{"id":"CVE-2026-90117","published":"2026-09-17T17:17:03.857","lastModified":"2026-09-17T17:17:03.857","description":"In the Linux kernel, the following vulnerability has been resolved:\n\nntfs: validate usa_ofs before preserving the update sequence number\n\nWhen ntfs_mft_record_alloc() reuses a free mft record it reads the old\nupdate sequence number straight from the on-disk record:\n\n     usn = *(__le16 *)((u8 *)m + le16_to_cpu(m->usa_ofs));\n\nHere m points into the raw $MFT page-cache folio, which still holds\nunvalidated, MST-protected bytes: the folio is read by a plain\niomap_read_folio() and neither post_read_mst_fixup() nor\nntfs_mft_record_check() has run on it (both work on private copies).\nm->usa_ofs is therefore an untrusted u16, and a corrupted record can put\nit past the end of the record so the two-byte read lands outside the\nfolio.  Reading such a record while creating a file gives, under KASAN:\n\n   BUG: KASAN: use-after-free in ntfs_mft_record_alloc+...\n   Read of size 2 at addr ...\n    ntfs_mft_record_alloc -> __ntfs_create -> ntfs_create -> path_openat\n\nOnly preserve the old update sequence number when usa_ofs is even and in\nrange, mirroring the check ntfs_mft_record_check() already applies;\notherwise leave usn zero, which the existing restore below skips.","cvssScore":null,"cvssSeverity":null,"cvssVector":null,"cwes":[],"vendors":[],"products":[],"references":[{"url":"https://git.kernel.org/stable/c/81684340963da2e898eabb8c1e274433d9375bc6","tags":[]},{"url":"https://git.kernel.org/stable/c/ea5a3c30710710c1dc5e483d3313309ec2e868f7","tags":[]}],"exploitRefs":[],"hasPoc":false,"ai":null}