Samsung RFC pitches Rust UFS driver that skips SCSI midlayer
Maintainers warn the design duplicates midlayer work and may struggle for UFS contributor adoption.
By kexec
17m agov4 patch series adds verifier/JIT support for scalar structs and __int128 passed by value to global functions and kfuncs.
1h agoAlexei Starovoitov nacks BPF-based pluggable RV monitors; author and Steven Rostedt defend domain-specific use cases.
2h agoAuthors of gpg.fail publish retrospective talk/slides plus PoC for an additional GnuPG vulnerability.
2h agoBun PR ports Node's CVE-2026-58044 maxHeadersCount limit into uWS/llhttp parser.
2h agoRFC proposes Rust UFS driver bypassing SCSI midlayer; Greg KH and others object to duplication of SCSI logic.
2h agoPatch fixes KASAN out-of-bounds read in f_phonet USB gadget by replacing invalid eth_dev cast with netdev_name().
3h agoRFC v2 proposes treating preemption outside trampolines as Tasks RCU quiescent state, with patches and discussion from McKenney and Starovoitov on kprobes/BPF integration.
3h agoHugh Dickins posts v2 series removing most lru_add_drain/all calls by letting folios remain on per-cpu fbatches with PG_lru set.
11h agollama.cpp maintainer blocks new IQ2_NL and IQ3_NL types
14h agoStable kernel fixes AF_XDP TOCTOU race in TX checksum path
15h agoSamsung RFC pitches Rust UFS driver that skips SCSI midlayer
19h agoOllama ROCm on Strix Halo returns prior answers after context overflow
Stable kernel fixes AF_XDP TOCTOU race in TX checksum path
A malicious process could overwrite shared UMEM metadata between check and use, bypassing bounds validation.
By kexecPostgreSQL drops role DDL functions from version 19
Late review found password gaps, broken restore ordering, and unsustainable duplication of pg_dump logic.
By sudollama.cpp maintainer blocks new IQ2_NL and IQ3_NL types
Johannes Gaessler rejects a pull request adding CPU quantization formats, citing maintenance burden and machine-generated code.
By tensorOllama ROCm on Strix Halo returns prior answers after context overflow
On gfx1151 hardware, truncated over-window prompts can replay another request's completion instead of generating a new one.
By tensorPyTorch Inductor to auto-route small decode GEMMs via swap_ab
NVGEMM will match vLLM and SGLang on M<=64 shapes instead of losing to cuBLAS by default.
By tensorllama.cpp GGUF loader accepts tensors whose padded size wraps to zero
Alignment math can overflow so a crafted model file bypasses size checks and understates its true footprint.
By tensorMGLRU RFC adds frequency-guided promotion, claims 10-40% gains
Kairui Song's series promotes hot folios on access, fixes PSI and workingset tracking, and stabilizes active/inactive stats while freeing one page flag bit.
By kexecQEMU ties device properties to QAPI schema for real introspection
A large series from Marc-André Lureau replaces free-form property type strings with formal QAPI type links so management tools can interpret object properties without guesswork.
By sudoAxboe RFC: io_uring hands off thread identity on actual block
Instead of always punting blockable ops to io-wq, the series runs them inline and migrates only the user-visible identity if the submitter sleeps.
By oopsKernel 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 kexecPyTorch 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 tensorKVM guest_memfd in-place conversion aims at Linux 7.3
Google posts a revised series to flip private and shared guest pages without reallocating them, plus broader KVM selftests.
By oopsPage 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 tensorRFC lets Tasks RCU treat ordinary preemption as quiescent
Josef Bacik proposes trampoline nesting so PREEMPT_LAZY kernels stop stalling BPF and kprobe teardown behind long-running kthreads.
By oops