fix(kernel): rebase stable base to Linux 6.18.44 #15
Loading…
Reference in a new issue
No description provided.
Delete branch "security/linux-6.18.44"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Picks up 14 CVEs across 6.18.43 and 6.18.44. Applicable to this fleet (headless AMD EPYC cloud VMs running untrusted CI workloads):
Not applicable - hardware this fleet does not have: CVE-2026-64586 (brcmfmac wifi) and 7 drm CVEs (xe, i915, amdgpu, gpusvm).
Also picks up two non-CVE fixes that matter here: mm/hugetlb swap-entry corruption on
fork()(this fleet leans on hugepages for mayastor) and an mm/huge_memory UAF in__folio_split().Verified before push
SUBLEVEL = 44v6.18.44fs/{Makefile,Kconfig}still wire ZFS inGenerated by
scripts/kernel/rebase-stable.sh(seanfarm PR #991), which reproduces this exact tree88ddd7e4550c269ae410896f84c4ee63cfae8aaa.Import the authenticated kernel.org linux-6.18.44 stable tarball and reapply the complete effective seanfarm overlay from 6.18.42. Upstream tarball SHA-256: 0f72d938f06828e82c90405174fe572287db7bfe089e2fc46572a99a7f240d43 Upstream stable commit: 1efe5d048a391de3ead2804b2e7f86376c356cc5 Upstream tree: 9853ab31251669994d0d57301d4090b04e2e9e2e ZFS remains source-derived 2.4.2; no upstream path overlaps the effective seanfarm overlay. Picks up 14 CVEs across 6.18.43 and 6.18.44. The ones that apply to this fleet (headless AMD EPYC cloud VMs running untrusted CI workloads): CVE-2026-68480 x86/bugs: Safe-RET robust against interrupt injection (SRSO -- our nodes are AMD EPYC; speculative data leak) CVE-2026-68166 userfaultfd: prevent registration of special VMAs (UFFDIO_COPY with attacker-chosen return addresses) CVE-2026-63978 net/handshake: drain pending requests at netns exit CVE-2026-63979 net/handshake: hand off pinned file ref to accept CVE-2026-64523 net/handshake: take a long-lived file ref at submit CVE-2026-68322 rds: NULL deref when IPv6 is disabled Not applicable, hardware this fleet does not have: CVE-2026-64586 (brcmfmac wifi) and 7 drm CVEs (xe, i915, amdgpu, gpusvm). Also picks up two non-CVE fixes that matter here: an mm/hugetlb swap-entry corruption on fork() (this fleet leans on hugepages for mayastor) and an mm/huge_memory use-after-free in __folio_split(). Overlay is 854 files: 472 under fs/zfs, 372 under include/zfs, both seanfarm_defconfigs, CHANGELOG.md, 2 genl_magic_func headers, and 5 files that modify upstream (fs/{Kconfig,Makefile}, drivers/block/{Kconfig,Makefile}, fs/iomap/iter.c). All 5 are byte-identical between v6.18.42 and v6.18.44 upstream, and none of the added paths exist upstream, so the overlay transplants with zero conflicts.