pc : unregister_netdevice_many+0xaa4/0xaec
lr : unregister_netdevice_many+0x700/0xaec
Call trace:
unregister_netdevice_many+0xaa4/0xaec
default_device_exit_batch+0x294/0x340
ops_exit_list+0xac/0xc4
cleanup_net+0x2e4/0x544
process_one_work+0x4ec/0xb40
---[ end trace 0000000000000000 ]---
unregister_netdevice: waiting for swp0 to become free. Usage count = 2
This is because since DSA user ports, since they started populating
dev->rtnl_link_ops in the blamed commit, gained a different treatment
from default_device_exit_net(), which thinks these interfaces can now be
unregistered.
They can't; so set netns_refund = true to restore the behavior prior to
populating dev->rtnl_link_ops.
Fixes: 95f510d0b792 ("net: dsa: allow the DSA master to be seen and changed through rtnetlink") Suggested-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Link: https://lore.kernel.org/r/20220921185428.1767001-1-vladimir.oltean@nxp.com Signed-off-by: Jakub Kicinski <kuba@kernel.org>
During LPC2022 I meetup with my page_pool co-maintainer Ilias. When
discussing page_pool code we realised/remembered certain optimizations
had not been fully utilised.
Since commit c07aea3ef4d4 ("mm: add a signature in struct page") struct
page have a direct pointer to the page_pool object this page was
allocated from.
Thus, with this info it is possible to skip the rhashtable_lookup to
find the page_pool object in __xdp_return().
The rcu_read_lock can be removed as it was tied to xdp_mem_allocator.
The page_pool object is still safe to access as it tracks inflight pages
and (potentially) schedules final release from a work queue.
Created a micro benchmark of XDP redirecting from mlx5 into veth with
XDP_DROP bpf-prog on the peer veth device. This increased performance
6.5% from approx 8.45Mpps to 9Mpps corresponding to using 7 nanosec
(27 cycles at 3.8GHz) less per packet.
Peilin Ye [Fri, 23 Sep 2022 04:59:26 +0000 (21:59 -0700)]
af_unix: Refactor unix_read_skb()
Similar to udp_read_skb(), delete the unnecessary while loop in
unix_read_skb() for readability. Since recv_actor() cannot return a
value greater than skb->len (see sk_psock_verdict_recv()), remove the
redundant check.
Peilin Ye [Fri, 23 Sep 2022 04:59:13 +0000 (21:59 -0700)]
udp: Refactor udp_read_skb()
Delete the unnecessary while loop in udp_read_skb() for readability.
Additionally, since recv_actor() cannot return a value greater than
skb->len (see sk_psock_verdict_recv()), remove the redundant check.
Commit 0a28bfd4971f added a metadata_dst to each tx_sc, but that's
only allocated when macsec_add_dev has run, which happens after device
registration. If the requested or computed SCI already exists, or if
linking to the lower device fails, we will panic because
metadata_dst_free can't handle NULL.
Reproducer:
ip link add link $lower type macsec
ip link add link $lower type macsec
Jakub Kicinski [Sat, 24 Sep 2022 03:56:31 +0000 (20:56 -0700)]
Merge branch 'net-ipa-another-set-of-cleanups'
Alex Elder says:
====================
net: ipa: another set of cleanups
This series contains another set of cleanups done in preparation for
an upcoming series that reworks how IPA registers and their fields
are defined.
The first replaces the use of u32_replace_bits() with a simple
logical AND operation in two places.
The second creates a new function to encapsulate some common code,
and renames another for consistency. The third restructures two
other functions that do similar things to make their similarity more
obvious.
The fourth defines the flag bits in a register using an enumerated
type. And the fifth updates "ipa_reg.h" so the values assigned to
enumerated type members are aligned consistently.
The last three encapsulate the code that assigns values to a few
registers into separate functions.
====================
Alex Elder [Thu, 22 Sep 2022 22:20:59 +0000 (17:20 -0500)]
net: ipa: encapsulate updating the COUNTER_CFG register
Create a new function that encapsulates setting the counter
configuration register value for versions prior to IPA v4.5.
Create another small function to represent configuring hardware
timing regardless of version.
Signed-off-by: Alex Elder <elder@linaro.org> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Alex Elder [Thu, 22 Sep 2022 22:20:57 +0000 (17:20 -0500)]
net: ipa: tidy up register enum definitions
Update a few enumerated type definitions in "ipa_reg.h" so that the
values assigned to each member align on the same column. Where a
"TX" or "RX" (or both) comment is present, move that annotation into
a separate comment between the member name and its value.
Signed-off-by: Alex Elder <elder@linaro.org> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Alex Elder [Thu, 22 Sep 2022 22:20:56 +0000 (17:20 -0500)]
net: ipa: define BCR values using an enum
The backward compatibility register (BCR) has a set of bit flags
that indicate ways in which the IPA hardware should operate in a
backward compatible way. The register is not supported starting
with IPA v4.5, and where it is supported, defined bits all have the
same numeric value.
Redefine these flags using an enumerated type, with each member's
value representing the bit position that encodes it in the BCR.
This replaces all of the single-bit field masks previously defined.
Signed-off-by: Alex Elder <elder@linaro.org> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Alex Elder [Thu, 22 Sep 2022 22:20:55 +0000 (17:20 -0500)]
net: ipa: rearrange functions for similarity
Both aggr_time_limit_encode() and hol_block_timer_encode() figure
out how to encode a millisecond time value so it can be programmed
into a register. Rearranging them a bit can make their similarity
more obvious, with both taking essentially the same form.
To do this:
- Return 0 immediately in aggr_time_limit_encode() if the
microseconds value supplied is zero.
- Reverse the test at top of aggr_time_limit_encode(), so we
compute and return the Qtime value in the "true" block,
and compute the result the old way otherwise.
- Open-code (and eliminate) hol_block_timer_qtime_encode() at the
top of hol_block_timer_encode() in the case we use Qtimer.
Signed-off-by: Alex Elder <elder@linaro.org> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Alex Elder [Thu, 22 Sep 2022 22:20:54 +0000 (17:20 -0500)]
net: ipa: introduce ipa_qtime_val()
Create a new function ipa_qtime_val() which returns a value that
indicates what should be encoded for a register with a time field
expressed using Qtime. Use it to factor out common code in
aggr_time_limit_encoded() and hol_block_timer_qtime_val().
Rename aggr_time_limit_encoded() and hol_block_timer_qtime_val() so
their names are both verbs ending in "encode". Rename the "limit"
argument to the former to be "milliseconds" for consistency.
Signed-off-by: Alex Elder <elder@linaro.org> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Alex Elder [Thu, 22 Sep 2022 22:20:53 +0000 (17:20 -0500)]
net: ipa: don't use u32p_replace_bits()
In two spots we use u32_replace_bits() to replace a set of bits in a
register while preserving the rest. Both of those cases just zero
the bits being replaced, and this can be done more simply without
using that function.
Signed-off-by: Alex Elder <elder@linaro.org> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The first 2 patches are by Ziyang Xuan and optimize registration and
the sending in the CAN BCM protocol a bit.
The next 8 patches target the gs_usb driver. 7 are by me and first fix
the time hardware stamping support (added during this net-next cycle),
rename a variable, convert the usb_control_msg + manual
kmalloc()/kfree() to usb_control_msg_{send,rev}(), clean up the error
handling and add switchable termination support. The patch by Rhett
Aultman and Vasanth Sadhasivan convert the driver from
usb_alloc_coherent()/usb_free_coherent() to kmalloc()/URB_FREE_BUFFER.
The last patch is by Shang XiaoJing and removes an unneeded call to
dev_err() from the ctucanfd driver.
* tag 'linux-can-next-for-6.1-20220923' of git://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can-next:
can: ctucanfd: Remove redundant dev_err call
can: gs_usb: remove dma allocations
can: gs_usb: add switchable termination support
can: gs_usb: gs_make_candev(): clean up error handling
can: gs_usb: convert from usb_control_msg() to usb_control_msg_{send,recv}()
can: gs_usb: gs_cmd_reset(): rename variable holding struct gs_can pointer to dev
can: gs_usb: gs_can_open(): initialize time counter before starting device
can: gs_usb: add missing lock to protect struct timecounter::cycle_last
can: gs_usb: gs_usb_get_timestamp(): fix endpoint parameter for usb_control_msg_recv()
can: bcm: check the result of can_send() in bcm_can_tx()
can: bcm: registration process optimization in bcm_module_init()
====================
====================
net: macsec: remove the preparation phase when offloading operations
It was reported[1] the 2-step phase offloading of MACsec operations did
not fit well and device drivers were mostly ignoring the first phase
(preparation). In addition the s/w fallback in case h/w rejected an
operation, which could have taken advantage of this design, never was
implemented and it's probably not a good idea anyway (at least
unconditionnally). So let's remove this logic which only makes the code
more complex for no advantage, before there are too many drivers
providing MACsec offloading.
This series removes the first phase (preparation) of the MACsec h/w
offloading. The modifications are split per-driver and in a way that
makes bissection working with logical steps; but I can squash some
patches if needed.
This was tested on the MSCC PHY but not on the Altantic nor mlx5e NICs.
Antoine Tenart [Wed, 21 Sep 2022 13:51:18 +0000 (15:51 +0200)]
net: macsec: remove the prepare flag from the MACsec offloading context
Now that the MACsec offloading preparation phase was removed from the
MACsec core implementation as well as from drivers implementing it, we
can safely remove the flag representing it.
Signed-off-by: Antoine Tenart <atenart@kernel.org> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Antoine Tenart [Wed, 21 Sep 2022 13:51:14 +0000 (15:51 +0200)]
net: macsec: remove the prepare phase when offloading
The hardware offloading in MACsec was initially supported using 2 phases.
This was proposed in the RFC as this could have allowed easier fallback
to the software implementation if the hardware did not support a feature
or had enough entries already. But this fallback wasn't implemented and
might not be a good idea after all. In addition it turned out this logic
didn't mapped well the hardware logic and device drivers were mostly
ignoring the preparation phase.
Let's remove this as it does not offer any advantage and is ignored by
drivers.
Signed-off-by: Antoine Tenart <atenart@kernel.org> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
It looks like this test has been accidentally dropped when resolving
conflicts in this Makefile.
Most probably because there were 3 different patches modifying this file
in parallel:
commit 152e8ec77640 ("selftests/bonding: add a test for bonding lladdr target")
commit bbb774d921e2 ("net: Add tests for bonding and team address list management")
commit 2ffd57327ff1 ("selftests: bonding: cause oops in bond_rr_gen_slave_id")
The first one was applied in 'net-next' while the two other ones were
recently applied in the 'net' tree.
But that's alright, easy to fix by re-adding the missing one!
DMA allocated buffers are a precious resource. If there is no need for
DMA allocations, then it might be worth to use non-dma allocated
buffers.
After testing the gs_usb driver with and without DMA allocation, there
does not seem to be a significant change in latency or CPU utilization
either way. Therefore, DMA allocation is not necessary and removed.
Internal buffers used within urbs were managed and freed manually.
These buffers are no longer needed to be managed by the driver. The
URB_FREE_BUFFER flag, allows for the buffers in question to be
automatically freed.
The candleLight community is working on switchable termination support
for the candleLight firmware. As the the Linux CAN framework supports
switchable termination add this feature to the gs_usb driver.
Devices supporting the feature should set the
GS_CAN_FEATURE_TERMINATION and implement the
GS_USB_BREQ_SET_TERMINATION and GS_USB_BREQ_GET_TERMINATION control
messages.
For now the driver assumes for activated termination the standard
termination value of 120Ω.
can: gs_usb: gs_make_candev(): clean up error handling
Introduce a label to free the allocated candev in case of an error and
make use of if. Fix a memory leak if the extended bit timing cannot be
read. Extend the error messages to print the number of the failing
channel and the symbolic error name.
can: gs_usb: convert from usb_control_msg() to usb_control_msg_{send,recv}()
Convert the driver to use usb_control_msg_{send,recv}() instead of
usb_control_msg(). These functions allow the data to be placed on the
stack. This makes the driver a lot easier as we don't have to deal
with dynamically allocated memory.
can: gs_usb: gs_can_open(): initialize time counter before starting device
On busy networks the CAN controller might receive CAN frames directly
after starting it but before the timecounter is setup. This will lead
to NULL pointer deref while converting the converting the CAN frame's
timestamp with the timecounter.
Close the race window by setting up the timecounter before starting
the CAN controller.
Fixes: 45dfa45f52e6 ("can: gs_usb: add RX and TX hardware timestamp support") Link: https://lore.kernel.org/all/20220921081329.385509-1-mkl@pengutronix.de Cc: John Whittington <git@jbrengineering.co.uk Tested-by: John Whittington <git@jbrengineering.co.uk> Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
can: gs_usb: add missing lock to protect struct timecounter::cycle_last
The struct timecounter::cycle_last is a 64 bit variable, read by
timecounter_cyc2time(), and written by timecounter_read(). On 32 bit
architectures this is not atomic.
Add a spinlock to protect access to struct timecounter::cycle_last. In
the gs_usb_timestamp_read() callback the lock is dropped to execute a
sleeping synchronous USB transfer. This is safe, as the variable we
want to protect is accessed during this call.
Fixes: 45dfa45f52e6 ("can: gs_usb: add RX and TX hardware timestamp support") Link: https://lore.kernel.org/all/20220920100416.959226-3-mkl@pengutronix.de Cc: John Whittington <git@jbrengineering.co.uk> Tested-by: John Whittington <git@jbrengineering.co.uk> Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
can: gs_usb: gs_usb_get_timestamp(): fix endpoint parameter for usb_control_msg_recv()
The 2nd argument of usb_control_msg_recv() is the "endpoint",
usb_control_msg_recv() will internally convert the endpoint into a
pipe with usb_rcvctrlpipe().
In gs_usb_get_timestamp() not the endpoint "0" is passed, but the
pipe. This worked by accident as endpoint is a __u8 and the lowest 8
bits of the pipe are 0. Fix this copy/paste error by using the correct
endpoint of "0".
Fixes: 45dfa45f52e6 ("can: gs_usb: add RX and TX hardware timestamp support") Link: https://lore.kernel.org/all/20220920100416.959226-2-mkl@pengutronix.de Cc: John Whittington <git@jbrengineering.co.uk> Tested-by: John Whittington <git@jbrengineering.co.uk> Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Ziyang Xuan [Thu, 15 Sep 2022 01:55:55 +0000 (09:55 +0800)]
can: bcm: registration process optimization in bcm_module_init()
Now, register_netdevice_notifier() and register_pernet_subsys() are both
after can_proto_register(). It can create CAN_BCM socket and process socket
once can_proto_register() successfully, so it is possible missing notifier
event or proc node creation because notifier or bcm proc directory is not
registered or created yet. Although this is a low probability scenario, it
is not impossible.
Move register_pernet_subsys() and register_netdevice_notifier() to the
front of can_proto_register(). In addition, register_pernet_subsys() and
register_netdevice_notifier() may fail, check their results are necessary.
Lan966x switch supports time-based egress shaping in hardware
according to IEEE 802.1Qbv. Add support for TAS configuration on
egress port of lan966x switch.
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com> Signed-off-by: David S. Miller <davem@davemloft.net>