Accidentally spotted while trying to understand what else needs
to be renamed to netif_ prefix. Most of the calls to dev_set_promiscuity
are adjacent to dev_set_allmulti or dev_disable_lro so it should
be safe to add the lock. Note that new netif_set_promiscuity is
currently unused, the locked paths call __dev_set_promiscuity directly.
Fixes: ad7c7b2172 ("net: hold netdev instance lock during sysfs operations")
Signed-off-by: Stanislav Fomichev <sdf@fomichev.me>
Link: https://patch.msgid.link/20250506011919.2882313-1-sdf@fomichev.me
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
__qdisc_destroy() calls into various qdiscs .destroy() op, which in turn
can call .ndo_setup_tc(), which requires the netdev instance lock.
This commit extends the critical section in
unregister_netdevice_many_notify() to cover dev_shutdown() (and
dev_tcx_uninstall() as a side-effect) and acquires the netdev instance
lock in __dev_change_net_namespace() for the other dev_shutdown() call.
This should now guarantee that for all qdisc ops, the netdev instance
lock is held during .ndo_setup_tc().
Fixes: a0527ee2df ("net: hold netdev instance lock during qdisc ndo_setup_tc")
Signed-off-by: Cosmin Ratiu <cratiu@nvidia.com>
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Link: https://patch.msgid.link/20250505194713.1723399-1-cratiu@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Use Device Serial Number instead of PCI bus/device/function for
the index of struct ice_adapter.
Functions on the same physical device should point to the very same
ice_adapter instance, but with two PFs, when at least one of them is
PCI-e passed-through to a VM, it is no longer the case - PFs will get
seemingly random PCI BDF values, and thus indices, what finally leds to
each of them being on their own instance of ice_adapter. That causes them
to don't attempt any synchronization of the PTP HW clock usage, or any
other future resources.
DSN works nicely in place of the index, as it is "immutable" in terms of
virtualization.
Fixes: 0e2bddf9e5 ("ice: add ice_adapter for shared data across PFs on the same NIC")
Suggested-by: Jacob Keller <jacob.e.keller@intel.com>
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Suggested-by: Jiri Pirko <jiri@resnulli.us>
Reviewed-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>
Signed-off-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Tested-by: Rinitha S <sx.rinitha@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Link: https://patch.msgid.link/20250505161939.2083581-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The purpose of resetting the TX queue is to reset the byte and packet
count as well as to clear the software flow control XOFF bit.
MediaTek developers pointed out that netdev_reset_queue would only
resets queue 0 of the network device.
Queues that are not reset may cause unexpected issues.
Packets may stop being sent after reset and "transmit timeout" log may
be displayed.
Import fix from MediaTek's SDK to resolve this issue.
Link: 319c0d9905
Fixes: f63959c7ee ("net: ethernet: mtk_eth_soc: implement multi-queue support for per-port queues")
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Link: https://patch.msgid.link/c9ff9adceac4f152239a0f65c397f13547639175.1746406763.git.daniel@makrotopia.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Strings from the kernel are guaranteed to be null terminated and
ynl_attr_validate() checks for this. But it doesn't check if the string
has a len of 0, which would cause problems when trying to access
data[len - 1]. Fix this by checking that len is positive.
Signed-off-by: David Wei <dw@davidwei.uk>
Link: https://patch.msgid.link/20250503043050.861238-1-dw@davidwei.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Mohsin Bashir says:
====================
selftests: drv: net: fix `ping.py` test failure
Fix `ping.py` test failure on an ipv6 system, and appropriately handle the
cases where either one of the two address families (ipv4, ipv6) is not
present.
====================
Link: https://patch.msgid.link/20250503013518.1722913-1-mohsin.bashr@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently, the test result does not differentiate between the cases when
either one of the address families are configured or if both the address
families are configured. Ideally, the result should report if a
particular case was skipped.
./drivers/net/ping.py
TAP version 13
1..7
ok 1 ping.test_default_v4 # SKIP Test requires IPv4 connectivity
ok 2 ping.test_default_v6
ok 3 ping.test_xdp_generic_sb
ok 4 ping.test_xdp_generic_mb
ok 5 ping.test_xdp_native_sb
ok 6 ping.test_xdp_native_mb
ok 7 ping.test_xdp_offload # SKIP device does not support offloaded XDP
Totals: pass:5 fail:0 xfail:0 xpass:0 skip:2 error:0
Fixes: 75cc19c8ff ("selftests: drv-net: add xdp cases for ping.py")
Signed-off-by: Mohsin Bashir <mohsin.bashr@gmail.com>
Reviewed-by: David Wei <dw@davidwei.uk>
Link: https://patch.msgid.link/20250503013518.1722913-4-mohsin.bashr@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
On a system with either of the ipv4 or ipv6 information missing, tests
are currently skipped. Ideally, the test should run as long as at least
one address family is present. This patch make test run whenever
possible.
Before:
./drivers/net/ping.py
TAP version 13
1..6
ok 1 ping.test_default # SKIP Test requires IPv4 connectivity
ok 2 ping.test_xdp_generic_sb # SKIP Test requires IPv4 connectivity
ok 3 ping.test_xdp_generic_mb # SKIP Test requires IPv4 connectivity
ok 4 ping.test_xdp_native_sb # SKIP Test requires IPv4 connectivity
ok 5 ping.test_xdp_native_mb # SKIP Test requires IPv4 connectivity
ok 6 ping.test_xdp_offload # SKIP device does not support offloaded XDP
Totals: pass:0 fail:0 xfail:0 xpass:0 skip:6 error:0
After:
./drivers/net/ping.py
TAP version 13
1..6
ok 1 ping.test_default
ok 2 ping.test_xdp_generic_sb
ok 3 ping.test_xdp_generic_mb
ok 4 ping.test_xdp_native_sb
ok 5 ping.test_xdp_native_mb
ok 6 ping.test_xdp_offload # SKIP device does not support offloaded XDP
Totals: pass:5 fail:0 xfail:0 xpass:0 skip:1 error:0
Fixes: 75cc19c8ff ("selftests: drv-net: add xdp cases for ping.py")
Signed-off-by: Mohsin Bashir <mohsin.bashr@gmail.com>
Link: https://patch.msgid.link/20250503013518.1722913-3-mohsin.bashr@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The `get_interface_info` call has ip version hard-coded which leads to
failures on an IPV6 system. The NetDrvEnv class already gathers
information about remote interface, so instead of fixing the local
implementation switch to using cfg.remote_ifname.
Before:
./drivers/net/ping.py
Traceback (most recent call last):
File "/new_tests/./drivers/net/ping.py", line 217, in <module>
main()
File "/new_tests/./drivers/net/ping.py", line 204, in main
get_interface_info(cfg)
File "/new_tests/./drivers/net/ping.py", line 128, in get_interface_info
raise KsftFailEx('Can not get remote interface')
net.lib.py.ksft.KsftFailEx: Can not get remote interface
After:
./drivers/net/ping.py
TAP version 13
1..6
ok 1 ping.test_default # SKIP Test requires IPv4 connectivity
ok 2 ping.test_xdp_generic_sb # SKIP Test requires IPv4 connectivity
ok 3 ping.test_xdp_generic_mb # SKIP Test requires IPv4 connectivity
ok 4 ping.test_xdp_native_sb # SKIP Test requires IPv4 connectivity
ok 5 ping.test_xdp_native_mb # SKIP Test requires IPv4 connectivity
ok 6 ping.test_xdp_offload # SKIP device does not support offloaded XDP
Totals: pass:0 fail:0 xfail:0 xpass:0 skip:6 error:0
Fixes: 75cc19c8ff ("selftests: drv-net: add xdp cases for ping.py")
Signed-off-by: Mohsin Bashir <mohsin.bashr@gmail.com>
Reviewed-by: David Wei <dw@davidwei.uk>
Link: https://patch.msgid.link/20250503013518.1722913-2-mohsin.bashr@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Guillaume Nault says:
====================
gre: Reapply IPv6 link-local address generation fix.
Reintroduce the IPv6 link-local address generation fix for GRE and its
kernel selftest. These patches were introduced by merge commit
b3fc5927de ("Merge branch
'gre-fix-regressions-in-ipv6-link-local-address-generation'") but have
been reverted by commit 8417db0be5 ("Merge branch
'gre-revert-ipv6-link-local-address-fix'"), because it uncovered
another bug in multipath routing. Now that this bug has been
investigated and fixed, we can apply the GRE link-local address fix
and its kernel selftest again.
For convenience, here's the original cover letter:
IPv6 link-local address generation has some special cases for GRE
devices. This has led to several regressions in the past, and some of
them are still not fixed. This series fixes the remaining problems,
like the ipv6.conf.<dev>.addr_gen_mode sysctl being ignored and the
router discovery process not being started (see details in patch 1).
To avoid any further regressions, patch 2 adds selftests covering
IPv4 and IPv6 gre/gretap devices with all combinations of currently
supported addr_gen_mode values.
====================
Link: https://patch.msgid.link/cover.1746225213.git.gnault@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
GRE devices have their special code for IPv6 link-local address
generation that has been the source of several regressions in the past.
Add selftest to check that all gre, ip6gre, gretap and ip6gretap get an
IPv6 link-link local address in accordance with the
net.ipv6.conf.<dev>.addr_gen_mode sysctl.
Note: This patch was originally applied as commit 6f50175cca ("selftests:
Add IPv6 link-local address generation tests for GRE devices.").
However, it was then reverted by commit 355d940f4d ("Revert "selftests:
Add IPv6 link-local address generation tests for GRE devices."")
because the commit it depended on was going to be reverted. Now that
the situation is resolved, we can add this selftest again (no changes
since original patch, appart from context update in
tools/testing/selftests/net/Makefile).
Signed-off-by: Guillaume Nault <gnault@redhat.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Tested-by: Ido Schimmel <idosch@nvidia.com>
Link: https://patch.msgid.link/2c3a5733cb3a6e3119504361a9b9f89fda570a2d.1746225214.git.gnault@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Use addrconf_addr_gen() to generate IPv6 link-local addresses on GRE
devices in most cases and fall back to using add_v4_addrs() only in
case the GRE configuration is incompatible with addrconf_addr_gen().
GRE used to use addrconf_addr_gen() until commit e5dd729460 ("ip/ip6_gre:
use the same logic as SIT interfaces when computing v6LL address")
restricted this use to gretap and ip6gretap devices, and created
add_v4_addrs() (borrowed from SIT) for non-Ethernet GRE ones.
The original problem came when commit 9af28511be ("addrconf: refuse
isatap eui64 for INADDR_ANY") made __ipv6_isatap_ifid() fail when its
addr parameter was 0. The commit says that this would create an invalid
address, however, I couldn't find any RFC saying that the generated
interface identifier would be wrong. Anyway, since gre over IPv4
devices pass their local tunnel address to __ipv6_isatap_ifid(), that
commit broke their IPv6 link-local address generation when the local
address was unspecified.
Then commit e5dd729460 ("ip/ip6_gre: use the same logic as SIT
interfaces when computing v6LL address") tried to fix that case by
defining add_v4_addrs() and calling it to generate the IPv6 link-local
address instead of using addrconf_addr_gen() (apart for gretap and
ip6gretap devices, which would still use the regular
addrconf_addr_gen(), since they have a MAC address).
That broke several use cases because add_v4_addrs() isn't properly
integrated into the rest of IPv6 Neighbor Discovery code. Several of
these shortcomings have been fixed over time, but add_v4_addrs()
remains broken on several aspects. In particular, it doesn't send any
Router Sollicitations, so the SLAAC process doesn't start until the
interface receives a Router Advertisement. Also, add_v4_addrs() mostly
ignores the address generation mode of the interface
(/proc/sys/net/ipv6/conf/*/addr_gen_mode), thus breaking the
IN6_ADDR_GEN_MODE_RANDOM and IN6_ADDR_GEN_MODE_STABLE_PRIVACY cases.
Fix the situation by using add_v4_addrs() only in the specific scenario
where the normal method would fail. That is, for interfaces that have
all of the following characteristics:
* run over IPv4,
* transport IP packets directly, not Ethernet (that is, not gretap
interfaces),
* tunnel endpoint is INADDR_ANY (that is, 0),
* device address generation mode is EUI64.
In all other cases, revert back to the regular addrconf_addr_gen().
Also, remove the special case for ip6gre interfaces in add_v4_addrs(),
since ip6gre devices now always use addrconf_addr_gen() instead.
Note:
This patch was originally applied as commit 183185a18f ("gre: Fix
IPv6 link-local address generation."). However, it was then reverted
by commit fc486c2d06 ("Revert "gre: Fix IPv6 link-local address
generation."") because it uncovered another bug that ended up
breaking net/forwarding/ip6gre_custom_multipath_hash.sh. That other
bug has now been fixed by commit 4d0ab3a688 ("ipv6: Start path
selection from the first nexthop"). Therefore we can now revive this
GRE patch (no changes since original commit 183185a18f ("gre: Fix
IPv6 link-local address generation.").
Fixes: e5dd729460 ("ip/ip6_gre: use the same logic as SIT interfaces when computing v6LL address")
Signed-off-by: Guillaume Nault <gnault@redhat.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://patch.msgid.link/a88cc5c4811af36007645d610c95102dccb360a6.1746225214.git.gnault@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Device Tree and Ethernet MAC driver writers often misunderstand RGMII
delays. Rewrite the Normative section in terms of the PCB, is the PCB
adding the 2ns delay. This meaning was previous implied by the
definition, but often wrongly interpreted due to the ambiguous wording
and looking at the definition from the wrong perspective. The new
definition concentrates clearly on the hardware, and should be less
ambiguous.
Add an Informative section to the end of the binding describing in
detail what the four RGMII delays mean. This expands on just the PCB
meaning, adding in the implications for the MAC and PHY.
Additionally, when the MAC or PHY needs to add a delay, which is
software configuration, describe how Linux does this, in the hope of
reducing errors. Make it clear other users of device tree binding may
implement the software configuration in other ways while still
conforming to the binding.
Fixes: 9d3de3c583 ("dt-bindings: net: Add YAML schemas for the generic Ethernet options")
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Acked-by: Conor Dooley <conor.dooley@microchip.com>
Link: https://patch.msgid.link/20250430-v6-15-rc3-net-rgmii-delays-v2-1-099ae651d5e5@lunn.ch
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Commit 4bc12818b3 ("virtio-net: disable delayed refill when pausing rx")
fixed a deadlock between reconfig paths and refill work trying to disable
the same NAPI instance. The refill work can't run in parallel with reconfig
because trying to double-disable a NAPI instance causes a stall under the
instance lock, which the reconfig path needs to re-enable the NAPI and
therefore unblock the stalled thread.
There are two cases where we re-enable refill too early. One is in the
virtnet_set_queues() handler. We call it when installing XDP:
virtnet_rx_pause_all(vi);
...
virtnet_napi_tx_disable(..);
...
virtnet_set_queues(..);
...
virtnet_rx_resume_all(..);
We want the work to be disabled until we call virtnet_rx_resume_all(),
but virtnet_set_queues() kicks it before NAPIs were re-enabled.
The other case is a more trivial case of mis-ordering in
__virtnet_rx_resume() found by code inspection.
Taking the spin lock in virtnet_set_queues() (requested during review)
may be unnecessary as we are under rtnl_lock and so are all paths writing
to ->refill_enabled.
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Bui Quang Minh <minhquangbui99@gmail.com>
Fixes: 4bc12818b3 ("virtio-net: disable delayed refill when pausing rx")
Fixes: 413f0271f3 ("net: protect NAPI enablement with netdev_lock()")
Link: https://patch.msgid.link/20250430163758.3029367-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Cong Wang says:
====================
net_sched: fix a regression in sch_htb
This patchset contains a fix for the regression reported by Alan and a
selftest to cover that case. Please see each patch description for more
details.
====================
Link: https://patch.msgid.link/20250428232955.1740419-1-xiyou.wangcong@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Alan reported a NULL pointer dereference in htb_next_rb_node()
after we made htb_qlen_notify() idempotent.
It turns out in the following case it introduced some regression:
htb_dequeue_tree():
|-> fq_codel_dequeue()
|-> qdisc_tree_reduce_backlog()
|-> htb_qlen_notify()
|-> htb_deactivate()
|-> htb_next_rb_node()
|-> htb_deactivate()
For htb_next_rb_node(), after calling the 1st htb_deactivate(), the
clprio[prio]->ptr could be already set to NULL, which means
htb_next_rb_node() is vulnerable here.
For htb_deactivate(), although we checked qlen before calling it, in
case of qlen==0 after qdisc_tree_reduce_backlog(), we may call it again
which triggers the warning inside.
To fix the issues here, we need to:
1) Make htb_deactivate() idempotent, that is, simply return if we
already call it before.
2) Make htb_next_rb_node() safe against ptr==NULL.
Many thanks to Alan for testing and for the reproducer.
Fixes: 5ba8b837b5 ("sch_htb: make htb_qlen_notify() idempotent")
Reported-by: Alan J. Wylie <alan@wylie.me.uk>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Link: https://patch.msgid.link/20250428232955.1740419-2-xiyou.wangcong@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pull networking fixes from Jakub Kicinski:
"Happy May Day.
Things have calmed down on our end (knock on wood), no outstanding
investigations. Including fixes from Bluetooth and WiFi.
Current release - fix to a fix:
- igc: fix lock order in igc_ptp_reset
Current release - new code bugs:
- Revert "wifi: iwlwifi: make no_160 more generic", fixes regression
to Killer line of devices reported by a number of people
- Revert "wifi: iwlwifi: add support for BE213", initial FW is too
buggy
- number of fixes for mld, the new Intel WiFi subdriver
Previous releases - regressions:
- wifi: mac80211: restore monitor for outgoing frames
- drv: vmxnet3: fix malformed packet sizing in vmxnet3_process_xdp
- eth: bnxt_en: fix timestamping FIFO getting out of sync on reset,
delivering stale timestamps
- use sock_gen_put() in the TCP fraglist GRO heuristic, don't assume
every socket is a full socket
Previous releases - always broken:
- sched: adapt qdiscs for reentrant enqueue cases, fix list
corruptions
- xsk: fix race condition in AF_XDP generic RX path, shared UMEM
can't be protected by a per-socket lock
- eth: mtk-star-emac: fix spinlock recursion issues on rx/tx poll
- btusb: avoid NULL pointer dereference in skb_dequeue()
- dsa: felix: fix broken taprio gate states after clock jump"
* tag 'net-6.15-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (83 commits)
net: vertexcom: mse102x: Fix RX error handling
net: vertexcom: mse102x: Add range check for CMD_RTS
net: vertexcom: mse102x: Fix LEN_MASK
net: vertexcom: mse102x: Fix possible stuck of SPI interrupt
net: hns3: defer calling ptp_clock_register()
net: hns3: fixed debugfs tm_qset size
net: hns3: fix an interrupt residual problem
net: hns3: store rx VLAN tag offload state for VF
octeon_ep: Fix host hang issue during device reboot
net: fec: ERR007885 Workaround for conventional TX
net: lan743x: Fix memleak issue when GSO enabled
ptp: ocp: Fix NULL dereference in Adva board SMA sysfs operations
net: use sock_gen_put() when sk_state is TCP_TIME_WAIT
bnxt_en: fix module unload sequence
bnxt_en: Fix ethtool -d byte order for 32-bit values
bnxt_en: Fix out-of-bound memcpy() during ethtool -w
bnxt_en: Fix coredump logic to free allocated buffer
bnxt_en: delay pci_alloc_irq_vectors() in the AER path
bnxt_en: call pci_alloc_irq_vectors() after bnxt_reserve_rings()
bnxt_en: Add missing skb_mark_for_recycle() in bnxt_rx_vlan()
...
Stefan Wahren says:
====================
net: vertexcom: mse102x: Fix RX handling
This series is the first part of two series for the Vertexcom driver.
It contains substantial fixes for the RX handling of the Vertexcom MSE102x.
====================
Link: https://patch.msgid.link/20250430133043.7722-1-wahrenst@gmx.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In case the CMD_RTS got corrupted by interferences, the MSE102x
doesn't allow a retransmission of the command. Instead the Ethernet
frame must be shifted out of the SPI FIFO. Since the actual length is
unknown, assume the maximum possible value.
Fixes: 2f207cbf0d ("net: vertexcom: Add MSE102x SPI support")
Signed-off-by: Stefan Wahren <wahrenst@gmx.net>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://patch.msgid.link/20250430133043.7722-5-wahrenst@gmx.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The MSE102x doesn't provide any SPI commands for interrupt handling.
So in case the interrupt fired before the driver requests the IRQ,
the interrupt will never fire again. In order to fix this always poll
for pending packets after opening the interface.
Fixes: 2f207cbf0d ("net: vertexcom: Add MSE102x SPI support")
Signed-off-by: Stefan Wahren <wahrenst@gmx.net>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://patch.msgid.link/20250430133043.7722-2-wahrenst@gmx.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When a VF is passthrough to a VM, and the VM is killed, the reported
interrupt may not been handled, it will remain, and won't be clear by
the nic engine even with a flr or tqp reset. When the VM restart, the
interrupt of the first vector may be dropped by the second enable_irq
in vfio, see the issue below:
https://gitlab.com/qemu-project/qemu/-/issues/2884#note_2423361621
We notice that the vfio has always behaved this way, and the interrupt
is a residue of the nic engine, so we fix the problem by moving the
vector enable process out of the enable_irq loop.
Fixes: 08a100689d ("net: hns3: re-organize vector handle")
Signed-off-by: Yonglong Liu <liuyonglong@huawei.com>
Signed-off-by: Jijie Shao <shaojijie@huawei.com>
Link: https://patch.msgid.link/20250430093052.2400464-3-shaojijie@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The VF driver missed to store the rx VLAN tag strip state when
user change the rx VLAN tag offload state. And it will default
to enable the rx vlan tag strip when re-init VF device after
reset. So if user disable rx VLAN tag offload, and trig reset,
then the HW will still strip the VLAN tag from packet nad fill
into RX BD, but the VF driver will ignore it for rx VLAN tag
offload disabled. It may cause the rx VLAN tag dropped.
Fixes: b2641e2ad4 ("net: hns3: Add support of hardware rx-vlan-offload to HNS3 VF driver")
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Jijie Shao <shaojijie@huawei.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250430093052.2400464-2-shaojijie@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Tony Nguyen says:
====================
Intel Wired LAN Driver Updates 2025-04-29 (idpf, igc)
For idpf:
Michal fixes error path handling to remove memory leak.
Larysa prevents reset from being called during shutdown.
For igc:
Jake adjusts locking order to resolve sleeping in atomic context.
* '200GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
igc: fix lock order in igc_ptp_reset
idpf: protect shutdown from reset
idpf: fix potential memory leak on kcalloc() failure
====================
Link: https://patch.msgid.link/20250429221034.3909139-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When the host loses heartbeat messages from the device,
the driver calls the device-specific ndo_stop function,
which frees the resources. If the driver is unloaded in
this scenario, it calls ndo_stop again, attempting to free
resources that have already been freed, leading to a host
hang issue. To resolve this, dev_close should be called
instead of the device-specific stop function.dev_close
internally calls ndo_stop to stop the network interface
and performs additional cleanup tasks. During the driver
unload process, if the device is already down, ndo_stop
is not called.
Fixes: 5cb96c29aa ("octeon_ep: add heartbeat monitor")
Signed-off-by: Sathesh B Edara <sedara@marvell.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250429114624.19104-1-sedara@marvell.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Activate TX hang workaround also in
fec_enet_txq_submit_skb() when TSO is not enabled.
Errata: ERR007885
Symptoms: NETDEV WATCHDOG: eth0 (fec): transmit queue 0 timed out
commit 37d6017b84 ("net: fec: Workaround for imx6sx enet tx hang when enable three queues")
There is a TDAR race condition for mutliQ when the software sets TDAR
and the UDMA clears TDAR simultaneously or in a small window (2-4 cycles).
This will cause the udma_tx and udma_tx_arbiter state machines to hang.
So, the Workaround is checking TDAR status four time, if TDAR cleared by
hardware and then write TDAR, otherwise don't set TDAR.
Fixes: 53bb20d1fa ("net: fec: add variable reg_desc_active to speed things up")
Signed-off-by: Mattias Barthel <mattias.barthel@atlascopco.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://patch.msgid.link/20250429090826.3101258-1-mattiasbarthel@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
On Adva boards, SMA sysfs store/get operations can call
__handle_signal_outputs() or __handle_signal_inputs() while the `irig`
and `dcf` pointers are uninitialized, leading to a NULL pointer
dereference in __handle_signal() and causing a kernel crash. Adva boards
don't use `irig` or `dcf` functionality, so add Adva-specific callbacks
`ptp_ocp_sma_adva_set_outputs()` and `ptp_ocp_sma_adva_set_inputs()` that
avoid invoking `irig` or `dcf` input/output routines.
Fixes: ef61f5528f ("ptp: ocp: add Adva timecard support")
Signed-off-by: Sagi Maimon <maimon.sagi@gmail.com>
Reviewed-by: Vadim Fedorenko <vadim.fedorenko@linux.dev>
Link: https://patch.msgid.link/20250429073320.33277-1-maimon.sagi@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Clang and GCC have different behaviors around disabling warnings
included in -Wall and -Wextra and the order in which flags are
specified, which is exposed by clang's new support for
-Wunterminated-string-initialization.
$ cat test.c
const char foo[3] = "FOO";
const char bar[3] __attribute__((__nonstring__)) = "BAR";
$ clang -fsyntax-only -Wextra test.c
test.c:1:21: warning: initializer-string for character array is too long, array size is 3 but initializer has size 4 (including the null terminating character); did you mean to use the 'nonstring' attribute? [-Wunterminated-string-initialization]
1 | const char foo[3] = "FOO";
| ^~~~~
$ clang -fsyntax-only -Wextra -Wno-unterminated-string-initialization test.c
$ clang -fsyntax-only -Wno-unterminated-string-initialization -Wextra test.c
test.c:1:21: warning: initializer-string for character array is too long, array size is 3 but initializer has size 4 (including the null terminating character); did you mean to use the 'nonstring' attribute? [-Wunterminated-string-initialization]
1 | const char foo[3] = "FOO";
| ^~~~~
$ gcc -fsyntax-only -Wextra test.c
test.c:1:21: warning: initializer-string for array of ‘char’ truncates NUL terminator but destination lacks ‘nonstring’ attribute (4 chars into 3 available) [-Wunterminated-string-initialization]
1 | const char foo[3] = "FOO";
| ^~~~~
$ gcc -fsyntax-only -Wextra -Wno-unterminated-string-initialization test.c
$ gcc -fsyntax-only -Wno-unterminated-string-initialization -Wextra test.c
Move -Wextra up right below -Wall in Makefile.extrawarn to ensure these
flags are at the beginning of the warning options list. Move the couple
of warning options that have been added to the main Makefile since
commit e88ca24319 ("kbuild: consolidate warning flags in
scripts/Makefile.extrawarn") to scripts/Makefile.extrawarn after -Wall /
-Wextra to ensure they get properly disabled for all compilers.
Fixes: 9d7a0577c9 ("gcc-15: disable '-Wunterminated-string-initialization' entirely for now")
Link: https://github.com/llvm/llvm-project/issues/10359
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull btrfs fixes from David Sterba:
- fix potential inode leak in iget() after memory allocation failure
- in subpage mode, fix extent buffer bitmap iteration when writing out
dirty sectors
- fix range calculation when falling back to COW for a NOCOW file
* tag 'for-6.15-rc4-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
btrfs: adjust subpage bit start based on sectorsize
btrfs: fix the inode leak in btrfs_iget()
btrfs: fix COW handling in run_delalloc_nocow()
Pull modules fixes from Petr Pavlu:
"A single series to properly handle the module_kobject creation.
This fixes a problem with missing /sys/module/<module>/drivers for
built-in modules"
* tag 'modules-6.15-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/modules/linux:
drivers: base: handle module_kobject creation
kernel: globalize lookup_or_create_module_kobject()
kernel: refactor lookup_or_create_module_kobject()
kernel: param: rename locate_module_kobject
Michael Chan says:
====================
bnxt_en: Misc. bug fixes
This series fixes a bug in the driver initialization path, MSIX
setup sequencing issue in the FW error and AER paths, a missing
skb_mark_for_recycle() in the VLAN error path, some ethtool coredump
fixes, an ethtool selftest fix, and an ethtool register dump byte order
fix.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
For version 1 register dump that includes the PCIe stats, the existing
code incorrectly assumes that all PCIe stats are 64-bit values. Fix it
by using an array containing the starting and ending index of the 32-bit
values. The loop in bnxt_get_regs() will use the array to do proper
endian swap for the 32-bit values.
Fixes: b5d600b027 ("bnxt_en: Add support for 'ethtool -d'")
Reviewed-by: Shruti Parab <shruti.parab@broadcom.com>
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Andy Gospodarek <andrew.gospodarek@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When retrieving the FW coredump using ethtool, it can sometimes cause
memory corruption:
BUG: KFENCE: memory corruption in __bnxt_get_coredump+0x3ef/0x670 [bnxt_en]
Corrupted memory at 0x000000008f0f30e8 [ ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ] (in kfence-#45):
__bnxt_get_coredump+0x3ef/0x670 [bnxt_en]
ethtool_get_dump_data+0xdc/0x1a0
__dev_ethtool+0xa1e/0x1af0
dev_ethtool+0xa8/0x170
dev_ioctl+0x1b5/0x580
sock_do_ioctl+0xab/0xf0
sock_ioctl+0x1ce/0x2e0
__x64_sys_ioctl+0x87/0xc0
do_syscall_64+0x5c/0xf0
entry_SYSCALL_64_after_hwframe+0x78/0x80
...
This happens when copying the coredump segment list in
bnxt_hwrm_dbg_dma_data() with the HWRM_DBG_COREDUMP_LIST FW command.
The info->dest_buf buffer is allocated based on the number of coredump
segments returned by the FW. The segment list is then DMA'ed by
the FW and the length of the DMA is returned by FW. The driver then
copies this DMA'ed segment list to info->dest_buf.
In some cases, this DMA length may exceed the info->dest_buf length
and cause the above BUG condition. Fix it by capping the copy
length to not exceed the length of info->dest_buf. The extra
DMA data contains no useful information.
This code path is shared for the HWRM_DBG_COREDUMP_LIST and the
HWRM_DBG_COREDUMP_RETRIEVE FW commands. The buffering is different
for these 2 FW commands. To simplify the logic, we need to move
the line to adjust the buffer length for HWRM_DBG_COREDUMP_RETRIEVE
up, so that the new check to cap the copy length will work for both
commands.
Fixes: c74751f4c3 ("bnxt_en: Return error if FW returns more data than dump length")
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Signed-off-by: Shruti Parab <shruti.parab@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When handling HWRM_DBG_COREDUMP_LIST FW command in
bnxt_hwrm_dbg_dma_data(), the allocated buffer info->dest_buf is
not freed in the error path. In the normal path, info->dest_buf
is assigned to coredump->data and it will eventually be freed after
the coredump is collected.
Free info->dest_buf immediately inside bnxt_hwrm_dbg_dma_data() in
the error path.
Fixes: c74751f4c3 ("bnxt_en: Return error if FW returns more data than dump length")
Reported-by: Michael Chan <michael.chan@broadcom.com>
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Signed-off-by: Shruti Parab <shruti.parab@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch is similar to the last patch to delay the
pci_alloc_irq_vectors() call in the AER path until after calling
bnxt_reserve_rings(). bnxt_reserve_rings() needs to properly map
the MSIX table first before we call pci_alloc_irq_vectors() which
may immediately write to the MSIX table in some architectures.
Move the bnxt_init_int_mode() call from bnxt_io_slot_reset() to
bnxt_io_resume() after calling bnxt_reserve_rings().
With this change, the AER path may call bnxt_open() ->
bnxt_hwrm_if_change() with bp->irq_tbl set to NULL. bp->irq_tbl is
cleared when we call bnxt_clear_int_mode() in bnxt_io_slot_reset().
So we cannot use !bp->irq_tbl to detect aborted FW reset. Add a
new BNXT_FW_RESET_STATE_ABORT to detect aborted FW reset in
bnxt_hwrm_if_change().
Signed-off-by: Kashyap Desai <kashyap.desai@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
On some architectures (e.g. ARM), calling pci_alloc_irq_vectors()
will immediately cause the MSIX table to be written. This will not
work if we haven't called bnxt_reserve_rings() to properly map
the MSIX table to the MSIX vectors reserved by FW.
Fix the FW error recovery path to delay the bnxt_init_int_mode() ->
pci_alloc_irq_vectors() call by removing it from bnxt_hwrm_if_change().
bnxt_request_irq() later in the code path will call it and by then the
MSIX table is properly mapped.
Fixes: 4343838ca5 ("bnxt_en: Replace deprecated PCI MSIX APIs")
Suggested-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: Kashyap Desai <kashyap.desai@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If bnxt_rx_vlan() fails because the VLAN protocol ID is invalid,
the SKB is freed but we're missing the call to recycle it. This
may cause the warning:
"page_pool_release_retry() stalled pool shutdown"
Add the missing skb_mark_for_recycle() in bnxt_rx_vlan().
Fixes: 86b05508f7 ("bnxt_en: Use the unified RX page pool buffers for XDP and non-XDP")
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When RDMA driver is loaded, running offline self test is not
supported and driver returns failure early. But it is not clearing
the input buffer and hence the application prints some junk
characters for individual test results.
Fix it by clearing the buffer before returning.
Fixes: 895621f1c8 ("bnxt_en: Don't support offline self test when RoCE driver is loaded")
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
WARN_ON() is triggered in __flush_work() if bnxt_init_chip() fails
because we call cancel_work_sync() on dim work that has not been
initialized.
WARNING: CPU: 37 PID: 5223 at kernel/workqueue.c:4201 __flush_work.isra.0+0x212/0x230
The driver relies on the BNXT_STATE_NAPI_DISABLED bit to check if dim
work has already been cancelled. But in the bnxt_open() path,
BNXT_STATE_NAPI_DISABLED is not set and this causes the error
path to think that it needs to cancel the uninitalized dim work.
Fix it by setting BNXT_STATE_NAPI_DISABLED during initialization.
The bit will be cleared when we enable NAPI and initialize dim work.
Fixes: 40452969a5 ("bnxt_en: Fix DIM shutdown")
Suggested-by: Somnath Kotur <somnath.kotur@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Signed-off-by: Shravya KN <shravya.k-n@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull crypto fix from Herbert Xu:
"This fixes a regression in scompress"
* tag 'v6.15-p6' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6:
crypto: scompress - increment scomp_scratch_users when already allocated
