freenode

← digests

Kernel: ipset double-free, mm_cid fixes, MTD list dispute

Kernel & Low-Level2026-11-18

Kernel lists on 18 November were dominated by a netfilter ipset double-free that can yield local privilege escalation, plus use-after-free fixes in the scheduler's mm_cid path. A public rebuke over LLM-generated MTD patches and several filesystem and memory-management series filled out the day.

ipset resize can double-free during garbage collection

Patches on the net and stable lists exclude add-path garbage collection while an ipset is resized. The bug is a double-free that can be used for local privilege escalation from unprivileged namespaces.

mm_cid series fixes use-after-free on exec and affinity

Mathieu Desnoyers posted a 17-patch series fixing six mm_cid scheduler bugs, including a use-after-free on exec and affinity changes, marked for stable. The series also adds new rseq selftests.

FUSE names with embedded NULs can leak kernel memory

A three-patch series covers FUSE and fsnotify handling of names that contain embedded NULs. FUSE READDIRPLUS can supply qstrs with embedded NULs, so strscpy copies in inotify and fanotify leak uninitialized kernel memory to userspace.

MTD maintainer asks submitter to stop flooding the list

A 58-patch v5 series proposes retaining raw NAND chip and controller state across removal. The maintainer publicly called out the submitter for flooding the list with hundreds of LLM-generated MTD patches and asked them to stop.

RFC refuses pseudo filesystem files as storage backends

Christian Brauner posted RFC v3 of a series that adds a common helper to refuse pseudo filesystem files (sysfs, procfs, bpffs) as storage backends. The helper covers loop, nvmet, the SCSI target, and overlayfs data layers.

Linus calls epoll broken in pipe wakeup thread

A patch would keep per-write wakeups for EPOLLEXCLUSIVE consumers, in a thread about pipe rd_wait ordering with epoll. Linus called epoll broken and suggested ignoring the report.

Series would leave folios on per-CPU fbatches

Hugh Dickins posted v2 of a 26-patch series that removes most lru_add_drain and lru_add_drain_all calls. Folios would remain on per-CPU fbatches with PG_lru set.