{"dataType":"CVE_RECORD","dataVersion":"5.2","cveMetadata":{"cveId":"CVE-2026-97520","assignerOrgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","state":"PUBLISHED","assignerShortName":"Linux","dateReserved":"2026-09-24T16:01:01.152Z","datePublished":"2026-09-24T16:05:20.188Z","dateUpdated":"2026-09-28T05:30:58.172Z"},"containers":{"cna":{"providerMetadata":{"orgId":"416baaa9-dc9f-4396-8d5f-8c081fb06d67","shortName":"Linux","dateUpdated":"2026-09-28T05:30:58.172Z"},"descriptions":[{"lang":"en","value":"In the Linux kernel, the following vulnerability has been resolved:\n\ngfs2: move quota_init qc iterator increment\n\nMove qc++ from the loop body into the for-loop increment\nexpression in gfs2_quota_init().\n\nThis keeps iterator progression explicit and avoids mixing pointer\nadvance with duplicate-slot handling in the loop body."}],"metrics":[{"cvssV3_1":{"version":"3.1","vectorString":"CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:H/A:H","baseScore":7.1,"baseSeverity":"HIGH"},"scenarios":[{"lang":"en","value":"AV:L - The malformed bytes are on-disk struct gfs2_quota_change records in the per-node quota_change file. gfs2_quota_init() reads them via gfs2_meta_ra() from gfs2_get_tree()→get_tree_bdev()→gfs2_fill_super()→gfs2_make_fs_rw(); no NFS/SMB/TCP message carries qc_id or qc_change.\nAC:L - A crafted quota_change file with two non-zero qc_change entries sharing qc_id, the duplicate in the last slot of a QC block, makes the pre-fix body qc++ then memset(qc,0,sizeof(*qc)) run on every read-write mount of that image, with no race.\nPR:N - Crafting the image needs no host account. gfs2_fs_type.fs_flags is only FS_REQUIRES_DEV (no FS_USERNS_MOUNT), so mount_capable() is capable(CAP_SYS_ADMIN) in init_user_ns; that privileged mount of the attacker image is scored as UI:R, not attacker privilege.\nUI:R - gfs2_quota_init() runs only from gfs2_make_fs_rw() after gfs2_fill_super() or gfs2_reconfigure() attaches the attacker-supplied block device. Because gfs2 is not FS_USERNS_MOUNT, a victim or administrator must mount that image read-write before the duplicate-slot memset runs.\nS:U - The out-of-bounds memset in gfs2_quota_init() corrupts page-cache memory in the host kernel that mounted the volume; this is not a KVM/Xen guest-to-host escape or an IOMMU/DMA bypass.\nC:N - The duplicate path only stores 16 zeros through the already-advanced qc pointer; it does not load adjacent kernel bytes or copy them to userspace.\nI:H - Pre-fix, qc++ runs before memset(qc,0,sizeof(*qc)). sd_qc_per_block is (sb_bsize-24)/16, leaving 8 bytes of padding, so a last-slot duplicate stores 8 zeros past bh->b_data+sb_bsize of the gfs2_getbuf() buffer_head: a kernel out-of-bounds write.\nA:H - Those 8 zeros past the QC buffer_head land in the next folio/page (or the next BH in the same folio when sb_bsize