{ "slug": "zero-copy-source-lifetime", "title": "Linux zero-copy source lifetime: eight file-to-pipe checks", "summary": "40 file-to-pipe observations, plus 10 separate tee(2) checks, distinguish source stability from enqueue and consumer completion.", "published": "2026-10-18", "updated": "2026-10-18", "method": "Sequential one-page file-to-pipe tests; source mutation finishes before pipe consumption except the consume-first control.", "limits": [ "One host and filesystem-magic label; not a filesystem comparison.", "Synthetic A/B/C payloads; measured API observations, not storage-fleet statistics.", "No TCP, MSG_ZEROCOPY, crash, power-loss, throughput or cache-eviction test.", "Mutation cases intentionally violate the documented sendfile source-lifetime requirement.", "A different supported path may copy data; post-mutation B/C results are observations, not a universal API guarantee." ], "license": "Hesela's original harness, generated observations and derived CSV are dedicated under CC0-1.0. Linked documentation and Linux source retain their own licenses and are not redistributed.", "licenseUrl": "https://creativecommons.org/publicdomain/zero/1.0/", "downloadPath": "/data/zero-copy-source-lifetime.csv", "rawPath": "/research/zero-copy-lifetime.json", "codePath": "/research/zero-copy-lifetime.py", "generatorPath": "/research/build-zero-copy-dataset.py", "testsPath": "/research/test_zero_copy_lifetime.py", "validationTestsPath": "/research/test_zero_copy_dataset.py", "humanPage": "https://hesela.com/analyses/zero-copy-source-lifetime/", "rowCount": 40, "columns": [ { "name": "repetition", "description": "Sequential repetition number; not an independent fleet sample." }, { "name": "case", "description": "Named sequence defined by the Python harness." }, { "name": "page_bytes", "description": "Requested and received payload length in bytes." }, { "name": "enqueue_bytes", "description": "Successful file-to-pipe or buffered pipe-write return value in bytes." }, { "name": "received_byte", "description": "Uniform ASCII byte observed at pipe consumption." }, { "name": "path_byte", "description": "Uniform ASCII byte read by opening the source pathname after the operation." }, { "name": "old_inode_byte", "description": "Uniform ASCII byte read through the original writable descriptor." }, { "name": "path_same_inode", "description": "Whether device and inode identity still match the initial source." }, { "name": "matches_before", "description": "Whether received SHA-256 equals the hash computed before enqueue." } ], "rows": [ { "repetition": 1, "case": "buffered-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "A", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": true }, { "repetition": 1, "case": "sendfile-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "B", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": false }, { "repetition": 1, "case": "splice-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "B", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": false }, { "repetition": 1, "case": "sendfile-overwrite-fsync", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "B", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": false }, { "repetition": 1, "case": "sendfile-replace-path", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "A", "path_byte": "B", "old_inode_byte": "A", "path_same_inode": false, "matches_before": true }, { "repetition": 1, "case": "sendfile-replace-then-write-old", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "C", "path_byte": "B", "old_inode_byte": "C", "path_same_inode": false, "matches_before": false }, { "repetition": 1, "case": "sendfile-close-then-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "B", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": false }, { "repetition": 1, "case": "sendfile-consume-then-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "A", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": true }, { "repetition": 2, "case": "buffered-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "A", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": true }, { "repetition": 2, "case": "sendfile-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "B", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": false }, { "repetition": 2, "case": "splice-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "B", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": false }, { "repetition": 2, "case": "sendfile-overwrite-fsync", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "B", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": false }, { "repetition": 2, "case": "sendfile-replace-path", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "A", "path_byte": "B", "old_inode_byte": "A", "path_same_inode": false, "matches_before": true }, { "repetition": 2, "case": "sendfile-replace-then-write-old", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "C", "path_byte": "B", "old_inode_byte": "C", "path_same_inode": false, "matches_before": false }, { "repetition": 2, "case": "sendfile-close-then-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "B", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": false }, { "repetition": 2, "case": "sendfile-consume-then-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "A", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": true }, { "repetition": 3, "case": "buffered-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "A", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": true }, { "repetition": 3, "case": "sendfile-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "B", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": false }, { "repetition": 3, "case": "splice-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "B", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": false }, { "repetition": 3, "case": "sendfile-overwrite-fsync", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "B", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": false }, { "repetition": 3, "case": "sendfile-replace-path", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "A", "path_byte": "B", "old_inode_byte": "A", "path_same_inode": false, "matches_before": true }, { "repetition": 3, "case": "sendfile-replace-then-write-old", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "C", "path_byte": "B", "old_inode_byte": "C", "path_same_inode": false, "matches_before": false }, { "repetition": 3, "case": "sendfile-close-then-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "B", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": false }, { "repetition": 3, "case": "sendfile-consume-then-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "A", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": true }, { "repetition": 4, "case": "buffered-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "A", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": true }, { "repetition": 4, "case": "sendfile-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "B", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": false }, { "repetition": 4, "case": "splice-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "B", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": false }, { "repetition": 4, "case": "sendfile-overwrite-fsync", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "B", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": false }, { "repetition": 4, "case": "sendfile-replace-path", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "A", "path_byte": "B", "old_inode_byte": "A", "path_same_inode": false, "matches_before": true }, { "repetition": 4, "case": "sendfile-replace-then-write-old", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "C", "path_byte": "B", "old_inode_byte": "C", "path_same_inode": false, "matches_before": false }, { "repetition": 4, "case": "sendfile-close-then-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "B", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": false }, { "repetition": 4, "case": "sendfile-consume-then-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "A", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": true }, { "repetition": 5, "case": "buffered-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "A", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": true }, { "repetition": 5, "case": "sendfile-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "B", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": false }, { "repetition": 5, "case": "splice-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "B", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": false }, { "repetition": 5, "case": "sendfile-overwrite-fsync", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "B", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": false }, { "repetition": 5, "case": "sendfile-replace-path", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "A", "path_byte": "B", "old_inode_byte": "A", "path_same_inode": false, "matches_before": true }, { "repetition": 5, "case": "sendfile-replace-then-write-old", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "C", "path_byte": "B", "old_inode_byte": "C", "path_same_inode": false, "matches_before": false }, { "repetition": 5, "case": "sendfile-close-then-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "B", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": false }, { "repetition": 5, "case": "sendfile-consume-then-overwrite", "page_bytes": 4096, "enqueue_bytes": 4096, "received_byte": "A", "path_byte": "B", "old_inode_byte": "B", "path_same_inode": true, "matches_before": true } ], "environment": { "kernel": "6.8.0-146-generic", "machine": "x86_64", "python": "3.12.3", "filesystem_magic_label": "ext2/ext3" }, "generatedAt": "2026-10-17T22:22:27.547266+00:00", "repetitions": 5, "citations": [ { "id": "sendfile", "title": "Linux man-pages 6.19: sendfile(2), return value and source lifetime", "url": "https://man7.org/linux/man-pages/man2/sendfile.2.html", "accessed": "2026-10-17" }, { "id": "splice", "title": "Linux man-pages 6.19: splice(2), pipe buffers and flags", "url": "https://man7.org/linux/man-pages/man2/splice.2.html", "accessed": "2026-10-17" }, { "id": "filemap", "title": "Linux v6.8 mm/filemap.c: splice_folio_into_pipe and filemap_splice_read", "url": "https://github.com/torvalds/linux/blob/v6.8/mm/filemap.c#L2614-L2758", "accessed": "2026-10-17" }, { "id": "rename", "title": "Linux man-pages 6.19: rename(2), replacement and open descriptors", "url": "https://man7.org/linux/man-pages/man2/rename.2.html", "accessed": "2026-10-17" }, { "id": "fsync", "title": "Linux man-pages 6.19: fsync(2), synchronization and directory entries", "url": "https://man7.org/linux/man-pages/man2/fsync.2.html", "accessed": "2026-10-17" }, { "id": "copy-path", "title": "Linux 6.12 documentation: splice and pipes, copy_splice_read", "url": "https://docs.kernel.org/6.12/filesystems/splice.html", "accessed": "2026-10-17" }, { "id": "msg-zerocopy", "title": "Linux kernel documentation: MSG_ZEROCOPY, notifications and deferred copies", "url": "https://docs.kernel.org/networking/msg_zerocopy.html", "accessed": "2026-10-17" }, { "id": "tee", "title": "Linux man-pages 6.19: tee(2), duplication without consuming and shared references", "url": "https://www.man7.org/linux/man-pages/man2/tee.2.html", "accessed": "2026-10-17" }, { "id": "io-lite", "title": "Pai, Druschel and Zwaenepoel: IO-Lite, OSDI 1999, sections 2.1, 2.5, 2.9 and 3", "url": "https://www.usenix.org/legacy/publications/library/proceedings/osdi99/full_papers/pai/pai_html/pai.html", "accessed": "2026-10-18" } ], "supplement": { "title": "tee(2) split-consumer observations", "rawPath": "/research/tee-source-lifetime.json", "codePath": "/research/tee-source-lifetime.py", "testsPath": "/research/test_tee_lifetime.py", "rowCount": 10, "sha256": { "tee-source-lifetime.json": "9b8254fee364d88b9a6910ee292dcb85532d5f5c19458899435fea6b12d02858", "tee-source-lifetime.py": "54c8a96b7d07ccccd510eedf513a545f3f48c8a0bbfc58107ae00fb08193261a", "test_tee_lifetime.py": "03675f358ec18789f701610f1f750551312cae39182fb83dd184753d016bb539" } }, "sha256": { "raw": "e2a6d276726e89b64a66e20aef642e06782785aeef5c5832f194f1b05b585454", "csv": "e3e547f949a54fe8aa871df17291d4dbd1ae57e77dceb791b1bb5946dde68af0", "harness": "b67fb20a42c36fa650d83815895df8a980ad023beec0cb21029c59dd8915da39", "tests": "e2f5576dcf07fb31f4d34f6b3bd60ccecd59e41a25caed48fe575058af9dd308", "generator": "bf0fb7e7c445eddc0e1e60e5e4617f37b36e7b0b1867b6119916be0619db53db", "validationTests": "18fb12a7cdf6250923e8a60f5a2c421c96c97d30d069561a5c7ed5c00b050ffa" } }