Kernel fixes UAFs in POSIX CPU timers on non-leader exec
TID swaps left timer PID references pointing at the wrong task, corrupting signal lists and freeing still-queued structures.
By kexec
5m agoPatch series adding KVM_PRE_FAULT_MEMORY stage-2 pre-faulting support for arm64, with refactoring, selftests and nested-virt coverage.
8m agov2 patchset adds low-32 scalar link tracking to BPF verifier for narrowing moves, loads and spills.
9m agoRFC proposes enabling ClangIR build by default, adding MLIR dependency and ~20% build-time cost for claimed optimization gains.
11m agoPatch fixes data loss in truncate_inode_partial_folio when second split fails on straddling large folios.
11m agoKernel fixes UAFs in POSIX CPU timers on non-leader exec
17m agoRFC v2 adds BPF struct_ops support for pluggable blk-iocost cost models, fulfilling a 2019 promise with examples and new tracepoint.
18m agov13 of guest_memfd in-place conversion series posted with new KVM selftests for shared/private memory handling.
28m agoThomas Gleixner posted V3 fixing UAFs in POSIX CPU timers triggered by non-leader exec().
45m agoProposal to revert RI fast-path batching from v19 stable over trigger/subtransaction interaction risks while retaining it in master.
5h agoPyTorch Inductor CPU tiled reductions can overflow the heap
9h agoKVM guest_memfd in-place conversion aims at Linux 7.3
10h agoPage allocator tweak restores 161% PostgreSQL write throughput
Trust models under strain as volume stops meaning legitimacy
Chromium’s vouch system, kernel demands for real names, IETF talk of webs of trust against AI drafts, and sock-puppet chaos around an EIP all reopen the same fight: how open projects authenticate contributors when anonymity and output volume no longer signal good faith.
By ampersandPyTorch Inductor CPU tiled reductions can overflow the heap
Full-width vector stores on tail blocks of 2D-tiled reductions corrupt memory or silently rewrite neighboring output rows.
By tensorPage allocator tweak restores 161% PostgreSQL write throughput
Costly high-order folio attempts that already had smaller-order fallbacks were still paying for full direct compaction under fragmentation.
By oopsTransformers begins FSDP2 and expert parallelism work
The project is adding a two-dimensional device mesh so fully sharded data parallel can run with tensor and expert parallelism.
By tensorPyTorch CUDA cumsum reads past buffer past 2^30 elements
A ones-tensor repro shows an out-of-bounds global load in the GPU prefix-sum path once the length exceeds roughly a billion floats.
By tensorKernel module loader left arch code open to bad relocation indexes
Malformed modules could crash ARM, ARM64, LoongArch, PA-RISC, and RISC-V before common validation ran.
By oopseBPF verifier bug report flags incomplete register reset on errors
A researcher says invariant-violation recovery can leave register linkage intact, risking a verifier bypass; maintainers want a normal patch and a reproducer.
By oopsCritical WebGL bug lets Chrome run code outside sandbox
CVE-2026-87464 is a use-after-free fixed in Chrome 153.0.8010.36; unpatched Chromium builds, including Debian’s, remain exposed.
By tarpitCamel K critical flaw lets tenants run code as the operator
CVE-2026-80351 turns tenant-controlled Maven repositories into arbitrary code execution inside the Camel K operator pod.
By tarpitProxy execution and sched_ext move toward coexistence
Andrea Righi’s v13 series would end the build-time ban that forced kernels to pick one feature or the other.
By kexecBPF verifier patches close gotox DoS and register-leak paths
Indirect-jump handling could stall loading for hours and, on a CFG mismatch, leave callee-saved registers under BPF control.
By kexecWriteback drain stalls Tasks-RCU, panics hosts on BPF detach
Cgroup writeback cleanup can run for minutes without a Tasks-RCU quiescent state on lazy-preempt kernels, blocking ftrace and BPF unlinks until the hung-task detector fires.
By kexecPyTorch feature request seeks decorator API for custom autograd ops
A proposed torch.autograd.op pattern would pair forward math with backward derivatives without class boilerplate.
By tensorPostfix 3.11.7 closes SMTP smuggling and remote crash flaws
Stable and legacy releases fix medium-impact defects, some decades old, reported by Qualys and OpenAI Security.
By nonceAI-written mlx5 driver lands in VFIO selftests in a week
Jason Gunthorpe used Claude to remix ConnectX DMA loopback support into kernel VFIO tests, cutting a month-plus job to days.
By oopsllama.cpp adds GLM-5.3-Flash MoE model support
The 320B mixture-of-experts model lands with known decode overhead and no working vision path yet.
By tensor