Daily update · Oct 8–9, 2026
Linux mainline update: WireGuard handshake fix, ASUS Wi-Fi power restored, ext4 data=journal deprecated
A networking-heavy day in mainline: the kernel stops sending uninitialized bytes onto the wire, MediaTek Wi-Fi gets a batch of fixes, Raspberry Pi Ethernet crashes are patched, and Intel SR-IOV bonding sheds a 10-second stall.
In brief
Today's mainline pull is dominated by networking fixes spanning the core stack, tunnels, wireless, and a wide range of NIC drivers. On the security side, the network stack no longer transmits stale buffer memory in ICMP replies and TX bounce buffers, IPv4 now validates checksum metadata before completing checksums, an unterminated net/sched algorithm name can no longer leak kernel stack bytes into module-loading command lines, and Xen netfront drops malformed backend packets instead of crashing. WireGuard received a fix for a handshake race that could resurrect superseded session state, and a bareudp GSO bug that throttled IPv6-over-bareudp from roughly 700 Mbit/s to 10 Mbit/s was corrected. MediaTek Wi-Fi hardware picked up multiple-link stability fixes, restoration of correct TX power on ASUS MT7922/MT7925 laptops, and a fix that lets mt7925 associate to normal networks after leaving a multi-link one. Raspberry Pi hardware benefits from Cadence MACB corruption fixes that could kill a Pi CM5 within seconds under pktgen and a bcmgenet fix that permanently disabled TX after Wake-on-LAN suspend/resume. Intel users get faster SR-IOV bonding setup on i40e/ice, an iavf MAC-change deadlock fix, and an e1000e quirk that eliminated 40% packet loss on an affected system. Azure VMs on MANA regain wire-speed forwarding, and ext4's data=journal mode is now formally deprecated with removal planned for January 2028.
Security and hardening
Network stack no longer transmits stale buffer memory
When skb_copy_and_csum_bits() hit unreadable frags it copied only the linear part and left the rest of the caller's buffer untouched, so uninitialized bytes could go out on the wire in ICMP error messages or driver TX bounce buffers. The unfilled portion is now zeroed.
Why it matters: Prevents uninitialized kernel memory from leaking onto the network via ICMP replies and transmit paths.
IPv4 validates checksum metadata before completing checksums
A packet with bad checksum metadata entering the IPv4 stack could cause skb_checksum_help to corrupt the network header. The stack now checks checksum_start before completing the checksum, as defence in depth until fuller input validation lands.
Why it matters: Hardens the kernel against packets with crafted checksum metadata, for example injected via tun or packet sockets, that could otherwise corrupt headers.
net/sched em_text no longer leaks stack bytes into module loader
em_text_change() passed a 16-byte algo array with no enforced NUL terminator to textsearch_prepare(), which on a TS_AUTOLOAD retry called request_module() with vsnprintf walking past the array into adjacent struct fields, feeding those bytes to the usermode helper command line.
Why it matters: A malicious netlink message could previously leak kernel stack contents into module-loading command lines.
Xen netfront hardened against malformed backend packets
Backend-controlled RX slot lengths shorter than an Ethernet header triggered a BUG() or an out-of-bounds read in eth_type_trans(); short packets are now dropped. A related fix frees the skb when a response chain overflows the frags array, which previously leaked the skb and its pages.
Why it matters: Xen guest kernels no longer crash or leak memory when a buggy or malicious backend sends malformed or oversized packet chains.
Networking core
IPv4 TCP SYN processing use-after-free fixed
tcp_v4_syn_recv_sock() transferred ireq_opt ownership to the child socket without copying, but a concurrent retransmitted SYN could read the same pointer under RCU only. The fix defers freeing via call_rcu.
Why it matters: High-connection-rate TCP servers could hit use-after-free races during SYN processing.
TCP zerocopy receive NULL pointer dereference on net_iov fragments
tcp_zerocopy_receive() could pass a net_iov fragment to can_map_frag(), which dereferenced a NULL page pointer from skb_frag_page(), causing a kernel panic.
Why it matters: TCP zerocopy receive on connections with dmabuf-backed pages could crash the kernel.
Open vSwitch soft lockup and panic during netlink flow dump
ovs_flow_cmd_dump() released stats->lock with spin_unlock_bh() once per CPU, each time running pending softirq backlog. On a busy CPU the backlog refilled as fast as it drained, so the dumping thread never slept and hit a soft lockup.
Why it matters: Dumping OVS flow tables on busy production hosts could lock up and panic the kernel.
Bonding XDP program reference leak on replacement or removal
When an XDP program was replaced or removed on a bonding interface, the old BPF program reference was not released, causing leaked programs to accumulate.
Why it matters: Repeatedly changing XDP programs on bond interfaces leaked memory and BPF program references.
Flow dissector u16 truncation caused bogus transport header offsets
The flow dissector truncated skb->len to u16 when computing the transport header offset, so packets larger than 65535 bytes got bogus small thoff values, pointing transport headers inside Ethernet headers.
Why it matters: Large packets, such as AF_PACKET with VNET_HDR and TSO frames, could be misclassified or mishandled by traffic classification, GSO validation, and other flow-dissection users.
IPv4 NULL derefs in PMTU and FIB dump under nexthop group replacement
Multiple IPv4 code paths independently loaded nh->nh_grp without ensuring consistency, so if a nexthop group was replaced with fewer paths between loads, fib_info_nhc() returned NULL and was dereferenced.
Why it matters: Concurrent nexthop group replacement operations could crash the kernel during PMTU updates, FIB dumps, or route notification sizing.
Tunnels and VPN
WireGuard handshake race fixed
Two threads processing a duplicate handshake response, interleaved with a subsequently queued initiation, could copy superseded handshake state back to the peer and start a new session from it. Response consumption after an intermediate initiation is now rejected.
Why it matters: WireGuard tunnel setup is now reliable under retransmitted or concurrently processed handshake messages.
bareudp GSO segmentation failure throttled IPv6 tunneling
bareudp did not correctly set the inner protocol for GSO segmentation, causing software segmentation to fail and drop packets, degrading IPv6-over-bareudp throughput from roughly 700 Mbit/s to about 10 Mbit/s.
Why it matters: Users tunneling traffic over bareudp, especially IPv6, saw dramatic throughput loss and thousands of retransmissions.
ip6_gre ERSPAN PMTU path crashable by unprivileged users
A legacy PMTU update in ip6erspan_tunnel_xmit() treated a metadata dst from tc tunnel_key as a route; since it had no output device, dst_dev(dst)->mtu dereferenced NULL and panicked the kernel.
Why it matters: A crash could be triggered from an unprivileged user in a network namespace using tc tunnel_key and mirred to redirect to an ip6erspan device.
Wi-Fi
mt76 multi-link (MLO) stability fixes for MediaTek hardware
mt7996 did not set the CONNECTION_MONITOR flag, so mac80211's legacy beacon monitor ran alongside the driver's own and removed associations still carrying traffic. Separately, for MLO peers, hardware completions arriving on a different link than the queuing link let the non-AQL packet counter drift upward until it blocked the station entirely.
Why it matters: Multi-link Wi-Fi on MediaTek hardware no longer randomly drops associations while traffic flows on other links, and MLO stations no longer eventually stop transmitting.
ASUS firmware SAR tables no longer clamp MediaTek TX power to minimum
Some ASUS firmware carries 0xff in ACPI SAR power slots, which the driver narrowed to s8 (-1), letting it win the min() against regulatory limits and cap every TX rate. SAR tables were also parsed with a version byte from a different ACPI table, dropping 6 GHz limits.
Why it matters: ASUS laptops with MT7922/MT7925 Wi-Fi, such as the Zenbook 14 and Vivobook S 14, suffered severely reduced TX power and throughput.
mt7925 legacy association failure after MLO teardown
Removing the last link of an MLD interface left stale firmware BSS/DEV entries with per-link MAC addresses, causing every subsequent legacy association to fail authentication until the interface was removed or the module reloaded.
Why it matters: MediaTek mt7925 users who disconnect from a multi-link network could not connect to any normal non-MLO network without reloading the driver.
brcmfmac P2P device double-removal crash on driver detach
When the driver was removed while wpa_supplicant was still exiting, the P2P device interface was removed twice, causing a kernel crash due to a race between brcmf_detach() and the userspace DEL_INTERFACE path.
Why it matters: Broadcom brcmfmac Wi-Fi users could hit kernel crashes when removing the driver while wpa_supplicant was still running.
Bluetooth
SCO connection use-after-free on receive path
sco_recv_scodata() could obtain a reference to an sco_conn that was concurrently freed, because the back-pointer was cleared without holding the device lock. This triggered KASAN slab-use-after-free reports.
Why it matters: Bluetooth SCO audio connections could crash the kernel under connection teardown races.
RFCOMM fixes: skb leak, deadlock, and NULL dereference
Three RFCOMM bugs were fixed: an skb leak when a DLC has no owner, a circular lock dependency between rfcomm_mutex and hdev->lock during connect, and a NULL tty_dev dereference in rfcomm_dev_shutdown when tty registration fails.
Why it matters: Bluetooth serial (RFCOMM) connections are more stable and no longer leak memory, deadlock, or crash under edge cases.
HCI socket race crash on controller unregistration
Rebinding an HCI socket after its controller was unregistered could race with monitor control replay, causing a NULL pointer dereference or use-after-free in create_monitor_ctrl_open().
Why it matters: Hot-removing Bluetooth controllers while sockets are active no longer crashes the kernel.
CAN bus
CAN RX skb header regression caused kernel panic via SOCK_PACKET
A regression left CAN RX skbs with an uninitialized mac_header sentinel, causing packet_rcv_spkt() to overflow its push-length calculation and trigger a full kernel panic.
Why it matters: Using legacy SOCK_PACKET sockets with CAN interfaces could panic the machine.
CAN unique skb identifier regression dropped legitimate frames under RPS
With RPS enabled, the flow dissector assigned every CAN frame the same non-zero hash, causing raw_rcv() to mistake legitimate frames for duplicates and drop them. The CAN UID now lives in a dedicated skb extension.
Why it matters: CAN frames could be silently dropped on multi-CPU systems with RPS, breaking CAN communication.
Intel wired networking
i40e and ice skip unnecessary VF reset when changing trust setting
Changing VF trust unconditionally triggered a VF reset, causing a roughly 10-second delay during bonding setup. The reset is now skipped when granting trust, and only performed on revocation when advanced features are configured.
Why it matters: Bonding setup on Intel i40e/ice SR-IOV VFs becomes significantly faster, avoiding roughly 10-second stalls.
iavf MAC address change deadlock with netdev instance lock
After the netdev instance lock was held during sysfs operations, iavf_set_mac() deadlocked because the watchdog and adminq tasks also needed the same lock. The MAC change request is now sent synchronously instead of being queued.
Why it matters: Changing the MAC address on Intel iavf (VF) interfaces no longer deadlocks the system.
e1000e disables K1 power saving on system with 40% packet loss
A specific system experienced 40% packet loss due to K1 power saving re-enablement; adding it to the disable list resolves the issue.
Why it matters: Users of the affected system with Intel e1000e networking regain normal LAN reliability.
Ethernet and NIC drivers
bnxt_en DMA mapping overrun on small padded packets
Packets smaller than 52 bytes, such as untagged ARP frames, were padded but the DMA mapping length was not increased, causing DMA read faults on page boundaries, especially on VLAN/macvlan interfaces with Intel IOMMU.
Why it matters: Broadcom NetXtreme-E NICs could experience TX queue timeouts and DMA faults when sending small packets like ARP on VLAN interfaces.
bnxt_en driver init fixed in kdump kernels behind PCIe switches
When the NIC is behind a PCIe switch in smart mode, the switch may need to see BARs initialized before passing memory read/write TLPs, causing fatal AER errors during kdump kernel initialization.
Why it matters: Systems with Broadcom bnxt_en NICs behind certain PCIe switches could fail to initialize networking in kdump/crashdump kernels, hampering network-based crash recovery.
mlx5e completion-queue doorbell ordering fixed
The driver updated the ICOSQ consumer counter before the CQ doorbell record, violating an ordering requirement that can cause a CQ overrun. The doorbell is now updated first, with a dma_wmb() before the consumer counter.
Why it matters: Prevents rare completion-queue overruns that can disrupt receive processing on Mellanox/NVIDIA ConnectX NICs.
stmmac 802.1ad (QinQ) VLAN stripping fixed
vlan_rx_hw() hardcoded ETH_P_8021Q instead of using the actual packet protocol, causing QinQ packets to be misidentified. Separately, rx-vlan-stag-hw-parse was advertised as enabled even when VLAN stripping was disabled.
Why it matters: QinQ networking on stmmac-based interfaces now works correctly, and ethtool feature reporting is accurate.
Cadence MACB crash and corruption fixes (hits Raspberry Pi CM5)
The driver modified skbs before checking TX ring space, so a NETDEV_TX_BUSY requeue could handle an already-freed or grown skb; it also appended the FCS in place to shared skbs, making pktgen clone_skb traffic hit a BQL BUG_ON that killed a Raspberry Pi CM5 within seconds. The ring is now checked first and shared skbs are copied.
Why it matters: Fixes crashes and memory corruption on systems using Cadence GEM/MACB Ethernet, including the Raspberry Pi CM5 via RP1.
bcmgenet TX permanently disabled after WoL suspend/resume
When the PHY link was down and UMAC was held in reset with WoL enabled, the resume sequence left UMAC_TX never enabled again on link up, permanently disabling transmission.
Why it matters: Raspberry Pi and other bcmgenet users could lose TX capability entirely after suspend/resume when WoL was enabled with the link down.
ASIX ax88179/178A USB Ethernet delivered frames with extra trailing bytes
A previous fix replaced skb_clone() with a copy that included a 2-byte IP alignment pseudo header, delivering non-last packets in bulk transfers with two extra trailing bytes, breaking MACsec and other length-sensitive protocols.
Why it matters: USB Ethernet adapters using the ASIX ax88179/178A chipset could deliver frames with incorrect lengths, breaking MACsec.
Virtualization and cloud
Azure MANA forwarding slowdown fixed
A memory-efficiency change handed out RX buffers with zero headroom, so every forwarded packet failed the skb_cow() check in ip_forward() and had to be reallocated and copied by pskb_expand_head(). RX buffers now reserve the needed headroom.
Why it matters: Restores wire-speed forwarding performance for virtual machines using Microsoft's MANA network interfaces on Azure.
virtio_net GSO packets always dissected for correct validation
A check added to avoid false positives in GSO validation inadvertently bypassed flow dissection for GSO packets from tun, virtio_net, and raw sockets because skb->network_header was zero-initialized, causing VLAN-tagged GSO packets to be rejected or mishandled.
Why it matters: Virtualization and tunneling users could experience dropped or missegmented GSO packets, especially with VLAN tagging.
vsock/virtio rx_bytes underflow broke stream reads
After a partial stream receive, virtio_transport_read_skb() passed the full packet length instead of the unread suffix, underflowing rx_bytes and causing SIOCINQ to report negative values, empty recv to return ELOOP, and poll to report empty sockets as readable.
Why it matters: vsock stream users could see broken receive behavior after partial reads, with data appearing unavailable or poll reporting readiness incorrectly.
Filesystems
ext4 data=journal mode marked deprecated
ext4's data=journal mount option, which journals file data as well as metadata, is now formally deprecated, with documentation noting it will be removed in January 2028.
Why it matters: Users or distros relying on full data journaling in ext4 need to plan a migration to another mode or filesystem before the removal date.
Source commits130 entries +
i40e: skip unnecessary VF reset when setting trust
Jose Ignacio Tornos Martinez · Apr 28, 2026 · 1 files
Revert "wifi: ath12k: add panic handler"
Yingying Tang · Jun 12, 2026 · 4 files
ice: skip unnecessary VF reset when setting trust
Jose Ignacio Tornos Martinez · Jun 15, 2026 · 1 files
iavf: send MAC change request synchronously
Jose Ignacio Tornos Martinez · Jun 22, 2026 · 3 files
wifi: nxpwifi: protect sta_list against concurrent add and delete
Linmao Li · Aug 7, 2026 · 2 files
wifi: nxpwifi: delete the station entry on the uAP deauth event
Linmao Li · Aug 7, 2026 · 1 files
wifi: mt76: mt7915: disable rx napi when removing device
Nicolas Cavallari · Aug 10, 2026 · 1 files
wifi: nxpwifi: wait for the wakeup timer before the adapter is freed
Linmao Li · Aug 13, 2026 · 1 files
wifi: nxpwifi: free the aggregation buffer when the RA list disappears
Linmao Li · Aug 13, 2026 · 1 files
wifi: nxpwifi: zero the channel statistics array
Linmao Li · Aug 13, 2026 · 2 files
wifi: mt76: mt7996: program a link again if the driver holds it
Felix Fietkau · Aug 18, 2026 · 1 files
wifi: mt76: mt7996: take over connection monitoring
Felix Fietkau · Aug 18, 2026 · 1 files
wifi: mt76: account non-AQL frames per peer rather than per link
Felix Fietkau · Aug 18, 2026 · 2 files
wifi: nxpwifi: fix the authentication frame length handling
Linmao Li · Aug 20, 2026 · 1 files
wifi: nxpwifi: handle authentication frame allocation failures
Linmao Li · Aug 20, 2026 · 1 files
wifi: mt76: mt7996: fix struct mt7996_mcu_wed_rro_ba_delete_event layout
Tao Gong · Aug 23, 2026 · 1 files
usb: f81604: fix struct f81604_int_data size mismatch
Ji-Ze Hong (Peter Hong) · Aug 24, 2026 · 1 files
wifi: nxpwifi: fix inverted check in Tx BA stream entry deletion
David Carlier · Aug 25, 2026 · 1 files
wifi: nxpwifi: do not delete Rx reorder entries under RCU
David Carlier · Aug 25, 2026 · 1 files
wifi: mt76: mt7925: restore the legacy BSS after MLO teardown
Aaron Ma · Aug 26, 2026 · 1 files
wifi: mt76: use ALTX queue for packets to disassociated stations
Felix Fietkau · Sep 1, 2026 · 3 files
wifi: mt76: mt7603: initialize global station WCID
Cristian Papa · Sep 4, 2026 · 1 files
wifi: mt76: mt792x: pick the SAR table layout from the table itself
Devin Wittmayer · Sep 5, 2026 · 2 files
wifi: mt76: mt792x: treat 0xff ACPI SAR entries as no limit
Junjie Cao · Sep 10, 2026 · 1 files
wifi: mt76: mt7925: don't put the band auto marker in TGID
Devin Wittmayer · Sep 11, 2026 · 1 files
Bluetooth: RFCOMM: free the skb when the DLC has no owner
Maxim Skokov · Sep 15, 2026 · 1 files
can: dev: init_can_skb(): restore skb header initialization
zjamg · Sep 17, 2026 · 1 files
MAINTAINERS: name the ext4 dev branch
Matthias Goergens · Sep 23, 2026 · 1 files
wifi: brcmfmac: fix P2P device removal race in brcmf_detach()
Michele Dionisio · Sep 24, 2026 · 1 files
ext4: mark data=journal as deprecated and will be removed in January 2028.
Theodore Ts'o · Sep 24, 2026 · 2 files
Bluetooth: MGMT: Fix status of pending commands flushed on power off
Iaroslav Voitovych · Sep 24, 2026 · 1 files
ibmveth: h_free logical LAN on open-fail after register
Mingming Cao · Sep 25, 2026 · 1 files
ibmveth: fix TX LTB and filter unwind on open-fail
Mingming Cao · Sep 25, 2026 · 1 files
net: bridge: avoid recursive multicast port cleanup
Zixuan Chai · Sep 26, 2026 · 1 files
Bluetooth: SCO: serialise sco_conn lifetime against sco_recv_scodata()
Aldo Ariel Panzardo · Sep 26, 2026 · 1 files
lib/dim: fix 32-bit overflow in dim_calc_stats() rates
Shashank Mohan Jain · Sep 27, 2026 · 1 files
lib/dim: add KUnit test for dim_calc_stats()
Shashank Mohan Jain · Sep 27, 2026 · 3 files
tipc: destroy topsrv workqueues before closing connections
Yuqi Xu · Sep 28, 2026 · 1 files
net/packet: guard the ll header push in packet_rcv_spkt()
Quchaosheng · Sep 28, 2026 · 1 files
tipc: hold a reference to nodes found by link name
Chengfeng Ye · Sep 28, 2026 · 1 files
net: xilinx: axienet: Free outstanding DMA buffers on dmaengine stop
Suraj Gupta · Sep 28, 2026 · 2 files
amt: send the relay's General Query directly from the receive path
Omar Ramadan · Sep 28, 2026 · 2 files
selftests: net: amt: check that the relay's queries bypass the amt device
Omar Ramadan · Sep 28, 2026 · 1 files
net: stmmac: Remove VLAN perfect matching dead code
Ovidiu Panait · Sep 28, 2026 · 3 files
net: stmmac: Stop toggling the EDVLP bit
Ovidiu Panait · Sep 28, 2026 · 1 files
net: stmmac: Rename double VLAN references to svlan
Ovidiu Panait · Sep 28, 2026 · 5 files
net: stmmac: Do not advertise S-VLAN stripping when it is disabled
Ovidiu Panait · Sep 28, 2026 · 1 files
net: stmmac: Disable S-Tag processing on dwmac4
Ovidiu Panait · Sep 28, 2026 · 1 files
ice: fix use-after-free in dynamic port cleanup
Xuanqiang Luo · Sep 28, 2026 · 1 files
ice: Restore Ordered MMIO Writes for Tx Doorbells
Bryan Fraschetti · Sep 28, 2026 · 1 files
ice: fix metadata_dst refcount handling on representor teardown
Tristan Madani · Sep 28, 2026 · 1 files
Bluetooth: btintel_pcie: Add shared HW reset for clean start
Sai Teja Aluvala · Sep 29, 2026 · 1 files
Bluetooth: btintel_pcie: fix TX descriptor bounds check
Ravindra · Sep 29, 2026 · 1 files
openvswitch: fix soft lockup in the netlink flow dump
Denis V. Lunev · Sep 29, 2026 · 1 files
Bluetooth: RFCOMM: connect the session socket without rfcomm_mutex
Mikhail Gavrilov · Sep 29, 2026 · 2 files
net: stmmac: propagate PTP addend and system time programming errors
Lorenzo Bianconi · Sep 29, 2026 · 2 files
cxgb4/ch_ktls: disable softirqs around tid_list erases
Sang-Hoon Choi · Sep 29, 2026 · 1 files
net: allwinner: remove dmaengine_desc_free()
Frank Li · Sep 29, 2026 · 1 files
sctp: copy zeroed stats to user in sctp_getsockopt_pr_streamstatus() when !streamoute
Hui Peng · Sep 30, 2026 · 1 files
net: airoha: Add retry mechanism to airoha_qdma_set_trtcm_param()
Leto Liu (刘涛) · Sep 30, 2026 · 2 files
net/sched: em_text: reject unterminated algo before autoload
Jamal Hadi Salim · Sep 30, 2026 · 1 files
Bluetooth: btintel_pcie: use managed IRQ teardown
Runyu Xiao · Sep 30, 2026 · 1 files
wifi: mt76: mt7996: fix uninitialized buf read in mt7996_variant_fem_init()
Lu Huang · Sep 30, 2026 · 1 files
tcp: reject net_iov in zerocopy receive mapping hints
Daehyeon Ko · Sep 30, 2026 · 1 files
pfcp: make sure the SEID is linear before reading it
Haishuang Yan · Sep 30, 2026 · 1 files
net/mlx5: Lag, split aggregate speed into oper and max helpers
Or Har-Toov · Sep 30, 2026 · 1 files
gve: fix XSK buffer leak when rings are stopped
Joshua Washington · Sep 30, 2026 · 1 files
gve: fix XSK buffer leak on error descriptor
Joshua Washington · Sep 30, 2026 · 1 files
gve: fix napi_disable deadlock when attempting to disable XSK pools
Joshua Washington · Sep 30, 2026 · 1 files
net: bcmgenet: if UMAC was suspended in SW_RESET, restore it to SW_RESET
Justin Chen · Sep 30, 2026 · 1 files
net: make dev_xdp_sb_prog_count() see programs attached through a link
Jakub Kicinski · Oct 1, 2026 · 1 files
bonding: fix the program leak when XDP is replaced
Jakub Kicinski · Oct 1, 2026 · 1 files
devlink: fix devlink_rel reference leak when notify work is pending
Haishuang Yan · Oct 1, 2026 · 1 files
can: fix unique skb identifier regression under RPS
Oliver Hartkopp · Oct 1, 2026 · 8 files
vsock/virtio: account only unread bytes in read_skb()
Daehyeon Ko · Oct 1, 2026 · 1 files
net/sched: cls_route: reject change with no routing attribute
Victor Nogueira · Oct 1, 2026 · 1 files
selftests: tc-testing: add cls_route no-routing-attribute change tests
Victor Nogueira · Oct 1, 2026 · 1 files
net: team: stop reusing skb after queue override
Weiming Shi · Oct 1, 2026 · 1 files
net: team: free skb when broadcast has no txable port
Weiming Shi · Oct 1, 2026 · 1 files
flow_dissector: avoid u16 truncation of skb->len when computing thoff
Eric Dumazet · Oct 1, 2026 · 1 files
net: always dissect GSO packets in __virtio_net_hdr_to_skb()
Eric Dumazet · Oct 1, 2026 · 4 files
selftests: net: tun: add test for VLAN-tagged GSO without NEEDS_CSUM
Eric Dumazet · Oct 1, 2026 · 1 files
ipv4: free inet_opt and ireq_opt after an RCU grace period
Eric Dumazet · Oct 1, 2026 · 2 files
e1000e: add system to disable K1 list
Tony Nguyen · Oct 1, 2026 · 1 files
ip6_gre: remove obsolete ERSPAN PMTU update
Daehyeon Ko · Oct 2, 2026 · 1 files
ipv4: stop PMTU walk when nexthop group shrinks
Daehyeon Ko · Oct 2, 2026 · 1 files
ipv4: stop exception dump when nexthop group shrinks
Daehyeon Ko · Oct 2, 2026 · 1 files
ipv4: stop route notification sizing when nexthop group shrinks
Daehyeon Ko · Oct 2, 2026 · 1 files
Bluetooth: RFCOMM: Fix NULL tty_dev dereference in rfcomm_dev_shutdown
Palla Raghunath · Oct 2, 2026 · 1 files
net: usb: ax88179_178a: fix rx frame length for non-last packets
Fredrik Nyberg · Oct 2, 2026 · 1 files
bareudp: set the inner protocol to the protocol of the packet
Haishuang Yan · Oct 2, 2026 · 1 files
net: mana: reserve RX buffer headroom to fix forwarding performance
Hamza Mahfooz · Oct 3, 2026 · 4 files
Revert "net/mlx5: E-Switch, preserve max tx speed on vport state modification"
Or Har-Toov · Oct 4, 2026 · 2 files
Bluetooth: hci_sock: Serialize dead-device detachment
Chengfeng Ye · Oct 4, 2026 · 1 files
Bluetooth: ISO: Reject concurrent BIS listener setup
Chengfeng Ye · Oct 4, 2026 · 1 files
Bluetooth: hci_sync: Fix command skb lifetime during scan setup
Chengfeng Ye · Oct 4, 2026 · 2 files
net: sparx5: start the TOD counters on non-PTP lan969x variants
Quentin Freimanis · Oct 5, 2026 · 1 files
Bluetooth: ISO: Serialize concurrent connect calls
Chengfeng Ye · Oct 5, 2026 · 1 files
strparser: make sure __strp_recv isn't running before tearing down the parser
Sabrina Dubroca · Oct 5, 2026 · 3 files
bnxt_en: Add bnxt_clear_bars() helper
Michael Chan · Oct 5, 2026 · 1 files
bnxt_en: Fix driver init in kdump kernel
Michael Chan · Oct 5, 2026 · 1 files
bnxt_en: Re-write the BARs following any type of PCIe errors
Pavan Chebbi · Oct 5, 2026 · 2 files
Revert "net: stmmac: propagate PTP addend and system time programming errors"
Jakub Kicinski · Oct 6, 2026 · 2 files
cipso: adjust cached option offsets when removing CIPSO
Daehyeon Ko · Oct 6, 2026 · 1 files
net: macb: check TX ring before modifying skb
Nicolai Buchwitz · Oct 6, 2026 · 1 files
net: macb: copy shared skbs before appending the FCS
Nicolai Buchwitz · Oct 6, 2026 · 1 files
wifi: mac80211: don't estimate airtime for unsupported rate widths
François Roux · Oct 6, 2026 · 1 files
net/mlx5e: Order ICOSQ cc update after CQ doorbell
Li RongQing · Oct 6, 2026 · 1 files
mlxsw: spectrum_flower: Fix port range register leak in tmplt_create()
Petr Machata · Oct 6, 2026 · 3 files
selftests: mlxsw: Test port range occupancy on template create
Petr Machata · Oct 6, 2026 · 1 files
veth: fix peer NETDEV_XDP_ACT_NDO_XMIT after GRO is toggled while down
Tianyi Gao · Oct 6, 2026 · 1 files
selftests: net: veth: test peer ndo-xmit after GRO toggle while down
Tianyi Gao · Oct 6, 2026 · 1 files
flow_dissector: avoid u16 truncation of hlen in bpf_flow_dissect()
Eric Dumazet · Oct 7, 2026 · 1 files
bnxt_en: fix DMA mapping length for padded small packets
Eric Dumazet · Oct 7, 2026 · 1 files
net: dsa: microchip: fix KSZ8765 fiber detection
Sebastien Royen · Oct 7, 2026 · 1 files
net: sparx5: free the matchall entry on destroy
Daniel Machon · Oct 7, 2026 · 1 files
net: skbuff: don't leave stale bytes in skb_copy_and_csum_bits()
Josef Bacik · Oct 7, 2026 · 1 files
xen/netfront: drop RX packets with a short Ethernet header
Josef Bacik · Oct 7, 2026 · 1 files
net/packet: call packet_parse_headers after virtio_net_hdr_to_skb
Willem de Bruijn · Oct 7, 2026 · 1 files
xen/netfront: don't leak the skb when xennet_fill_frags() fails
Josef Bacik · Oct 7, 2026 · 1 files
ptp: ocp: fix PCIe delay estimation calculation
Vadim Fedorenko · Oct 7, 2026 · 1 files
Revert "wifi: libertas: reject short monitor TX frames"
Johannes Berg · Oct 7, 2026 · 1 files
ipv4: validate checksum_start before completing checksum
Michael S. Tsirkin · Oct 7, 2026 · 6 files
ipv4: do not warn on route notification size race
Daehyeon Ko · Oct 8, 2026 · 1 files
ipv6: do not warn on route notification size race
Daehyeon Ko · Oct 8, 2026 · 1 files
net/smc: protect clcsock lifetime in smc_getname
Chengfeng Ye · Oct 8, 2026 · 1 files
net: openvswitch: validate transport header presence in set_ipv6_addr
Fernando Fernandez Mancera · Oct 8, 2026 · 1 files
wireguard: queueing: preserve tstamp_type when encapsulating packet
Ramses de Norre · Oct 8, 2026 · 1 files
wireguard: noise: reject response consumption after intermediate initiation
Jason A. Donenfeld · Oct 8, 2026 · 1 files
vsock: Fix memory leak in vmci_transport_recv_dgram_cb()
Ilia Gavrilov · Oct 8, 2026 · 1 files