Takahiro Itazuri says:
====================
ptp: vmclock: Add VM generation counter and ACPI notification
Similarly to live migration, starting a VM from some serialized state
(aka snapshot) is an event which calls for adjusting guest clocks, hence
a hypervisor should increase the disruption_marker before resuming the
VM vCPUs, letting the guest know.
However, loading a snapshot, is slightly different than live migration,
especially since we can start multiple VMs from the same serialized
state. Apart from adjusting clocks, the guest needs to take additional
action during such events, e.g. recreate UUIDs, reset network
adapters/connections, reseed entropy pools, etc. These actions are not
necessary during live migration. This calls for a differentiation
between the two triggering events.
We differentiate between the two events via an extra field in the
vmclock_abi, called vm_generation_counter. Whereas hypervisors should
increase the disruption marker in both cases, they should only increase
vm_generation_counter when a snapshot is loaded in a VM (not during live
migration).
Additionally, we attach an ACPI notification to VMClock. Implementing
the notification is optional for the device. VMClock device will declare
that it implements the notification by setting
VMCLOCK_FLAG_NOTIFICATION_PRESENT bit in vmclock_abi flags. Hypervisors
that implement the notification must send an ACPI notification every
time seq_count changes to an even number. The driver will propagate
these notifications to userspace via the poll() interface.
====================
Link: https://patch.msgid.link/20260130173704.12575-1-itazur@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
To output UTC would involve complex calculations about whether the time
elapsed since the reference time has crossed the end of the month when
a leap second takes effect. I've prototyped that, but it made me sad.
Much better to report TAI, which is what PHCs should do anyway.
And much much simpler.
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Babis Chalios <bchalios@amazon.es>
Tested-by: Takahiro Itazuri <itazur@amazon.com>
Link: https://patch.msgid.link/20260130173704.12575-8-itazur@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Now that we added device tree support we can remove dependency on
CONFIG_ACPI.
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Babis Chalios <bchalios@amazon.es>
Tested-by: Takahiro Itazuri <itazur@amazon.dom>
Link: https://patch.msgid.link/20260130173704.12575-7-itazur@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
As we finalised the spec, we spotted that vmgenid actually says that the
_HID is supposed to be hypervisor-specific. Although in the 13 years
since the original vmgenid doc was published, nobody seems to have cared
about using _HID to distinguish between implementations on different
hypervisors, and we only ever use the _CID.
For consistency, match the _CID of "VMCLOCK" too.
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Babis Chalios <bchalios@amazon.es>
Tested-by: Takahiro Itazuri <itazur@amazon.com>
Link: https://patch.msgid.link/20260130173704.12575-6-itazur@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add device tree support to the ptp_vmclock driver, allowing it to probe
via device tree in addition to ACPI.
Handle optional interrupt for clock disruption notifications, mirroring
the ACPI notification behaviour.
Although the interrupt is marked as 'optional' in the DT bindings, if
the device *advertises* the VMCLOCK_FLAG_NOTIFICATION_ABSENT then it
*should* have an interrupt. The driver will refuse to initialize if not.
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Babis Chalios <bchalios@amazon.es>
Signed-off-by: Takahiro Itazuri <itazur@amazon.com>
Tested-by: Takahiro Itazuri <itazur@amazon.com>
Link: https://patch.msgid.link/20260130173704.12575-5-itazur@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The vmclock device provides a PTP clock source and precise timekeeping
across live migration and snapshot/restore operations.
The binding has a required memory region containing the vmclock_abi
structure and an optional interrupt for clock disruption notifications.
The full spec is at https://uapi-group.org/specifications/specs/vmclock/
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Babis Chalios <bchalios@amazon.es>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@oss.qualcomm.com>
Tested-by: Takahiro Itazuri <itazur@amazon.com>
Link: https://patch.msgid.link/20260130173704.12575-4-itazur@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add optional support for device notifications in VMClock. When
supported, the hypervisor will send a device notification every time it
updates the seq_count to a new even value.
Moreover, add support for poll() in VMClock as a means to propagate this
notification to user space. poll() will return a POLLIN event to
listeners every time seq_count changes to a value different than the one
last seen (since open() or last read()/pread()). This means that when
poll() returns a POLLIN event, listeners need to use read() to observe
what has changed and update the reader's view of seq_count. In other
words, after a poll() returned, all subsequent calls to poll() will
immediately return with a POLLIN event until the listener calls read().
The device advertises support for the notification mechanism by setting
flag VMCLOCK_FLAG_NOTIFICATION_PRESENT in vmclock_abi flags field. If
the flag is not present the driver won't setup the ACPI notification
handler and poll() will always immediately return POLLHUP.
Signed-off-by: Babis Chalios <bchalios@amazon.es>
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Takahiro Itazuri <itazur@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Tested-by: Takahiro Itazuri <itazur@amazon.com>
Link: https://patch.msgid.link/20260130173704.12575-3-itazur@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Similar to live migration, loading a VM from some saved state (aka
snapshot) is also an event that calls for clock adjustments in the
guest. However, guests might want to take more actions as a response to
such events, e.g. as discarding UUIDs, resetting network connections,
reseeding entropy pools, etc. These are actions that guests don't
typically take during live migration, so add a new field in the
vmclock_abi called vm_generation_counter which informs the guest about
such events.
Hypervisor advertises support for vm_generation_counter through the
VMCLOCK_FLAG_VM_GEN_COUNTER_PRESENT flag. Users need to check the
presence of this bit in vmclock_abi flags field before using this flag.
Signed-off-by: Babis Chalios <bchalios@amazon.es>
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Tested-by: Takahiro Itazur <itazur@amazon.com>
Link: https://patch.msgid.link/20260130173704.12575-2-itazur@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
All inet6_cork users also use one inet_cork_full.
Reduce number of parameters and increase data locality.
This saves ~275 bytes of code on x86_64.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20260130210303.3888261-9-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When we expect an IPv4 dst, use dst4_mtu() instead of dst_mtu()
to save some code space.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20260130210303.3888261-8-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When we expect an IPv6 dst, use dst6_mtu() instead of dst_mtu()
to save some code space.
Due to current dst6_mtu() implementation, only convert
users in IPv6 stack.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20260130210303.3888261-7-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
With CONFIG_MITIGATION_RETPOLINE=y dst_mtu() is a bit fat,
because it is generic.
Indeed, clang does not always inline it.
Add dst4_mtu() and dst6_mtu() helpers for callers that
expect either ipv4_mtu() or ip6_mtu() to be called.
These helpers are always inlined.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20260130210303.3888261-6-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When a too big packet is dropped, use SKB_DROP_REASON_PKT_TOO_BIG.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20260130210303.3888261-5-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
ip6_xmit() makes sure there is enough headroom in the skb,
it can uses __skb_push() instead of the out-of-line skb_push().
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20260130210303.3888261-4-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
1) daddr is unlikely a multicast in ip6_finish_output2().
2) ip6_finish_output_gso_slowpath_drop() should not be called often.
3) ip6_fragment() should not be called often.
4) opt is unlikely to be set.
5) ip6_xmit() and ip6_forward() mostly sends not too big packets.
6) Most __ip6_make_skb() calls are for UDP packets,
not ICMPV6 ones.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20260130210303.3888261-3-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
With CONFIG_STACKPROTECTOR_STRONG=y, it is better to avoid passing
a pointer to an automatic variable.
Change these exported functions to return 'u8 proto'
instead of void.
- ipv6_push_nfrag_opts()
- ipv6_push_frag_opts()
For instance, replace
ipv6_push_frag_opts(skb, opt, &proto);
with:
proto = ipv6_push_frag_opts(skb, opt, proto);
Note that even after this change, ip6_xmit() has to use a stack canary
because of @first_hop variable.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20260130210303.3888261-2-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Many network drivers have unnecessary empty module_init and module_exit
functions. Remove them (including some that just print a message). Note
that if a module_init function exists, a module_exit function must also
exist; otherwise, the module cannot be unloaded.
Signed-off-by: Ethan Nelson-Moore <enelsonmoore@gmail.com>
Acked-by: Marc Kleine-Budde <mkl@pengutronix.de>
Acked-by: Michael Grzeschik <m.grzeschik@pengutronix.de>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Acked-by: Ping-Ke Shih <pkshih@realtek.com>
Acked-by: Toke Høiland-Jørgensen <toke@toke.dk>
Link: https://patch.msgid.link/20260131004327.18112-1-enelsonmoore@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The module version is useless, and the only thing these drivers' init
routines did besides pci_register_driver was to print the driver name
and/or version.
Acked-by: Francois Romieu <romieu@fr.zoreil.com> (epic100)
Reviewed-by: Simon Horman <horms@kernel.org> (epic100, sis900)
Reviewed-by: Sai Krishna <saikrishnag@marvell.com> (epic100)
Signed-off-by: Ethan Nelson-Moore <enelsonmoore@gmail.com>
Link: https://patch.msgid.link/20260131022441.56274-1-enelsonmoore@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add the following check, to detect bugs sooner for CONFIG_DEBUG_NET=y
builds.
DEBUG_NET_WARN_ON_ONCE(skb->data < skb->head);
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20260130160253.2936789-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Sean Anderson says:
====================
net: phy: dp83867: Always program R/SGMII enable bits
The hardware designers at my company neglected to read the datasheet for
this PHY and did not add appropriate resistors to configure it for
SGMII. Add support for configuring the it based on phy-mode instead of
relying on the resistors for a suitable default.
====================
Link: https://patch.msgid.link/20260129171205.3868605-1-sean.anderson@linux.dev
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
If the board designers have neglected to populate the appropriate
resistors on the strapping pins then the phy may default to the wrong
interface mode. Enable/disable the RGMII/SGMII enable bits as necessary
to select the correct interface.
The dp83867 strapping pins have four levels and typically configure two
features at once. LED_0 controls both port mirroring and whether SGMII
is enabled. If it is pulled to VDDIO, both port mirroring and SGMII
will be enabled. For variants of the dp83867 that do not support SGMII,
this will prevent data from being transferred. As we now explicitly set
the SGMII and RGMII enable bits, we do not need to detect whether SGMII
has been inadvertently enabled.
Signed-off-by: Sean Anderson <sean.anderson@linux.dev>
Link: https://patch.msgid.link/20260129171205.3868605-3-sean.anderson@linux.dev
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
All supported interfaces use the TX FIFO register at least some of the
time, so there's no point in checking the interface. Retain the check
for the RX FIFO level since it is only used by SGMII.
Signed-off-by: Sean Anderson <sean.anderson@linux.dev>
Link: https://patch.msgid.link/20260129171205.3868605-2-sean.anderson@linux.dev
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Extend the RCU section a bit so that we can use the safer
skb_dst_dev_rcu() helper.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20260130191906.3781856-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
It appears that in commit 7efd79c0e6 ("bnxt_en: Add drop action
support for ntuple"), bnxt gained support for ntuple filters for packet
drops.
However, support for this does not seem to work in recent kernels or
against net-next:
% sudo ethtool -U eth0 flow-type udp4 src-ip 1.1.1.1 action -1
rmgr: Cannot insert RX class rule: Operation not supported
Cannot insert classification rule
The issue is that the existing code uses ethtool_get_flow_spec_ring_vf,
which will return a non-zero value if the ring_cookie is set to
RX_CLS_FLOW_DISC, which then causes bnxt_add_ntuple_cls_rule to return
-EOPNOTSUPP because it thinks the user is trying to set an ntuple filter
for a vf.
Fix this by first checking that the ring_cookie is not RX_CLS_FLOW_DISC.
After this patch, ntuple filters for drops can be added:
% sudo ethtool -U eth0 flow-type udp4 src-ip 1.1.1.1 action -1
Added rule with ID 0
% ethtool -n eth0
44 RX rings available
Total 1 rules
Filter: 0
Rule Type: UDP over IPv4
Src IP addr: 1.1.1.1 mask: 0.0.0.0
Dest IP addr: 0.0.0.0 mask: 255.255.255.255
TOS: 0x0 mask: 0xff
Src port: 0 mask: 0xffff
Dest port: 0 mask: 0xffff
Action: Drop
Reviewed-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: Joe Damato <joe@dama.to>
Link: https://patch.msgid.link/20260131003042.2570434-1-joe@dama.to
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Make the default (non-JSON) output more compact. Looking at RSS
context dumps is pretty much impossible without this, because
default print shows the indirection table with line per entry:
'indir': [0,
1,
2,
...
And indirection tables have 100-200 entries each.
The compact output is far more readable:
'indir': [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29,
Reviewed-by: Donald Hunter <donald.hunter@gmail.com>
Link: https://patch.msgid.link/20260131203029.1173492-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Spell out the recommendation that the RSS table should be
4x the queue count to avoid traffic imbalance. Include minor
rephrasing and removal of the explicit 128 entry example
since a 128 entry table is inadequate on modern machines.
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20260131225454.1225151-2-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add a test which checks that the RSS table is at least 4x the max
queue count supported by the device. The original RSS spec from
Microsoft stated that the RSS indirection table should be 2 to 8
times the CPU count, presumably assuming queue per CPU. If the
CPU count is not a power of two, however, a power-of-2 table
2x larger than queue count results in a 33% traffic imbalance.
Validate that the indirection table is at least 4x the queue
count. This lowers the imbalance to 16% which empirically
appears to be more acceptable to memcache-like workloads.
Reviewed-by: Willem de Bruijn <willemb@google.com>
Link: https://patch.msgid.link/20260131225454.1225151-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Print the PHY driver used and interrupt status after connection.
Signed-off-by: Chukun Pan <amadeus@jmu.edu.cn>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://patch.msgid.link/20260201100001.33102-1-amadeus@jmu.edu.cn
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
-----BEGIN PGP SIGNATURE-----
iQFHBAABCgAxFiEEn/sM2K9nqF/8FWzzDHRl3/mQkZwFAml91NkTHG1rbEBwZW5n
dXRyb25peC5kZQAKCRAMdGXf+ZCRnJkSB/wOdPfielzOfa4tmEQZGXW/tHXRBTbM
l7o0WjZtpWdTlqGYyHNngwPC8ie8SwP+vo6H6D9hPmLyxoQhIg8Kgi0GXh2t6n3s
UgpC+aTTtkbOtIaJL5r4xamuoyxw0KN0Y0FJSX+JnV03qe4BsQvTNAPbTdST29Bu
OC7GquBY37B2qe04B51AgeSEuKlVV1jfUmwa0N2Ugdb/LEiKe8W8PmGIiv3RnHQI
Ec2R+Tw+VZ6yX/qIMKJVeykejAxPw9GVbtfH0sBb1ZgWKPGxAYdMmlulqTzpBl2A
8KDNmmWX23giO6k1iTnDVsfNFtGXZBMuH3EIJMkM+hBWk9ZN2h24ROY3
=q55P
-----END PGP SIGNATURE-----
Merge tag 'linux-can-next-for-6.20-20260131' of git://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can-next
Marc Kleine-Budde says:
====================
pull-request: can-next 2026-01-31
This first 2 patches are by Biju Das, target the rcar_canfd driver and
add support for FD-only mode.
Lad Prabhakar's patches, also for the rcar_canfd driver add support
for the RZ/T2H SoC.
The last 2 patches are by Michael Tretter and me, target the sja1000
driver and clean up the CAN state handling.
* tag 'linux-can-next-for-6.20-20260131' of git://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can-next:
can: sja1000: sja1000_err(): use error counter for error state
can: sja1000: sja1000_err(): make use of sja1000_get_berr_counter() to read error counters
can: rcar_canfd: Add RZ/T2H support
dt-bindings: can: renesas,rcar-canfd: Document RZ/T2H and RZ/N2H SoCs
dt-bindings: can: renesas,rcar-canfd: Document RZ/V2H(P) and RZ/V2N SoCs
dt-bindings: can: renesas,rcar-canfd: Specify reset-names
can: rcar_canfd: Add support for FD-Only mode
dt-bindings: can: renesas,rcar-canfd: Document renesas,fd-only property
====================
Link: https://patch.msgid.link/20260131101512.1958907-1-mkl@pengutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This reverts commit d7cd421da9.
As reported by Al Viro, the TCP ULP support for SMC is fundamentally
broken. The implementation attempts to convert an active TCP socket
into an SMC socket by modifying the underlying `struct file`, dentry,
and inode in-place, which violates core VFS invariants that assume
these structures are immutable for an open file, creating a risk of
use after free errors and general system instability.
Given the severity of this design flaw and the fact that cleaner
alternatives (e.g., LD_PRELOAD, BPF) exist for legacy application
transparency, the correct course of action is to remove this feature
entirely.
Fixes: d7cd421da9 ("net/smc: Introduce TCP ULP support")
Link: https://lore.kernel.org/netdev/Yus1SycZxcd+wHwz@ZenIV/
Reported-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: D. Wythe <alibuda@linux.alibaba.com>
Reviewed-by: Tony Lu <tonylu@linux.alibaba.com>
Reviewed-by: Dust Li <dust.li@linux.alibaba.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20260128055452.98251-1-alibuda@linux.alibaba.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The AX25_DAMA_MASTER option has been unimplemented and marked broken
ever since it was introduced in 2007 in commit 954b2e7f4c ("[NET]
AX.25 Kconfig and docs updates and fixes"). At this point, it is very
unlikely it will be implemented. Remove it.
Signed-off-by: Ethan Nelson-Moore <enelsonmoore@gmail.com>
Link: https://patch.msgid.link/20260129080908.44710-1-enelsonmoore@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Slark Xiao says:
====================
net: wwan: add NMEA port type support
The series introduces a long discussed NMEA port type support for the
WWAN subsystem. There are two goals. From the WWAN driver perspective,
NMEA exported as any other port type (e.g. AT, MBIM, QMI, etc.). From
user space software perspective, the exported chardev belongs to the
GNSS class what makes it easy to distinguish desired port and the WWAN
device common to both NMEA and control (AT, MBIM, etc.) ports makes it
easy to locate a control port for the GNSS receiver activation.
Done by exporting the NMEA port via the GNSS subsystem with the WWAN
core acting as proxy between the WWAN modem driver and the GNSS
subsystem.
The series starts from a cleanup patch. Then three patches prepares the
WWAN core for the proxy style operation. Followed by a patch introding a
new WWNA port type, integration with the GNSS subsystem and demux. The
series ends with a couple of patches that introduce emulated EMEA port
to the WWAN HW simulator.
The series is the product of the discussion with Loic about the pros and
cons of possible models and implementation. Also Muhammad and Slark did
a great job defining the problem, sharing the code and pushing me to
finish the implementation. Daniele has caught an issue on driver
unloading and suggested an investigation direction. What was concluded
by Loic. Many thanks.
====================
Link: https://patch.msgid.link/20260126062158.308598-1-slark_xiao@163.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
For MHI WWAN device, we need a match between NMEA channel and
WWAN_PORT_NMEA type. Then the GNSS subsystem could create the
gnss device succssfully.
Signed-off-by: Slark Xiao <slark_xiao@163.com>
Reviewed-by: Loic Poulain <loic.poulain@oss.qualcomm.com>
Link: https://patch.msgid.link/20260126062158.308598-9-slark_xiao@163.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Support NMEA port emulation for the WWAN core GNSS port testing purpose.
Emulator produces pair of GGA + RMC sentences every second what should
be enough to fool gpsd into believing it is working with a NMEA GNSS
receiver.
If the GNSS system is enabled then one NMEA port will be created
automatically for the simulated WWAN device. Manual NMEA port creation
is not supported at the moment.
Signed-off-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Reviewed-by: Loic Poulain <loic.poulain@oss.qualcomm.com>
Link: https://patch.msgid.link/20260126062158.308598-8-slark_xiao@163.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Just introduced WWAN NMEA port type needs a testing option. The WWAN HW
simulator was developed with the AT port type in mind and cannot be
easily extended. Refactor it now to make it capable to support more port
types.
No big functional changes, mostly renaming with a little code
rearrangement.
Signed-off-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Reviewed-by: Loic Poulain <loic.poulain@oss.qualcomm.com>
Link: https://patch.msgid.link/20260126062158.308598-7-slark_xiao@163.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Many WWAN modems come with embedded GNSS receiver inside and have a
dedicated port to output geopositioning data. On the one hand, the
GNSS receiver has little in common with WWAN modem and just shares a
host interface and should be exported using the GNSS subsystem. On the
other hand, GNSS receiver is not automatically activated and needs a
generic WWAN control port (AT, MBIM, etc.) to be turned on. And a user
space software needs extra information to find the control port.
Introduce the new type of WWAN port - NMEA. When driver asks to register
a NMEA port, the core allocates common parent WWAN device as usual, but
exports the NMEA port via the GNSS subsystem and acts as a proxy between
the device driver and the GNSS subsystem.
From the WWAN device driver perspective, a NMEA port is registered as a
regular WWAN port without any difference. And the driver interacts only
with the WWAN core. From the user space perspective, the NMEA port is a
GNSS device which parent can be used to enumerate and select the proper
control port for the GNSS receiver management.
CC: Slark Xiao <slark_xiao@163.com>
CC: Muhammad Nuzaihan <zaihan@unrealasia.net>
CC: Qiang Yu <quic_qianyu@quicinc.com>
CC: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
CC: Johan Hovold <johan@kernel.org>
Suggested-by: Loic Poulain <loic.poulain@oss.qualcomm.com>
Signed-off-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Reviewed-by: Loic Poulain <loic.poulain@oss.qualcomm.com>
Link: https://patch.msgid.link/20260126062158.308598-6-slark_xiao@163.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Upcoming GNSS (NMEA) port type support requires exporting it via the
GNSS subsystem. On another hand, we still need to do basic WWAN core
work: call the port stop operation, purge queues, release the parent
WWAN device, etc. To reuse as much code as possible, split the port
unregistering function into the deregistration of a regular WWAN port
device, and the common port tearing down code.
In order to keep more code generic, break the device_unregister() call
into device_del() and put_device(), which release the port memory
uniformly.
Signed-off-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Reviewed-by: Loic Poulain <loic.poulain@oss.qualcomm.com>
Link: https://patch.msgid.link/20260126062158.308598-5-slark_xiao@163.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Upcoming GNSS (NMEA) port type support requires exporting it via the
GNSS subsystem. On another hand, we still need to do basic WWAN core
work: find or allocate the WWAN device, make it the port parent, etc. To
reuse as much code as possible, split the port creation function into
the registration of a regular WWAN port device, and basic port struct
initialization.
To be able to use put_device() uniformly, break the device_register()
call into device_initialize() and device_add() and call device
initialization earlier.
While at it, fix a minor number leak upon WWAN port registration
failure.
Signed-off-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Link: https://patch.msgid.link/20260126062158.308598-4-slark_xiao@163.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We need information about existing WWAN device children since we remove
the device after removing the last child. Previously, we tracked users
implicitly by checking whether ops was registered and existence of a
child device of the wwan_class class. Upcoming GNSS (NMEA) port type
support breaks this approach by introducing a child device of the
gnss_class class.
And a modem driver can easily trigger a kernel Oops by removing regular
(e.g., MBIM, AT) ports first and then removing a GNSS port. The WWAN
device will be unregistered on removal of a last regular WWAN port. And
subsequent GNSS port removal will cause NULL pointer dereference in
simple_recursive_removal().
In order to support ports of classes other than wwan_class, switch to
explicit references counting. Introduce a dedicated counter to the WWAN
device struct, increment it on every wwan_create_dev() call, decrement
on wwan_remove_dev(), and actually unregister the WWAN device when there
are no more references.
Run tested with wwan_hwsim with NMEA support patches applied and
different port removing sequences.
Reported-by: Daniele Palmas <dnlplm@gmail.com>
Closes: https://lore.kernel.org/netdev/CAGRyCJE28yf-rrfkFbzu44ygLEvoUM7fecK1vnrghjG_e9UaRA@mail.gmail.com/
Suggested-by: Loic Poulain <loic.poulain@oss.qualcomm.com>
Signed-off-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Link: https://patch.msgid.link/20260126062158.308598-3-slark_xiao@163.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
It was used initially for a port id allocation, then removed, and then
accidently introduced again, but it is still unused. Drop it again to
keep code clean.
Signed-off-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Reviewed-by: Loic Poulain <loic.poulain@oss.qualcomm.com>
Link: https://patch.msgid.link/20260126062158.308598-2-slark_xiao@163.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Mike Marciniszyn says:
====================
eth fbnic: Add debugfs for mbx and tx/rx
This patches adds debugfs read of the firmware mailbox tx/rx
and of the tx/rx rings for each napi vector.
====================
Link: https://patch.msgid.link/20260127200644.11640-1-mike.marciniszyn@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add debugfs hooks to display tx/rx rings for each napi
vector.
Note that the cloning mechanism in fbnic_ethtool.c for configuration
changes protects against concurrency issues with simultaneous config
changes along with debugs ring accesses.
The configuration switch builds up the new configuration offline,
takes the current config down, which removes the debugfs nv files, and
switches to the new configuration. The new configuration is brought
up which brings the debugfs files back on top of the new configuration
rings.
The interaction with fbnic_queue_stop() and fbnic_queue_start() will
similarly delete and add the files for the indicated vector.
Signed-off-by: Mike Marciniszyn (Meta) <mike.marciniszyn@gmail.com>
Link: https://patch.msgid.link/20260127200644.11640-3-mike.marciniszyn@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This patch adds reporting the Rx and Tx information
interfacing with the firmware.
The result of reading fbnic/fw_mbx is:
Rx
Rdy: 1 Head: 11 Tail: 10
Idx Len E Addr F H Raw
----------------------------------
00 4096 0 000101fea000 0 1 1000000101fea001
01 4096 0 000101feb000 0 1 1000000101feb001
.
.
.
15 4096 0 000101fe9000 0 1 1000000101fe9001
Tx
Rdy: 1 Head: 4 Tail: 4
Idx Len E Addr F H Raw
----------------------------------
00 0004 1 00010321b000 1 1 000440010321b003
01 0004 1 00010228d000 1 1 000440010228d003
.
.
.
15 0004 1 00010321b000 1 1 000440010321b003
Signed-off-by: Mike Marciniszyn (Meta) <mike.marciniszyn@gmail.com>
Link: https://patch.msgid.link/20260127200644.11640-2-mike.marciniszyn@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Many USB network drivers define get_drvinfo functions which add no
value over usbnet_get_drvinfo, only setting the driver name and
version. usbnet_get_drvinfo automatically sets the driver name, and
separate driver versions are now frowned upon in the kernel. Remove all
driver versions and replace these get_drvinfo functions with references
to usbnet_get_drvinfo where possible. Where that is not possible,
remove unnecessary code to set the driver name. Also remove two
unnecessary initializations from aqc111_get_drvinfo, an inaccurate
comment in pegasus.c, and an unused macro in catc.c.
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Peter Korsgaard <peter@korsgaard.com> (for dm9601.c)
Signed-off-by: Ethan Nelson-Moore <enelsonmoore@gmail.com>
Link: https://patch.msgid.link/20260129042435.13395-2-enelsonmoore@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add the support to read the rx alignment errors and update
them in the standard rtnl_link_stats64 structure.
Signed-off-by: Raju Rangoju <Raju.Rangoju@amd.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://patch.msgid.link/20260129111520.1567097-1-Raju.Rangoju@amd.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This commit adds SNMP drop count increment for the packets in
per NUMA queues which were introduced in commit b650bf0977
("udp: remove busylock and add per NUMA queues"). note that SNMP
counters are incremented currently by the caller for skb. And
that these skbs on the intermediate queue cannot be counted
there so need similar logic in their error path.
Signed-off-by: Mahdi Faramarzpour <mahdifrmx@gmail.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20260129083806.204752-1-mahdifrmx@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
By default, when a kmem_cache is created with SLAB_TYPESAFE_BY_RCU,
slub has to use extra storage for the freelist pointer after each
object, because slub assumes that any bit in the object
can be used by RCU readers.
Because proto_register() is also using SLAB_HWCACHE_ALIGN,
this forces slub to use one extra cache line per object.
We can instead put the slub freelist anywhere in the object,
granted the concurrent RCU readers are not supposed to
use the pointer value.
Add a new (struct sock)sk_freeptr field, in an union
with sk_rcu: No RCU readers would need to look at sk_rcu,
which is only used at free phase.
Tested:
grep . /sys/kernel/slab/TCP/{object_size,slab_size,objs_per_slab}
grep . /sys/kernel/slab/TCPv6/{object_size,slab_size,objs_per_slab}
Before:
/sys/kernel/slab/TCP/object_size:2368
/sys/kernel/slab/TCP/slab_size:2432
/sys/kernel/slab/TCP/objs_per_slab:13
/sys/kernel/slab/TCPv6/object_size:2496
/sys/kernel/slab/TCPv6/slab_size:2560
/sys/kernel/slab/TCPv6/objs_per_slab:12
After this patch, we can pack one more TCPv6 object per slab,
and object_size == slab_size.
/sys/kernel/slab/TCP/object_size:2368
/sys/kernel/slab/TCP/slab_size:2368
/sys/kernel/slab/TCP/objs_per_slab:13
/sys/kernel/slab/TCPv6/object_size:2496
/sys/kernel/slab/TCPv6/slab_size:2496
/sys/kernel/slab/TCPv6/objs_per_slab:13
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20260129153458.4163797-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Bhargava Marreddy says:
====================
bng_en: enhancements for RX and TX datapath
This series enhances the bng_en driver by adding:
1. Tx support (standard + TSO)
2. Rx support (standard + LRO/TPA)
====================
Link: https://patch.msgid.link/20260128185623.26559-1-bhargava.marreddy@broadcom.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
