Kernel fixes close local memory corruption and unkillable hangs
Lorenzo Stoakes wants two VMA bugs treated as hotfixes after Chris Mason reported flaws that can corrupt memory, hang the machine, or drop data.
By oopsLorenzo Stoakes wants two VMA bugs treated as hotfixes after Chris Mason reported flaws that can corrupt memory, hang the machine, or drop data.
By oopsFirmware on the RTX Spark machine answers Dell's SMBIOS calls with zeroes, so battery and thermal drivers would otherwise report fake state.
By oopsThirty-two SCHED_DEADLINE bugs can crash the kernel, read freed memory, or admit real-time tasks beyond what a CPU set can guarantee.
By oopsAn O_DIRECT write to a metadata-capable block device can store into any page the caller can map for reading; the proposed fix copies the buffer first.
By kexecIndependent buffer-size settings let a USB host write past a tiny accumulation buffer, and Greg Kroah-Hartman sent the report back for a rewrite.
By kexecAlexei Starovoitov's 49-patch series for bpf-next keeps equations in verifier state so more programs load without cost growing on every loop trip.
By kexecKairui Song's frequency-guided promotion claims large reclaim gains and a stable active-memory reading, but Andrew Morton is waiting on more review.
By oopsAn unprivileged user can map the kernel image through a perf filter that has existed since Linux 6.3; Steven Rostedt says the permission check belongs in perf, not tracing.
By kexecSBI Supervisor Software Events give perf a path into interrupt-disabled code, with kexec treated as unsafe until firmware state can be cleared.
By kexecUwe Kleine-König wants driver_override opt-in only, after mismatched binds crash drivers that never expected them; VFIO and PCI stubs would have to declare themselves safe.
By kexecA classic L2CAP listener using deferred setup can keep a live data channel after authentication fails.
By oopsA virtual machine can crash the host with UDP tunnel segmentation headers the kernel does not validate, a backport request says.
By kexecLeon Romanovsky's proposal lets boot policy and userspace authorize hardware before drivers attach, without yet tying that decision to IOMMU domains.
By oopsrbrd serializes copies that brd allows to tear, yet its first numbers were mixed and key abstractions were already in flight.
By kexecA wakeup bug in Linux pipes shows exclusive epoll watchers can swallow a wake meant for a blocked reader and never pass it on.
By kexecFailed execve and affinity changes could hand threads a bogus ID, lock up small machines, or panic the kernel.
By kexecChristian Marangi's proposal treats coding sublayers like clocks, so MACs can pick them up from device tree or ACPI even if the PCS driver arrives or leaves later.
By kexecA double-free during set resize was missed by an earlier garbage-collection fix and has been turned into a local privilege escalation.
By kexecA proposed kernel fix closes a duplicate free during hash-set resize, but a maintainer wants a real name before taking it.
By kexecHe says structured comments next to kernel interfaces would pollute the source, and that any contract belongs outside the tree for tools rather than humans.
By kexecMathieu Desnoyers posted fixes for 33 SCHED_DEADLINE accounting errors, three of which take the machine down.
By kexecA 58-patch series keeps chip metadata valid after unbind and rejects further hardware access instead of touching freed controller state.
By kexecAlexei Starovoitov proposes growing tables for maps, subprograms and kfuncs, and faster checks so safe programs are no longer turned away for size.
By kexecThe change would let Android userdata survive a move from 4KB to 16KB pages without rewriting the on-disk format.
By oopsUprobes, IMA, IPE and a BPF helper have been treating a failed data lookup as if the overlay inode were the real file.
By kexecA 16-patch series from Xiaomi and Tencent makes Multi-Gen LRU split anon and file scans by refault rates, closing a gap that produced weak swappiness response and premature out-of-memory kills.
By kexecJarkko Sakkinen's series replaces the kernel's only user of assoc_array with a hash table, and fixes lifetime bugs in s390's certificate keyring.
By kexecLoop, NVMe, the SCSI target and similar paths would refuse pseudo filesystem files after a capability bypass let unprivileged tasks trigger kernel actions.
By kexecAndrew Morton queued a change that makes the main Linux list a fallback after duplicated traffic, including AI-generated patches, made the volume unsustainable.
By oopsIan Rogers' series lets users override event encodings and metrics from directories, without rebuilding perf or rescanning every sysfs event.
By oopsIan Rogers replaces the untyped table with a type-safe API after years of NULL-check mistakes and strict-aliasing violations.
By oopsUser ringbuf drain callbacks could read and write reused kernel stack since Linux 7.3; array map iteration leaked four bytes.
By oopsFlipping the official read-only flag without the mount lock could hang every writer after a freeze, and break emergency remount.
By kexecA drained open-coded iterator violates the verifier's NULL contract; maintainers say the same pattern shows up in other iterators.
By oopsChristian Marangi's phylink changes would let coding-sublayer drivers probe and disappear independently of the MAC.
By oopsA crafted GSO packet of stacked Network Service Headers could overflow the kernel stack when the nsh module was loaded.
By oopsSasha Levin wants allmodconfig to catch userspace breakage that currently slips past bots, but Andrew Morton says module-loading tests still have a separate problem.
By kexecA build-time scan would tag more than 12,000 error-returning functions so fmod_ret programs can skip them, with the authors warning the rule is not fully safe.
By kexecCarlos López proposes skipping secondary-MMU zaps when host permissions only widen, cutting snapshot-time refaults on x86 and s390.
By kexecDavid Hildenbrand's urgent series is already in the mm tree; networking paths that can still multiply page references remain open.
By kexec