$ the-wire · showcase
net-next 7.4 lands, keys stop leaking kernel addresses, Omarchy virtualizes huge tailnets
By RepoJournal · Filed · About Linux · Composed from the cited sources · methodology
The 7.4 merge window brings a large networking pull with deferred driver fixes, a keyring fix that stops leaking local kernel addresses through index values, and an Omarchy change that cuts a ~4.6s shell stall per poll on large tailnets.
Merge tag 'keys-v7.4-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/jarkko/linux-tpmdd torvalds/linux
Jarkko Sakkinen's keys pull stops the associative-array index from being computed over the description, type and domain addresses, so a sequence of dummy keys can no longer be used to leak local kernel addresses; the change also complements the ongoing rhashtable migration.
Merge tag 'net-next-7.4' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next torvalds/linux
Linus merged net-next 7.4, which carries a significant number of low-priority fixes the last cycle's RCs deferred, especially in NIC drivers, alongside continued work on removing rtnl_lock dependencies such as per-netns ARP and ND neighbour tables.
tailscale: virtualize the machine list so huge tailnets don't melt the shell (#7623) basecamp/omarchy
The tailscale widget now virtualizes the machine list; on a corporate tailnet with ~3,900 machines, `tailscale status --json` is ~4.6MB and the 30s poll rebuilt one eager row per peer, roughly 500k QML object creations per refresh and ~15% average CPU. Large tailnets previously made the shell unusable.
tipc: fix broadcast sender hang on last acker departure torvalds/linux
by Henry Martin
Henry Martin's TIPC fix restores *grp->open when the last outstanding broadcast acker departs; tipc_group_delete_member() zeroed grp->bc_ackers without reopening the group, so poll(EPOLLOUT) stopped reporting writability on the sender.
Merge tag 'hardening-v7.4-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/kees/linux torvalds/linux
Kees Cook's hardening pull adds a kmalloc_obj*-style API for devm_kmalloc(), improves seq_buf for saner C string handling along with several bug fixes from Cook and Bill Wendling, and unexports native_write_cr4(); the powerpc pull adds coregroup support on PowerNV and moves __softirq_pending out of irq_stat.
