{"api_version":"1","generated_at":"2026-11-20T19:06:09+00:00","cve":"CVE-2026-99648","urls":{"html":"https://cve.report/CVE-2026-99648","api":"https://cve.report/api/cve/CVE-2026-99648.json","docs":"https://cve.report/api","cve_org":"https://www.cve.org/CVERecord?id=CVE-2026-99648","nvd":"https://nvd.nist.gov/vuln/detail/CVE-2026-99648"},"summary":{"title":"mm/vma: call ->may_split() before a partial merge","description":"In the Linux kernel, the following vulnerability has been resolved:\n\nmm/vma: call ->may_split() before a partial merge\n\nA driver can use the ->may_split() hook to forbid a VMA boundary at\ncertain addresses.\n\nA device-dax device is a character device that gives raw access to a\nrange of memory through mmap(). device-dax refuses any address that is not\na multiple of the device's alignment through vm_ops->may_split.\n\nvma_merge_existing_range() can move boundaries, but does not call\nvm_ops->may_split.\n\nIf a range at the edge of a VMA gets the attributes of the adjacent VMA,\nthe function moves only that range over to the neighbour: one VMA shrinks\nand the other grows. This is a partial merge, and vma_modify() tries it\nbefore any split.\n\nTake a device-dax VMA that lies next to another VMA of the same file.\nThe user calls mprotect() on the one page of the first VMA that touches\nthe second VMA, and asks for the protection that the second VMA already\nhas. The result is a partial merge. It puts the boundary inside a 1 GiB\nmapping, which is a single leaf entry at the PUD level.\n\nAs part of the merge, the kernel looks for a huge PMD to split at the\nnew boundary. That code does not expect a PUD leaf:\n\n    vma_merge_existing_range()\n      commit_merge()\n        vma_adjust_trans_huge()\n          split_huge_pmd_address()\n            mm_find_pmd()\n              treats the PUD leaf as a pointer to a PMD table,\n              so the \"table\" is the device memory that the leaf maps\n            __split_huge_pmd()\n              reads a PMD that the user wrote to the device memory\n              folio_put() on the folio that this PMD names\n\nBad. While the code must at some point be taught to split PUD entries as\nwell, the issue here is that we move VMA boundaries such that they are\nincompatible with the driver.\n\nFix this by calling ->may_split() on the address that a partial merge\nwould move the boundary to, and by abandoning the merge if the driver\nrefuses. vma_modify() then tries the split, which fails with -EINVAL.\n\nFactor out existing split handling into check_split() and reuse it.\n\nFix generated by an LLM and sent by Chris Mason in private.\n\n[ david: improve description; cleanup the code ]","state":"PUBLISHED","assigner":"Linux","published_at":"2026-11-20 06:17:58","updated_at":"2026-11-20 06:17:58"},"problem_types":[],"metrics":[],"references":[{"url":"https://git.kernel.org/stable/c/f437fcb76d448a396940498aba278ac0ad8d715a","name":"https://git.kernel.org/stable/c/f437fcb76d448a396940498aba278ac0ad8d715a","refsource":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","tags":[],"title":"","mime":"","httpstatus":"","archivestatus":"0"},{"url":"https://git.kernel.org/stable/c/456932cdc6eda70b5ad4c628556098846699ab03","name":"https://git.kernel.org/stable/c/456932cdc6eda70b5ad4c628556098846699ab03","refsource":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","tags":[],"title":"","mime":"","httpstatus":"","archivestatus":"0"},{"url":"https://git.kernel.org/stable/c/0193e83368022a308d19ee0293f55a4f5d548e40","name":"https://git.kernel.org/stable/c/0193e83368022a308d19ee0293f55a4f5d548e40","refsource":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","tags":[],"title":"","mime":"","httpstatus":"","archivestatus":"0"},{"url":"https://www.cve.org/CVERecord?id=CVE-2026-99648","name":"CVE Program record","refsource":"CVE.ORG","tags":["canonical"]},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2026-99648","name":"NVD vulnerability detail","refsource":"NVD","tags":["canonical","analysis"]}],"affected":[{"source":"CNA","vendor":"Linux","product":"Linux","version":"affected 38607c62b34b46317c46d5baf1df03ac6e48a1c6 f437fcb76d448a396940498aba278ac0ad8d715a git","platforms":[]},{"source":"CNA","vendor":"Linux","product":"Linux","version":"affected 38607c62b34b46317c46d5baf1df03ac6e48a1c6 456932cdc6eda70b5ad4c628556098846699ab03 git","platforms":[]},{"source":"CNA","vendor":"Linux","product":"Linux","version":"affected 38607c62b34b46317c46d5baf1df03ac6e48a1c6 0193e83368022a308d19ee0293f55a4f5d548e40 git","platforms":[]},{"source":"CNA","vendor":"Linux","product":"Linux","version":"affected 6.15","platforms":[]},{"source":"CNA","vendor":"Linux","product":"Linux","version":"unaffected 6.15 semver","platforms":[]},{"source":"CNA","vendor":"Linux","product":"Linux","version":"unaffected 6.18.64 6.18.* semver","platforms":[]},{"source":"CNA","vendor":"Linux","product":"Linux","version":"unaffected 7.3.5 7.3.* semver","platforms":[]},{"source":"CNA","vendor":"Linux","product":"Linux","version":"unaffected 7.4-rc2 * original_commit_for_fix","platforms":[]}],"timeline":[],"solutions":[],"workarounds":[],"exploits":[],"credits":[],"nvd_cpes":[],"vendor_comments":[],"enrichments":{"kev":null,"epss":null,"legacy_qids":[]},"source_records":{"cve_program":{"containers":{"cna":{"affected":[{"defaultStatus":"unaffected","product":"Linux","programFiles":["mm/vma.c"],"repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","vendor":"Linux","versions":[{"lessThan":"f437fcb76d448a396940498aba278ac0ad8d715a","status":"affected","version":"38607c62b34b46317c46d5baf1df03ac6e48a1c6","versionType":"git"},{"lessThan":"456932cdc6eda70b5ad4c628556098846699ab03","status":"affected","version":"38607c62b34b46317c46d5baf1df03ac6e48a1c6","versionType":"git"},{"lessThan":"0193e83368022a308d19ee0293f55a4f5d548e40","status":"affected","version":"38607c62b34b46317c46d5baf1df03ac6e48a1c6","versionType":"git"}]},{"defaultStatus":"affected","product":"Linux","programFiles":["mm/vma.c"],"repo":"https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git","vendor":"Linux","versions":[{"status":"affected","version":"6.15"},{"lessThan":"6.15","status":"unaffected","version":"0","versionType":"semver"},{"lessThanOrEqual":"6.18.*","status":"unaffected","version":"6.18.64","versionType":"semver"},{"lessThanOrEqual":"7.3.*","status":"unaffected","version":"7.3.5","versionType":"semver"},{"lessThanOrEqual":"*","status":"unaffected","version":"7.4-rc2","versionType":"original_commit_for_fix"}]}],"cpeApplicability":[{"nodes":[{"cpeMatch":[{"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionEndExcluding":"6.18.64","versionStartIncluding":"6.15","vulnerable":true},{"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionEndExcluding":"7.3.5","versionStartIncluding":"6.15","vulnerable":true},{"criteria":"cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*","versionEndExcluding":"7.4-rc2","versionStartIncluding":"6.15","vulnerable":true}],"negate":false,"operator":"OR"}]}],"descriptions":[{"lang":"en","value":"In the Linux kernel, the following vulnerability has been resolved:\n\nmm/vma: call ->may_split() before a partial merge\n\nA driver can use the ->may_split() hook to forbid a VMA boundary at\ncertain addresses.\n\nA device-dax device is a character device that gives raw access to a\nrange of memory through mmap(). device-dax refuses any address that is not\na multiple of the device's alignment through vm_ops->may_split.\n\nvma_merge_existing_range() can move boundaries, but does not call\nvm_ops->may_split.\n\nIf a range at the edge of a VMA gets the attributes of the adjacent VMA,\nthe function moves only that range over to the neighbour: one VMA shrinks\nand the other grows. This is a partial merge, and vma_modify() tries it\nbefore any split.\n\nTake a device-dax VMA that lies next to another VMA of the same file.\nThe user calls mprotect() on the one page of the first VMA that touches\nthe second VMA, and asks for the protection that the second VMA already\nhas. The result is a partial merge. It puts the boundary inside a 1 GiB\nmapping, which is a single leaf entry at the PUD level.\n\nAs part of the merge, the kernel looks for a huge PMD to split at the\nnew boundary. That code does not expect a PUD leaf:\n\n    vma_merge_existing_range()\n      commit_merge()\n        vma_adjust_trans_huge()\n          split_huge_pmd_address()\n            mm_find_pmd()\n              treats the PUD leaf as a pointer to a PMD table,\n              so the \"table\" is the device memory that the leaf maps\n            __split_huge_pmd()\n              reads a PMD that the user wrote to the device memory\n              folio_put() on the folio that this PMD names\n\nBad. While the code must at some point be taught to split PUD entries as\nwell, the issue here is that we move VMA boundaries such that they are\nincompatible with the driver.\n\nFix this by calling ->may_split() on the address that a partial merge\nwould move the boundary to, and by abandoning the merge if the driver\nrefuses. vma_modify() then tries the split, which fails with -EINVAL.\n\nFactor out existing split handling into check_split() and reuse it.\n\nFix generated by an LLM and sent by Chris Mason in private.\n\n[ david: improve description; cleanup the code ]"}],"providerMetadata":{"dateUpdated":"2026-11-20T05:50:02.280Z","orgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","shortName":"Linux"},"references":[{"url":"https://git.kernel.org/stable/c/f437fcb76d448a396940498aba278ac0ad8d715a"},{"url":"https://git.kernel.org/stable/c/456932cdc6eda70b5ad4c628556098846699ab03"},{"url":"https://git.kernel.org/stable/c/0193e83368022a308d19ee0293f55a4f5d548e40"}],"title":"mm/vma: call ->may_split() before a partial merge","x_generator":{"engine":"bippy-1.2.0"}}},"cveMetadata":{"assignerOrgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","assignerShortName":"Linux","cveId":"CVE-2026-99648","datePublished":"2026-11-20T05:50:02.280Z","dateReserved":"2026-09-25T10:25:14.426Z","dateUpdated":"2026-11-20T05:50:02.280Z","state":"PUBLISHED"},"dataType":"CVE_RECORD","dataVersion":"5.2"},"nvd":{"publishedDate":"2026-11-20 06:17:58","lastModifiedDate":"2026-11-20 06:17:58","problem_types":[],"metrics":[],"configurations":[]},"legacy_mitre":{"record":{"CveYear":"2026","CveId":"99648","Ordinal":"1","Title":"mm/vma: call ->may_split() before a partial merge","CVE":"CVE-2026-99648","Year":"2026"},"notes":[{"CveYear":"2026","CveId":"99648","Ordinal":"1","NoteData":"In the Linux kernel, the following vulnerability has been resolved:\n\nmm/vma: call ->may_split() before a partial merge\n\nA driver can use the ->may_split() hook to forbid a VMA boundary at\ncertain addresses.\n\nA device-dax device is a character device that gives raw access to a\nrange of memory through mmap(). device-dax refuses any address that is not\na multiple of the device's alignment through vm_ops->may_split.\n\nvma_merge_existing_range() can move boundaries, but does not call\nvm_ops->may_split.\n\nIf a range at the edge of a VMA gets the attributes of the adjacent VMA,\nthe function moves only that range over to the neighbour: one VMA shrinks\nand the other grows. This is a partial merge, and vma_modify() tries it\nbefore any split.\n\nTake a device-dax VMA that lies next to another VMA of the same file.\nThe user calls mprotect() on the one page of the first VMA that touches\nthe second VMA, and asks for the protection that the second VMA already\nhas. The result is a partial merge. It puts the boundary inside a 1 GiB\nmapping, which is a single leaf entry at the PUD level.\n\nAs part of the merge, the kernel looks for a huge PMD to split at the\nnew boundary. That code does not expect a PUD leaf:\n\n    vma_merge_existing_range()\n      commit_merge()\n        vma_adjust_trans_huge()\n          split_huge_pmd_address()\n            mm_find_pmd()\n              treats the PUD leaf as a pointer to a PMD table,\n              so the \"table\" is the device memory that the leaf maps\n            __split_huge_pmd()\n              reads a PMD that the user wrote to the device memory\n              folio_put() on the folio that this PMD names\n\nBad. While the code must at some point be taught to split PUD entries as\nwell, the issue here is that we move VMA boundaries such that they are\nincompatible with the driver.\n\nFix this by calling ->may_split() on the address that a partial merge\nwould move the boundary to, and by abandoning the merge if the driver\nrefuses. vma_modify() then tries the split, which fails with -EINVAL.\n\nFactor out existing split handling into check_split() and reuse it.\n\nFix generated by an LLM and sent by Chris Mason in private.\n\n[ david: improve description; cleanup the code ]","Type":"Description","Title":"mm/vma: call ->may_split() before a partial merge"}]}}}