{"api_version":"1","generated_at":"2026-09-27T04:23:48+00:00","cve":"CVE-2026-97945","urls":{"html":"https://cve.report/CVE-2026-97945","api":"https://cve.report/api/cve/CVE-2026-97945.json","docs":"https://cve.report/api","cve_org":"https://www.cve.org/CVERecord?id=CVE-2026-97945","nvd":"https://nvd.nist.gov/vuln/detail/CVE-2026-97945"},"summary":{"title":"x86/mm: Fix user-space data loss with MADV_FREE and THP","description":"In the Linux kernel, the following vulnerability has been resolved:\n\nx86/mm: Fix user-space data loss with MADV_FREE and THP\n\nSome of users of Polars (a data analytics library) have lost production\ndata from this bug. They seem to have just the right combination of\nhuge pages, MADV_FREE and heavy reclaim pressure.\n\npmd_modify() masks the old value with (_HPAGE_CHG_MASK & ~_PAGE_DIRTY),\nsilently discarding the hardware dirty bit.  The subsequent\npmd_mksaveddirty() call is supposed to transfer _PAGE_DIRTY into\n_PAGE_SAVED_DIRTY when write-protecting, but the dirty bit was already\nstripped from the value, so there is nothing left to transfer.\n\nContrast with pte_modify(), which keeps _PAGE_DIRTY_BITS in its mask,\nand pud_modify(), which keeps _HPAGE_CHG_MASK untouched: pmd_modify()\nis the odd one out.  Any pmd_modify() on a writable, dirty PMD loses\nthe dirty state.\n\nOne visible consequence is data loss with MADV_FREE on PMD-mapped THP:\n\n    memset(buf, 0x5A, size);          // PMD-mapped THP, PMD dirty\n    madvise(buf, size, MADV_FREE);    // PMD cleaned but left writable,\n                                      // folio marked lazyfree\n    memset(buf, 0x5A, size);          // hardware sets _PAGE_DIRTY again\n    mprotect(buf, size, PROT_READ);   // pmd_modify() drops the dirty bit\n    mprotect(buf, size, PROT_READ|PROT_WRITE);\n    // ... memory pressure ...\n\nReclaim (e.g. under memcg pressure) then finds the lazyfree folio with\nno dirty bit set anywhere and frees it in\n__discard_anon_folio_pmd_locked(), even though the data was rewritten\nafter MADV_FREE; subsequent reads fault in fresh zero pages.  NUMA\nhinting alone can trigger the same loss, as do_huge_pmd_numa_page()\nrestores the PMD through pmd_modify() as well.\n\nPMD-mapped file THPs are affected too: mprotect()/NUMA hinting dropping\nthe dirty bit means rewritten data is never written back.\n\nFix it by keeping _PAGE_DIRTY in the preserved mask, exactly like\npte_modify() and pud_modify() do.  The existing\npmd_mksaveddirty()/pmd_clear_saveddirty() pair then performs the\nhardware-dirty <-> saved-dirty transition based on the write bit,\npreserving the shadow-stack encoding rules.","state":"PUBLISHED","assigner":"Linux","published_at":"2026-09-25 11:17:21","updated_at":"2026-09-25 11:17:21"},"problem_types":[],"metrics":[],"references":[{"url":"https://git.kernel.org/stable/c/e4d569fe2f436a9b58c048fefabf56eab68c1b7e","name":"https://git.kernel.org/stable/c/e4d569fe2f436a9b58c048fefabf56eab68c1b7e","refsource":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","tags":[],"title":"","mime":"","httpstatus":"","archivestatus":"0"},{"url":"https://git.kernel.org/stable/c/a5ad97cbcfcec31430268441c10eedde73200177","name":"https://git.kernel.org/stable/c/a5ad97cbcfcec31430268441c10eedde73200177","refsource":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","tags":[],"title":"","mime":"","httpstatus":"","archivestatus":"0"},{"url":"https://git.kernel.org/stable/c/55ddbc2ca6d5db64d44c3fb8f8705b4d39eb30b7","name":"https://git.kernel.org/stable/c/55ddbc2ca6d5db64d44c3fb8f8705b4d39eb30b7","refsource":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","tags":[],"title":"","mime":"","httpstatus":"","archivestatus":"0"},{"url":"https://git.kernel.org/stable/c/f7491d7c81db0e7c304a7bd757a76d2fbeaff80e","name":"https://git.kernel.org/stable/c/f7491d7c81db0e7c304a7bd757a76d2fbeaff80e","refsource":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","tags":[],"title":"","mime":"","httpstatus":"","archivestatus":"0"},{"url":"https://www.cve.org/CVERecord?id=CVE-2026-97945","name":"CVE Program record","refsource":"CVE.ORG","tags":["canonical"]},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2026-97945","name":"NVD vulnerability detail","refsource":"NVD","tags":["canonical","analysis"]}],"affected":[{"source":"CNA","vendor":"Linux","product":"Linux","version":"affected bb3aadf7d446aaf22c725b274e2c194ac5cb2111 e4d569fe2f436a9b58c048fefabf56eab68c1b7e git","platforms":[]},{"source":"CNA","vendor":"Linux","product":"Linux","version":"affected bb3aadf7d446aaf22c725b274e2c194ac5cb2111 55ddbc2ca6d5db64d44c3fb8f8705b4d39eb30b7 git","platforms":[]},{"source":"CNA","vendor":"Linux","product":"Linux","version":"affected bb3aadf7d446aaf22c725b274e2c194ac5cb2111 a5ad97cbcfcec31430268441c10eedde73200177 git","platforms":[]},{"source":"CNA","vendor":"Linux","product":"Linux","version":"affected bb3aadf7d446aaf22c725b274e2c194ac5cb2111 f7491d7c81db0e7c304a7bd757a76d2fbeaff80e git","platforms":[]},{"source":"CNA","vendor":"Linux","product":"Linux","version":"affected 6.6","platforms":[]},{"source":"CNA","vendor":"Linux","product":"Linux","version":"unaffected 6.6 semver","platforms":[]},{"source":"CNA","vendor":"Linux","product":"Linux","version":"unaffected 6.12.111 6.12.* semver","platforms":[]},{"source":"CNA","vendor":"Linux","product":"Linux","version":"unaffected 6.18.53 6.18.* semver","platforms":[]},{"source":"CNA","vendor":"Linux","product":"Linux","version":"unaffected 7.2.7 7.2.* semver","platforms":[]},{"source":"CNA","vendor":"Linux","product":"Linux","version":"unaffected 7.3-rc4 * original_commit_for_fix","platforms":[]}],"timeline":[],"solutions":[],"workarounds":[],"exploits":[],"credits":[],"nvd_cpes":[],"vendor_comments":[],"enrichments":{"kev":null,"epss":{"cve_year":"2026","cve_id":"97945","cve":"CVE-2026-97945","epss":"0.002000000","percentile":"0.088480000","score_date":"2026-09-26","updated_at":"2026-09-27 00:04:20"},"legacy_qids":[]},"source_records":{"cve_program":{"containers":{"cna":{"affected":[{"defaultStatus":"unaffected","product":"Linux","programFiles":["arch/x86/include/asm/pgtable.h"],"repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","vendor":"Linux","versions":[{"lessThan":"e4d569fe2f436a9b58c048fefabf56eab68c1b7e","status":"affected","version":"bb3aadf7d446aaf22c725b274e2c194ac5cb2111","versionType":"git"},{"lessThan":"55ddbc2ca6d5db64d44c3fb8f8705b4d39eb30b7","status":"affected","version":"bb3aadf7d446aaf22c725b274e2c194ac5cb2111","versionType":"git"},{"lessThan":"a5ad97cbcfcec31430268441c10eedde73200177","status":"affected","version":"bb3aadf7d446aaf22c725b274e2c194ac5cb2111","versionType":"git"},{"lessThan":"f7491d7c81db0e7c304a7bd757a76d2fbeaff80e","status":"affected","version":"bb3aadf7d446aaf22c725b274e2c194ac5cb2111","versionType":"git"}]},{"defaultStatus":"affected","product":"Linux","programFiles":["arch/x86/include/asm/pgtable.h"],"repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","vendor":"Linux","versions":[{"status":"affected","version":"6.6"},{"lessThan":"6.6","status":"unaffected","version":"0","versionType":"semver"},{"lessThanOrEqual":"6.12.*","status":"unaffected","version":"6.12.111","versionType":"semver"},{"lessThanOrEqual":"6.18.*","status":"unaffected","version":"6.18.53","versionType":"semver"},{"lessThanOrEqual":"7.2.*","status":"unaffected","version":"7.2.7","versionType":"semver"},{"lessThanOrEqual":"*","status":"unaffected","version":"7.3-rc4","versionType":"original_commit_for_fix"}]}],"cpeApplicability":[{"nodes":[{"cpeMatch":[{"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionEndExcluding":"6.12.111","versionStartIncluding":"6.6","vulnerable":true},{"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionEndExcluding":"6.18.53","versionStartIncluding":"6.6","vulnerable":true},{"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionEndExcluding":"7.2.7","versionStartIncluding":"6.6","vulnerable":true},{"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionEndExcluding":"7.3-rc4","versionStartIncluding":"6.6","vulnerable":true}],"negate":false,"operator":"OR"}]}],"descriptions":[{"lang":"en","value":"In the Linux kernel, the following vulnerability has been resolved:\n\nx86/mm: Fix user-space data loss with MADV_FREE and THP\n\nSome of users of Polars (a data analytics library) have lost production\ndata from this bug. They seem to have just the right combination of\nhuge pages, MADV_FREE and heavy reclaim pressure.\n\npmd_modify() masks the old value with (_HPAGE_CHG_MASK & ~_PAGE_DIRTY),\nsilently discarding the hardware dirty bit.  The subsequent\npmd_mksaveddirty() call is supposed to transfer _PAGE_DIRTY into\n_PAGE_SAVED_DIRTY when write-protecting, but the dirty bit was already\nstripped from the value, so there is nothing left to transfer.\n\nContrast with pte_modify(), which keeps _PAGE_DIRTY_BITS in its mask,\nand pud_modify(), which keeps _HPAGE_CHG_MASK untouched: pmd_modify()\nis the odd one out.  Any pmd_modify() on a writable, dirty PMD loses\nthe dirty state.\n\nOne visible consequence is data loss with MADV_FREE on PMD-mapped THP:\n\n    memset(buf, 0x5A, size);          // PMD-mapped THP, PMD dirty\n    madvise(buf, size, MADV_FREE);    // PMD cleaned but left writable,\n                                      // folio marked lazyfree\n    memset(buf, 0x5A, size);          // hardware sets _PAGE_DIRTY again\n    mprotect(buf, size, PROT_READ);   // pmd_modify() drops the dirty bit\n    mprotect(buf, size, PROT_READ|PROT_WRITE);\n    // ... memory pressure ...\n\nReclaim (e.g. under memcg pressure) then finds the lazyfree folio with\nno dirty bit set anywhere and frees it in\n__discard_anon_folio_pmd_locked(), even though the data was rewritten\nafter MADV_FREE; subsequent reads fault in fresh zero pages.  NUMA\nhinting alone can trigger the same loss, as do_huge_pmd_numa_page()\nrestores the PMD through pmd_modify() as well.\n\nPMD-mapped file THPs are affected too: mprotect()/NUMA hinting dropping\nthe dirty bit means rewritten data is never written back.\n\nFix it by keeping _PAGE_DIRTY in the preserved mask, exactly like\npte_modify() and pud_modify() do.  The existing\npmd_mksaveddirty()/pmd_clear_saveddirty() pair then performs the\nhardware-dirty <-> saved-dirty transition based on the write bit,\npreserving the shadow-stack encoding rules."}],"providerMetadata":{"dateUpdated":"2026-09-25T10:22:55.981Z","orgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","shortName":"Linux"},"references":[{"url":"https://git.kernel.org/stable/c/e4d569fe2f436a9b58c048fefabf56eab68c1b7e"},{"url":"https://git.kernel.org/stable/c/55ddbc2ca6d5db64d44c3fb8f8705b4d39eb30b7"},{"url":"https://git.kernel.org/stable/c/a5ad97cbcfcec31430268441c10eedde73200177"},{"url":"https://git.kernel.org/stable/c/f7491d7c81db0e7c304a7bd757a76d2fbeaff80e"}],"title":"x86/mm: Fix user-space data loss with MADV_FREE and THP","x_generator":{"engine":"bippy-1.2.0"}}},"cveMetadata":{"assignerOrgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","assignerShortName":"Linux","cveId":"CVE-2026-97945","datePublished":"2026-09-25T10:22:55.981Z","dateReserved":"2026-09-25T10:18:58.204Z","dateUpdated":"2026-09-25T10:22:55.981Z","state":"PUBLISHED"},"dataType":"CVE_RECORD","dataVersion":"5.2"},"nvd":{"publishedDate":"2026-09-25 11:17:21","lastModifiedDate":"2026-09-25 11:17:21","problem_types":[],"metrics":[],"configurations":[]},"legacy_mitre":{"record":{"CveYear":"2026","CveId":"97945","Ordinal":"1","Title":"x86/mm: Fix user-space data loss with MADV_FREE and THP","CVE":"CVE-2026-97945","Year":"2026"},"notes":[{"CveYear":"2026","CveId":"97945","Ordinal":"1","NoteData":"In the Linux kernel, the following vulnerability has been resolved:\n\nx86/mm: Fix user-space data loss with MADV_FREE and THP\n\nSome of users of Polars (a data analytics library) have lost production\ndata from this bug. They seem to have just the right combination of\nhuge pages, MADV_FREE and heavy reclaim pressure.\n\npmd_modify() masks the old value with (_HPAGE_CHG_MASK & ~_PAGE_DIRTY),\nsilently discarding the hardware dirty bit.  The subsequent\npmd_mksaveddirty() call is supposed to transfer _PAGE_DIRTY into\n_PAGE_SAVED_DIRTY when write-protecting, but the dirty bit was already\nstripped from the value, so there is nothing left to transfer.\n\nContrast with pte_modify(), which keeps _PAGE_DIRTY_BITS in its mask,\nand pud_modify(), which keeps _HPAGE_CHG_MASK untouched: pmd_modify()\nis the odd one out.  Any pmd_modify() on a writable, dirty PMD loses\nthe dirty state.\n\nOne visible consequence is data loss with MADV_FREE on PMD-mapped THP:\n\n    memset(buf, 0x5A, size);          // PMD-mapped THP, PMD dirty\n    madvise(buf, size, MADV_FREE);    // PMD cleaned but left writable,\n                                      // folio marked lazyfree\n    memset(buf, 0x5A, size);          // hardware sets _PAGE_DIRTY again\n    mprotect(buf, size, PROT_READ);   // pmd_modify() drops the dirty bit\n    mprotect(buf, size, PROT_READ|PROT_WRITE);\n    // ... memory pressure ...\n\nReclaim (e.g. under memcg pressure) then finds the lazyfree folio with\nno dirty bit set anywhere and frees it in\n__discard_anon_folio_pmd_locked(), even though the data was rewritten\nafter MADV_FREE; subsequent reads fault in fresh zero pages.  NUMA\nhinting alone can trigger the same loss, as do_huge_pmd_numa_page()\nrestores the PMD through pmd_modify() as well.\n\nPMD-mapped file THPs are affected too: mprotect()/NUMA hinting dropping\nthe dirty bit means rewritten data is never written back.\n\nFix it by keeping _PAGE_DIRTY in the preserved mask, exactly like\npte_modify() and pud_modify() do.  The existing\npmd_mksaveddirty()/pmd_clear_saveddirty() pair then performs the\nhardware-dirty <-> saved-dirty transition based on the write bit,\npreserving the shadow-stack encoding rules.","Type":"Description","Title":"x86/mm: Fix user-space data loss with MADV_FREE and THP"}]}}}