Russell King says:
====================
net: stmmac: cleanups
This series removes various redundant items in the stmmac driver:
- the unused TBI and RTBI PCS flags
- the NULL pointer initialisations for PCS methods in dwxgmac2
- the stmmac_pcs_rane() method which is never called, and it's
associated implementations
- the redundant netif_carrier_off()s
Finally, it replaces asm/io.h with the preferred linux/io.h.
====================
Link: https://lore.kernel.org/r/Zlbp7xdUZAXblOZJ@shell.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
It is incorrect to call netif_carrier_off(), or in fact any driver
teardown, before unregister_netdev() has been called.
unregister_netdev() unpublishes the network device from userspace, and
takes the interface down if it was up prior to returning. Therefore,
once the call has returned, we are guaranteed that .ndo_stop() will
have been called for an interface that was up. Phylink will take the
carrier down via phylink_stop(), making any manipulation of the carrier
in the remove path unnecessary.
In the stmmac_release() path, the netif_carrier_off() call follows the
call to phylink_stop(), so this call is redundant.
Reviewed-by: Andrew Halaney <ahalaney@redhat.com>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Serge Semin <fancer.lancer@gmail.com>
Link: https://lore.kernel.org/r/E1sCErZ-00EOPx-PF@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
First of all the flags are never set by any of the driver parts. If nobody
have them set then the respective statements will always have the same
result. Thus the statements can be simplified or even dropped with no risk
to break things.
Secondly shall any of the TBI or RTBI flag is set the MDIO-bus
registration will be bypassed. Why? It really seems weird. It's perfectly
fine to have a TBI/RTBI-capable PHY configured over the MDIO bus
interface.
Based on the notes above the TBI/RTBI PCS flags can be freely dropped thus
simplifying the driver code.
Reviewed-by: Andrew Halaney <ahalaney@redhat.com>
Signed-off-by: Serge Semin <fancer.lancer@gmail.com>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Serge Semin <fancer.lancer@gmail.com>
Link: https://lore.kernel.org/r/E1sCErK-00EOPf-EP@rmk-PC.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
A comment in define_trace.h clearly states:
TRACE_INCLUDE_PATH if the path is something other than core kernel
vvvvvvvvvvvvvv
include/trace then this macro can define the path to use. Note, the path
is relative to define_trace.h, not the file including it. Full path names
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
for out of tree modules must be used.
fjes uses path relative to itself. Which (somehow) works most of
the time. Except when the kernel tree is "nested" in another
kernel tree, and ../drivers/net/fjes actually exists. In which
case build will use the header file from the wrong directory.
I've been trying to figure out why net NIPA builder is constantly
failing for the last 5 days, with:
include/trace/../../../drivers/net/fjes/fjes_trace.h:88:17: error: ‘__assign_str’ undeclared (first use in this function)
88 | __assign_str(err, err);
| ^~~~~~~~~~~~
when the line in the tree clearly has only one "err". NIPA does
indeed have "nested" trees, because it uses git work-trees and
the tree on the "outside" is not very up to date.
Link: https://lore.kernel.org/r/20240529023322.3467755-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Recovery from broken states can be hard. If the LIF reset in
the fw_down path didn't work because the PCI link was broken,
the FW won't be in the right state for proper restart. We can
fire another LIF reset in the fw_up path to be sure things
are clean on restart.
Signed-off-by: Shannon Nelson <shannon.nelson@amd.com>
Link: https://lore.kernel.org/r/20240529000259.25775-3-shannon.nelson@amd.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Each object file contains information about which module it gets linked
into, so linking the same file into multiple modules now causes a warning:
scripts/Makefile.build:254: drivers/net/ethernet/hisilicon/hns3/Makefile: hns3_common/hclge_comm_cmd.o is added to multiple modules: hclge hclgevf
scripts/Makefile.build:254: drivers/net/ethernet/hisilicon/hns3/Makefile: hns3_common/hclge_comm_rss.o is added to multiple modules: hclge hclgevf
scripts/Makefile.build:254: drivers/net/ethernet/hisilicon/hns3/Makefile: hns3_common/hclge_comm_tqp_stats.o is added to multiple modules: hclge hclgevf
Change the way that hns3 is built by moving the three common files into a
separate module with exported symbols instead.
Fixes: 5f20be4e90 ("net: hns3: refactor hns3 makefile to support hns3_common module")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://lore.kernel.org/r/20240528161603.2443125-1-arnd@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Each object file contains information about which module it gets linked
into, so linking the same file into multiple modules now causes a warning:
scripts/Makefile.build:254: drivers/net/ethernet/marvell/octeontx2/nic/Makefile: otx2_devlink.o is added to multiple modules: rvu_nicpf rvu_nicvf
Change the way that octeontx2 ethernet is built by moving the common
file into a separate module with exported symbols instead.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://lore.kernel.org/r/20240528152527.2148092-1-arnd@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Joe Damato says:
====================
mlx4: Add support for netdev-genl API
There are no functional changes from v5, which I mistakenly sent right
after net-next was closed (oops). This revision, however, includes
Tariq's Reviewed-by tags of the v5 in each commit message. See the
changelog below.
This series adds support to mlx4 for the netdev-genl API which makes it
much easier for users and user programs to map NAPI IDs back to
ifindexes, queues, and IRQs. This is extremely useful for a number of
use cases, including epoll-based busy poll.
In addition, this series includes a patch to generate per-queue
statistics using the netlink API, as well.
To facilitate the stats, patch 1/3 adds a field "alloc_fail" to the ring
structure. This is incremented by the driver in an appropriate place and
used in patch 3/3 as alloc_fail.
Please note: I do not have access to mlx4 hardware, but I've been
working closely with Martin Karsten from University of Waterloo (CC'd)
who has very graciously tested my patches on their mlx4 hardware (hence
his Tested-by attribution in each commit). His latest research work is
particularly interesting [1] and this series helps to support that (and
future) work.
Martin re-test v4 using Jakub's suggested tool [2] and the
stats.pkt_byte_sum and stats.qstat_by_ifindex tests passed. He also
adjusted the queue count and re-ran test to confirm it still passed even
if the queue count was modified.
[1]: https://dl.acm.org/doi/pdf/10.1145/3626780
[2]: https://lore.kernel.org/lkml/20240423175718.4ad4dc5a@kernel.org/
====================
Link: https://lore.kernel.org/r/20240528181139.515070-1-jdamato@fastly.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
MD Danish Anwar says:
====================
Introduce switch mode support for ICSSG driver
This series adds support for switch-mode for ICSSG driver. This series
also introduces helper APIs to configure firmware maintained FDB
(Forwarding Database) and VLAN tables. These APIs are later used by ICSSG
driver in switch mode.
Now the driver will boot by default in dual EMAC mode. When first ICSSG
interface is added to bridge driver will still be in EMAC mode. As soon as
second ICSSG interface is added to same bridge, switch-mode will be
enabled and switch firmwares will be loaded to PRU cores. The driver will
remain in dual EMAC mode if ICSSG interfaces are added to two different
bridges or if two different interfaces (One ICSSG, one other) is added to
the same bridge. We'll only enable is_switch_mode flag when two ICSSG
interfaces are added to same bridge.
We start in dual MAC mode. Let's say lan0 and lan1 are ICSSG interfaces
ip link add name br0 type bridge
ip link set lan0 master br0
At this point, we get a CHANGEUPPER event. Only one port is a member of
the bridge, so we will still be in dual MAC mode.
ip link set lan1 master br0
We get a second CHANGEUPPER event, the second interface lan1 is also ICSSG
interface so we will set the is_switch_mode flag and when interfaces are
brought up again, ICSSG switch firmwares will be loaded to PRU Cores.
There are some other cases to consider as well.
ip link add name br0 type bridge
ip link add name br1 type bridge
ip link set lan0 master br0
ip link set ppp0 master br0
Here we are adding lan0 (ICSSG) and ppp0 (non ICSSG) to same bridge, as
they both are not ICSSG, we will still be running in dual EMAC mode.
ip link set lan1 master br1
ip link set vpn0 master br1
Here we are adding lan1 (ICSSG) and vpn0 (non ICSSG) to same bridge, as
they both are not ICSSG, we will still be running in dual EMAC mode.
This is v6 of the series.
Changes from v5 to v6:
*) Removed __packed from structures in icssg_config.h file.
*) Added RB tags of Andrew Lunn <andrew@lunn.ch> to patch 2/3 and patch
3/3 of this series.
Changes from v4 to v5:
*) Rebased on 6.10-rc1.
*) Dropped the RFC tag.
Changes from v3 to v4:
*) Added RFC tag as net-next is closed now.
*) Modified the driver to remove the need of bringing interfaces up / down
for enabling / disabling switch mode. Now switch mode can be enabled
without bringig interfaces up / down as requested by Andrew Lunn
<andrew@lunn.ch>
*) Modified commit message of patch 3/3.
Changes from v2 to v3:
*) Dropped RFC tag.
*) Used ether_addr_copy() instead of manually copying mac address using
for loop in patch 1/3 as suggested by Andrew Lunn <andrew@lunn.ch>
*) Added helper API icssg_fdb_setup() in patch 1/3 to reduce code
duplication as suggested by Andrew Lunn <andrew@lunn.ch>
*) In prueth_switchdev_stp_state_set() removed BR_STATE_LEARNING as
learning without forwarding is not supported by ICSSG firmware.
*) Used ether_addr_equal() wherever possible in patch 2/3 as suggested
by Andrew Lunn <andrew@lunn.ch>
*) Fixed typo "nit: s/prueth_switchdevice_nb/prueth_switchdev_nb/" in
patch 2/3 as suggested by Simon Horman <horms@kernel.org>
*) Squashed "#include "icssg_mii_rt.h" to patch 2/3 from patch 3/3 as
suggested by Simon Horman <horms@kernel.org>
*) Rebased on latest net-next/main.
Changes from v1 to v2:
*) Removed TAPRIO support patch from this series.
*) Stopped using devlink for enabling switch-mode as suggested by Andrew L
*) Added read_poll_timeout() in patch 1 / 3 as suggested by Andrew L.
v1 https://lore.kernel.org/all/20230830110847.1219515-4-danishanwar@ti.com/
v2 https://lore.kernel.org/all/20240118071005.1514498-1-danishanwar@ti.com/
v3 https://lore.kernel.org/all/20240327114054.1907278-1-danishanwar@ti.com/
v4 https://lore.kernel.org/all/20240515060320.2783244-1-danishanwar@ti.com/
v5 https://lore.kernel.org/all/20240527052738.152821-1-danishanwar@ti.com/
Thanks and Regards,
Md Danish Anwar
====================
Link: https://lore.kernel.org/r/20240528113734.379422-1-danishanwar@ti.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Add support for ICSSG switch firmware using existing Dual EMAC driver
with switchdev.
Limitations:
VLAN offloading is limited to 0-256 IDs.
MDB/FDB static entries are limited to 511 entries and different FDBs can
hash to same bucket and thus may not completely offloaded
Example assuming ETH1 and ETH2 as ICSSG2 interfaces:
Switch to ICSSG Switch mode:
ip link add name br0 type bridge
ip link set dev eth1 master br0
ip link set dev eth2 master br0
ip link set dev br0 up
bridge vlan add dev br0 vid 1 pvid untagged self
Going back to Dual EMAC mode:
ip link set dev br0 down
ip link set dev eth1 nomaster
ip link set dev eth2 nomaster
ip link del name br0 type bridge
By default, Dual EMAC firmware is loaded, and can be changed to switch
mode by above steps
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: MD Danish Anwar <danishanwar@ti.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
ICSSG can operating in switch mode with 2 ext port and 1 host port with
VLAN/FDB/MDB and STP offloading. Add switchdev based driver to
support the same.
Driver itself will be integrated with icssg_prueth in future commits
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: MD Danish Anwar <danishanwar@ti.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Introduce helper functions to configure firmware FDB tables, VLAN tables
and Port VLAN ID settings to aid adding Switch mode support.
Signed-off-by: MD Danish Anwar <danishanwar@ti.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Alexander Sverdlin says:
====================
net: ethernet: ti: am65-cpsw-nuss: support stacked switches
Currently an external Ethernet switch connected to a am65-cpsw-nuss CPU
port will not be probed successfully because of_find_net_device_by_node()
will not be able to find the netdev of the CPU port.
It's necessary to populate of_node of the struct device for the
am65-cpsw-nuss ports. DT nodes of the ports are already stored in per-port
private data, but because of some legacy reasons the naming ("phy_node")
was misleading.
====================
Link: https://lore.kernel.org/r/20240528075954.3608118-1-alexander.sverdlin@siemens.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Eric Dumazet says:
====================
tcp: fix tcp_poll() races
Flakes in packetdrill tests stressing epoll_wait()
were root caused to bad ordering in tcp_write_err()
Precisely, we have to call sk_error_report() after
tcp_done().
When fixing this issue, we discovered tcp_abort(),
tcp_v4_err() and tcp_v6_err() had similar issues.
Since tcp_reset() has the correct ordering,
first patch takes part of it and creates
tcp_done_with_error() helper.
====================
Link: https://lore.kernel.org/r/20240528125253.1966136-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
These functions have races when they:
1) Write sk->sk_err
2) call sk_error_report(sk)
3) call tcp_done(sk)
As described in prior patches in this series:
An smp_wmb() is missing.
We should call tcp_done() before sk_error_report(sk)
to have consistent tcp_poll() results on SMP hosts.
Use tcp_done_with_error() where we centralized the
correct sequence.
Fixes: 1da177e4c3 ("Linux-2.6.12-rc2")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Link: https://lore.kernel.org/r/20240528125253.1966136-5-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
I noticed flakes in a packetdrill test, expecting an epoll_wait()
to return EPOLLERR | EPOLLHUP on a failed connect() attempt,
after multiple SYN retransmits. It sometimes return EPOLLERR only.
The issue is that tcp_write_err():
1) writes an error in sk->sk_err,
2) calls sk_error_report(),
3) then calls tcp_done().
tcp_done() is writing SHUTDOWN_MASK into sk->sk_shutdown,
among other things.
Problem is that the awaken user thread (from 2) sk_error_report())
might call tcp_poll() before tcp_done() has written sk->sk_shutdown.
tcp_poll() only sees a non zero sk->sk_err and returns EPOLLERR.
This patch fixes the issue by making sure to call sk_error_report()
after tcp_done().
tcp_write_err() also lacks an smp_wmb().
We can reuse tcp_done_with_error() to factor out the details,
as Neal suggested.
Fixes: 1da177e4c3 ("Linux-2.6.12-rc2")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Link: https://lore.kernel.org/r/20240528125253.1966136-3-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
If a user provides an invalid netconsole configuration during boot time
(e.g., specifying an invalid ethX interface), netconsole will be
entirely disabled. Consequently, the user won't be able to create new
entries in /sys/kernel/config/netconsole/ as that directory does not
exist.
Apart from misconfiguration, another issue arises when ethX is loaded as
a module and the netconsole= line in the command line points to ethX,
resulting in an obvious failure. This renders netconsole unusable, as
/sys/kernel/config/netconsole/ will never appear. This is more annoying
since users reconfigure (or just toggle) the configuratin later (see
commit 5fbd6cdbe3 ("netconsole: Attach cmdline target to dynamic
target"))
Create /sys/kernel/config/netconsole/ even if the command line arguments
are invalid, so, users can create dynamic entries in netconsole.
Reported-by: Aijay Adams <aijay@meta.com>
Signed-off-by: Breno Leitao <leitao@debian.org>
Link: https://lore.kernel.org/r/20240528084225.3215853-1-leitao@debian.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
An earlier commit deleted the TSO support in the Cortina Gemini
driver because the driver was confusing gso_size and MTU,
probably because what the Linux kernel calls "gso_size" was
called "MTU" in the datasheet.
Restore the functionality properly reading the gso_size from
the skbuff.
Tested with iperf3, running a server on a different machine
and client on the device with the cortina gemini ethernet:
Connecting to host 192.168.1.2, port 5201
60008000.ethernet-port eth0: segment offloading mss = 05ea len=1c8a
60008000.ethernet-port eth0: segment offloading mss = 05ea len=1c8a
60008000.ethernet-port eth0: segment offloading mss = 05ea len=27da
60008000.ethernet-port eth0: segment offloading mss = 05ea len=0b92
60008000.ethernet-port eth0: segment offloading mss = 05ea len=2bda
(...)
(The hardware MSS 0x05ea here includes the ethernet headers.)
If I disable all segment offloading on the receiving host and
dump packets using tcpdump -xx like this:
ethtool -K enp2s0 gro off gso off tso off
tcpdump -xx -i enp2s0 host 192.168.1.136
I get segmented packages such as this when running iperf3:
23:16:54.024139 IP OpenWrt.lan.59168 > Fecusia.targus-getdata1:
Flags [.], seq 1486:2934, ack 1, win 4198,
options [nop,nop,TS val 3886192908 ecr 3601341877], length 1448
0x0000: fc34 9701 a0c6 14d6 4da8 3c4f 0800 4500
0x0010: 05dc 16a0 4000 4006 9aa1 c0a8 0188 c0a8
0x0020: 0102 e720 1451 ff25 9822 4c52 29cf 8010
0x0030: 1066 ac8c 0000 0101 080a e7a2 990c d6a8
(...)
0x05c0: 5e49 e109 fe8c 4617 5e18 7a82 7eae d647
0x05d0: e8ee ae64 dc88 c897 3f8a 07a4 3a33 6b1b
0x05e0: 3501 a30f 2758 cc44 4b4a
Several such packets often follow after each other verifying
the segmentation into 0x05a8 (1448) byte packages also on the
reveiving end. As can be seen, the ethernet frames are
0x05ea (1514) in size.
Performance with iperf3 before this patch: ~15.5 Mbit/s
Performance with iperf3 after this patch: ~175 Mbit/s
This was running a 60 second test (twice) the best measurement
was 179 Mbit/s.
For comparison if I run iperf3 with UDP I get around 1.05 Mbit/s
both before and after this patch.
While this is a gigabit ethernet interface, the CPU is a cheap
D-Link DIR-685 router (based on the ARMv5 Faraday FA526 at
~50 MHz), and the software is not supposed to drive traffic,
as the device has a DSA chip, so this kind of numbers can be
expected.
Fixes: ac631873c9 ("net: ethernet: cortina: Drop TSO support")
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
