With the ksz_chip_id enums moved to the platform include file for ksz
switches, platform code that instantiates a device can now use these to
set ksz_platform_data::chip_id.
Signed-off-by: Daniel Danzberger <dd@embedd.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
The ksz driver has bits and pieces of platform_data probing support, but
it doesn't work.
The conventional thing to do is to have an encapsulating structure for
struct dsa_chip_data that gets put into dev->platform_data. This driver
expects a struct ksz_platform_data, but that doesn't contain a struct
dsa_chip_data as first element, which will obviously not work with
dsa_switch_probe() -> dsa_switch_parse().
Pointing dev->platform_data to a struct dsa_chip_data directly is in
principle possible, but that doesn't work either. The driver has
ksz_switch_detect() to read the device ID from hardware, followed by
ksz_check_device_id() to compare it against a predetermined expected
value. This protects against early errors in the SPI/I2C communication.
With platform_data, the mechanism in ksz_check_device_id() doesn't work
and even leads to NULL pointer dereferences, since of_device_get_match_data()
doesn't work in that probe path.
So obviously, the platform_data support is actually missing, and the
existing handling of struct ksz_platform_data is bogus. Complete the
support by adding a struct dsa_chip_data as first element, and fixing up
ksz_check_device_id() to pick up the platform_data instead of the
unavailable of_device_get_match_data().
The early dev->chip_id assignment from ksz_switch_register() is also
bogus, because ksz_switch_detect() sets it to an initial value. So
remove it.
Also, ksz_platform_data :: enabled_ports isn't used anywhere, delete it.
Link: https://lore.kernel.org/netdev/20231204154315.3906267-1-dd@embedd.com/
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Daniel Danzberger <dd@embedd.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Ante Knezic says:
====================
net: dsa: microchip: enable setting rmii reference
KSZ88X3 devices can select between internal and external RMII reference clock.
This patch series introduces new device tree property for setting reference
clock to internal.
---
V5:
- move rmii-clk-internal to be a port device tree property.
V4:
- remove rmii_clk_internal from ksz_device, as its not needed any more
- move rmii clk config as well as ksz8795_cpu_interface_select to
ksz8_config_cpu_port
V3:
- move ksz_cfg from global switch config to port config as suggested by Vladimir
Oltean
- reverse patch order as suggested by Vladimir Oltean
- adapt dt schema as suggested by Conor Dooley
V2:
- don't rely on default register settings - enforce set/clear property as
suggested by Andrew Lunn
- enforce dt schema as suggested by Conor Dooley
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Microchip KSZ8863/KSZ8873 have the ability to select between internal
and external RMII reference clock. By default, reference clock
needs to be provided via REFCLKI_3 pin. If required, device can be
setup to provide RMII clock internally so that REFCLKI_3 pin can be
left unconnected.
Add a new "microchip,rmii-clk-internal" property which will set
RMII clock reference to internal. If property is not set, reference
clock needs to be provided externally.
While at it, move the ksz8795_cpu_interface_select() to
ksz8_config_cpu_port() to get a cleaner call path for cpu port.
Signed-off-by: Ante Knezic <ante.knezic@helmholz.de>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Due to linkwatch_forget_dev() (and perhaps others?) checking for
list_empty(&dev->link_watch_list), we must have all manipulations
of even the local on-stack list 'wrk' here under spinlock, since
even that list can be reached otherwise via dev->link_watch_list.
This is already the case, but makes this a bit counter-intuitive,
often local lists are used to _not_ have to use locking for their
local use.
Remove the local list as it doesn't seem to serve any purpose.
While at it, move a variable declaration into the loop using it.
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://lore.kernel.org/r/20231205170011.56576dcc1727.I698b72219d9f6ce789bd209b8f6dffd0ca32a8f2@changeid
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently the checks are:
[1] unlikely(q->features & IONIC_TXQ_F_HWSTAMP)
[2] !unlikely(q->features & IONIC_TXQ_F_HWSTAMP)
[1] is clear enough, but [2] isn't exactly obvious to the
reader because !unlikely reads as likely. However, that's
not what this means.
[2] means that it's unlikely that the q has
IONIC_TXQ_F_HWSTAMP enabled.
Write an inline helper function to hide the unlikely
optimization to make these checks more readable.
Signed-off-by: Brett Creeley <brett.creeley@amd.com>
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Rahul Rameshbabu <rrameshbabu@nvidia.com>
Link: https://lore.kernel.org/r/20231204210936.16587-5-shannon.nelson@amd.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Each time a VF attribute is set via iproute a call to get the VF
configuration is also made. This is currently problematic because for
each VF configuration call there are multiple commands sent to the
device. Unfortunately, this doesn't scale well. Fix this by reporting
the cached VF attributes.
The original change to query the device for getting the VF attributes
f16f5be310 ("ionic: Query FW when getting VF info via ndo_get_vf_config")
was made to remain consistent with device set VF attributes. However,
after further investigation there is no need to query the device.
Signed-off-by: Brett Creeley <brett.creeley@amd.com>
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Rahul Rameshbabu <rrameshbabu@nvidia.com>
Link: https://lore.kernel.org/r/20231204210936.16587-2-shannon.nelson@amd.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The variables are organized according in the following way:
- TX read-mostly hotpath cache lines
- TXRX read-mostly hotpath cache lines
- RX read-mostly hotpath cache lines
- TX read-write hotpath cache line
- TXRX read-write hotpath cache line
- RX read-write hotpath cache line
Fastpath cachelines end after rcvq_space.
Cache line boundaries are enforced only between read-mostly and
read-write. That is, if read-mostly tx cachelines bleed into
read-mostly txrx cachelines, we do not care. We care about the
boundaries between read and write cachelines because we want
to prevent false sharing.
Fast path variables span cache lines before change: 12
Fast path variables span cache lines after change: 8
Suggested-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Wei Wang <weiwan@google.com>
Signed-off-by: Coco Li <lixiaoyan@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20231204201232.520025-3-lixiaoyan@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Inquire firmware on supported offloads, as well as convey offloads
enabled dynamically to firmware. New control net API functionality is
required for the above. Implement control net API framework for
offloads.
Additionally, fetch/insert offload metadata from hardware RX/TX
buffer respectively during receive/transmit.
Currently supported offloads include checksum and TSO.
Signed-off-by: Shinas Rasheed <srasheed@marvell.com>
Link: https://lore.kernel.org/r/20231204154940.2583140-1-srasheed@marvell.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Tobias Waldekranz says:
====================
net: mvmdio: Performance related improvements
Observations of the XMDIO bus on a CN9130-based system during a
firmware download showed a very low bus utilization, which stemmed
from the 150us (10x the average access time) sleep which would take
place when the first poll did not succeed.
With this series in place, bus throughput increases by about 10x,
multiplied by whatever gain you are able to extract from running the
MDC at a higher frequency (hardware dependent).
====================
Link: https://lore.kernel.org/r/20231204100811.2708884-1-tobias@waldekranz.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Before this change, when operating in polled mode, i.e. no IRQ is
available, every individual C45 access would be hit with a 150us sleep
after the bus access.
For example, on a board with a CN9130 SoC connected to an MV88X3310
PHY, a single C45 read would take around 165us:
root@infix:~$ mdio f212a600.mdio-mii mmd 4:1 bench 0xc003
Performed 1000 reads in 165ms
By replacing the long sleep with a tighter poll loop, we observe a 10x
increase in bus throughput:
root@infix:~$ mdio f212a600.mdio-mii mmd 4:1 bench 0xc003
Performed 1000 reads in 15ms
Signed-off-by: Tobias Waldekranz <tobias@waldekranz.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Tested-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20231204100811.2708884-3-tobias@waldekranz.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
ynl.h has a growing amount of "internal" stuff, which may confuse
users who try to take a look at the external API. Currently the
internals are at the bottom of the file with a banner in between,
but this arrangement makes it hard to add external APIs / inline
helpers which need internal definitions.
Move internals to a separate header.
Link: https://lore.kernel.org/r/20231202211225.342466-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Commit 1768d8a767 ("tools/net/ynl: Add support for create flags")
added support for setting legacy netlink CRUD flags on netlink
messages (NLM_F_REPLACE, _EXCL, _CREATE etc.).
Most of genetlink won't need these, don't force callers to pass
in an empty argument to each do() call.
Reviewed-by: Donald Hunter <donald.hunter@gmail.com>
Link: https://lore.kernel.org/r/20231202211005.341613-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is ignored (apart
from emitting a warning) and this typically results in resource leaks.
To improve here there is a quest to make the remove callback return
void. In the first step of this quest all drivers are converted to
.remove_new(), which already returns void. Eventually after all drivers
are converted, .remove_new() will be renamed to .remove().
Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.
Link: https://lore.kernel.org/r/20231117095922.876489-7-u.kleine-koenig@pengutronix.de
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/8c9ffca75ea24810f9ba05a514d5ad59847cc4fe.1701713943.git.u.kleine-koenig@pengutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is ignored (apart
from emitting a warning) and this typically results in resource leaks.
To improve here there is a quest to make the remove callback return
void. In the first step of this quest all drivers are converted to
.remove_new(), which already returns void. Eventually after all drivers
are converted, .remove_new() will be renamed to .remove().
Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.
Link: https://lore.kernel.org/r/20231117095922.876489-4-u.kleine-koenig@pengutronix.de
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/4889ac6a7ffa9b02fa5cdd2d3212e739741f80b8.1701713943.git.u.kleine-koenig@pengutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is ignored (apart
from emitting a warning) and this typically results in resource leaks.
To improve here there is a quest to make the remove callback return
void. In the first step of this quest all drivers are converted to
.remove_new(), which already returns void. Eventually after all drivers
are converted, .remove_new() will be renamed to .remove().
Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.
Link: https://lore.kernel.org/r/20231117095922.876489-3-u.kleine-koenig@pengutronix.de
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/c43193b9a002e88da36b111bb44ce2973ecde722.1701713943.git.u.kleine-koenig@pengutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This function exceeds the stack frame warning limit:
drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_debugfs.c: In function 'hclge_dbg_dump_tm_pri':
drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_debugfs.c:1039:1: error: the frame size of 1408 bytes is larger than 1024 bytes [-Werror=frame-larger-than=]
Use dynamic allocation for the largest stack object instead. It
would be nice to rewrite this file to completely avoid the extra
buffer and just use the one that was already allocated by debugfs,
but that is a much larger change.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Link: https://lore.kernel.org/r/20231204085735.4112882-1-arnd@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The ynl-generated user space C code is already above 25kLoC
and is growing.
The initial reason to commit these files was to make reviewing changes
to the generator easier. Unfortunately, it has the opposite effect on
reviewing changes to specs, and we get far more changes to specs
than to the generator.
Uncommit those fails, as they are generated on the fly as needed.
netdev patchwork now runs a script on each series to create a diff
of generated code on the fly, for the rare cases when looking at
it is helpful:
https://github.com/kuba-moo/nipa/blob/master/tests/series/ynl/ynl.sh
Suggested-by: Paolo Abeni <pabeni@redhat.com>
Acked-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The .parse_protocol hook function in the ipvlan_header_ops structure is
not implemented. As a result, when the AF_PACKET family is used to send
packets, skb->protocol will be set to 0.
Ipvlan is a device of type ARPHRD_ETHER (ether_setup). Therefore, use
eth_header_parse_protocol function to obtain the protocol.
Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Link: https://lore.kernel.org/r/20231202130438.2266343-1-shaozhengchao@huawei.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Hangbin Liu says:
====================
Conver net selftests to run in unique namespace (Part 1)
As Guillaume pointed, many selftests create namespaces with very common
names (like "client" or "server") or even (partially) run directly in init_net.
This makes these tests prone to failure if another namespace with the same
name already exists. It also makes it impossible to run several instances
of these tests in parallel.
This patch set intend to conver all the net selftests to run in unique namespace,
so we can update the selftest freamwork to run all tests in it's own namespace
in parallel. After update, we only need to wait for the test which need
longest time.
As the total patch set is too large. I break it to severl parts. This is
the first part.
v2 -> v3:
- Convert all ip netns del to cleanup_ns (Justin Iurman)
v1 -> v2:
- Split the large patch set to small parts for easy review (Paolo Abeni)
- Move busywait from forwarding/lib.sh to net/lib.sh directly (Petr Machata)
- Update setup_ns/cleanup_ns struct (Petr Machata)
- Remove default trap in lib.sh (Petr Machata)
====================
Link: https://lore.kernel.org/r/20231202020110.362433-1-liuhangbin@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Here is the test result after conversion.
# ./unicast_extensions.sh
/usr/bin/which: no nettest in (/root/.local/bin:/root/bin:/usr/share/Modules/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin)
###########################################################################
Unicast address extensions tests (behavior of reserved IPv4 addresses)
###########################################################################
TEST: assign and ping within 240/4 (1 of 2) (is allowed) [ OK ]
TEST: assign and ping within 240/4 (2 of 2) (is allowed) [ OK ]
TEST: assign and ping within 0/8 (1 of 2) (is allowed) [ OK ]
...
TEST: assign and ping class D address (is forbidden) [ OK ]
TEST: routing using class D (is forbidden) [ OK ]
TEST: routing using 127/8 (is forbidden) [ OK ]
Acked-by: David Ahern <dsahern@kernel.org>
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Here is the test result after conversion.
]# ./sctp_vrf.sh
Testing For SCTP VRF:
TEST 01: nobind, connect from client 1, l3mdev_accept=1, Y [PASS]
...
TEST 12: bind vrf-2 & 1 in server, connect from client 1 & 2, N [PASS]
***v6 Tests Done***
Acked-by: David Ahern <dsahern@kernel.org>
Reviewed-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Here is the test result after conversion.
]# ./l2tp.sh
TEST: IPv4 basic L2TP tunnel [ OK ]
TEST: IPv4 route through L2TP tunnel [ OK ]
TEST: IPv6 basic L2TP tunnel [ OK ]
TEST: IPv6 route through L2TP tunnel [ OK ]
TEST: IPv4 basic L2TP tunnel - with IPsec [ OK ]
TEST: IPv4 route through L2TP tunnel - with IPsec [ OK ]
TEST: IPv6 basic L2TP tunnel - with IPsec [ OK ]
TEST: IPv6 route through L2TP tunnel - with IPsec [ OK ]
TEST: IPv4 basic L2TP tunnel [ OK ]
TEST: IPv4 route through L2TP tunnel [ OK ]
TEST: IPv6 basic L2TP tunnel - with IPsec [ OK ]
TEST: IPv6 route through L2TP tunnel - with IPsec [ OK ]
TEST: IPv4 basic L2TP tunnel - after IPsec teardown [ OK ]
TEST: IPv4 route through L2TP tunnel - after IPsec teardown [ OK ]
TEST: IPv6 basic L2TP tunnel - after IPsec teardown [ OK ]
TEST: IPv6 route through L2TP tunnel - after IPsec teardown [ OK ]
Tests passed: 16
Tests failed: 0
Acked-by: David Ahern <dsahern@kernel.org>
Reviewed-by: James Chapman <jchapman@katalix.com>
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Here is the test result after conversion.
]# ./ioam6.sh
--------------------------------------------------------------------------
OUTPUT tests
--------------------------------------------------------------------------
TEST: Unknown IOAM namespace (inline mode) [ OK ]
TEST: Unknown IOAM namespace (encap mode) [ OK ]
TEST: Missing trace room (inline mode) [ OK ]
TEST: Missing trace room (encap mode) [ OK ]
TEST: Trace type with bit 0 only (inline mode) [ OK ]
...
TEST: Full supported trace (encap mode) [ OK ]
--------------------------------------------------------------------------
GLOBAL tests
--------------------------------------------------------------------------
TEST: Forward - Full supported trace (inline mode) [ OK ]
TEST: Forward - Full supported trace (encap mode) [ OK ]
- Tests passed: 88
- Tests failed: 0
Acked-by: David Ahern <dsahern@kernel.org>
Reviewed-by: Justin Iurman <justin.iurman@uliege.be>
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Here is the test result after conversion.
# ./icmp_redirect.sh
###########################################################################
Legacy routing
###########################################################################
TEST: IPv4: redirect exception [ OK ]
...
TEST: IPv4: mtu exception plus redirect [ OK ]
TEST: IPv6: mtu exception plus redirect [ OK ]
Tests passed: 40
Tests failed: 0
Tests xfailed: 0
Acked-by: David Ahern <dsahern@kernel.org>
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
