{ "dataType": "CVE_RECORD", "dataVersion": "5.2", "cveMetadata": { "cveId": "CVE-2025-39779", "assignerOrgId": "416baaa9-dc9f-4396-8d5f-8c081fb06d67", "state": "PUBLISHED", "assignerShortName": "Linux", "dateReserved": "2025-04-16T07:20:57.130Z", "datePublished": "2025-09-11T16:56:30.395Z", "dateUpdated": "2026-08-05T12:05:04.307Z" }, "containers": { "cna": { "providerMetadata": { "orgId": "416baaa9-dc9f-4396-8d5f-8c081fb06d67", "shortName": "Linux", "dateUpdated": "2026-08-05T12:05:04.307Z" }, "descriptions": [ { "lang": "en", "value": "In the Linux kernel, the following vulnerability has been resolved:\n\nbtrfs: subpage: keep TOWRITE tag until folio is cleaned\n\nbtrfs_subpage_set_writeback() calls folio_start_writeback() the first time\na folio is written back, and it also clears the PAGECACHE_TAG_TOWRITE tag\neven if there are still dirty blocks in the folio. This can break ordering\nguarantees, such as those required by btrfs_wait_ordered_extents().\n\nThat ordering breakage leads to a real failure. For example, running\ngeneric/464 on a zoned setup will hit the following ASSERT. This happens\nbecause the broken ordering fails to flush existing dirty pages before the\nfile size is truncated.\n\n assertion failed: !list_empty(&ordered->list) :: 0, in fs/btrfs/zoned.c:1899\n ------------[ cut here ]------------\n kernel BUG at fs/btrfs/zoned.c:1899!\n Oops: invalid opcode: 0000 [#1] SMP NOPTI\n CPU: 2 UID: 0 PID: 1906169 Comm: kworker/u130:2 Kdump: loaded Not tainted 6.16.0-rc6-BTRFS-ZNS+ #554 PREEMPT(voluntary)\n Hardware name: Supermicro Super Server/H12SSL-NT, BIOS 2.0 02/22/2021\n Workqueue: btrfs-endio-write btrfs_work_helper [btrfs]\n RIP: 0010:btrfs_finish_ordered_zoned.cold+0x50/0x52 [btrfs]\n RSP: 0018:ffffc9002efdbd60 EFLAGS: 00010246\n RAX: 000000000000004c RBX: ffff88811923c4e0 RCX: 0000000000000000\n RDX: 0000000000000000 RSI: ffffffff827e38b1 RDI: 00000000ffffffff\n RBP: ffff88810005d000 R08: 00000000ffffdfff R09: ffffffff831051c8\n R10: ffffffff83055220 R11: 0000000000000000 R12: ffff8881c2458c00\n R13: ffff88811923c540 R14: ffff88811923c5e8 R15: ffff8881c1bd9680\n FS: 0000000000000000(0000) GS:ffff88a04acd0000(0000) knlGS:0000000000000000\n CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033\n CR2: 00007f907c7a918c CR3: 0000000004024000 CR4: 0000000000350ef0\n Call Trace:\n \n ? srso_return_thunk+0x5/0x5f\n btrfs_finish_ordered_io+0x4a/0x60 [btrfs]\n btrfs_work_helper+0xf9/0x490 [btrfs]\n process_one_work+0x204/0x590\n ? srso_return_thunk+0x5/0x5f\n worker_thread+0x1d6/0x3d0\n ? __pfx_worker_thread+0x10/0x10\n kthread+0x118/0x230\n ? __pfx_kthread+0x10/0x10\n ret_from_fork+0x205/0x260\n ? __pfx_kthread+0x10/0x10\n ret_from_fork_asm+0x1a/0x30\n \n\nConsider process A calling writepages() with WB_SYNC_NONE. In zoned mode or\nfor compressed writes, it locks several folios for delalloc and starts\nwriting them out. Let's call the last locked folio folio X. Suppose the\nwrite range only partially covers folio X, leaving some pages dirty.\nProcess A calls btrfs_subpage_set_writeback() when building a bio. This\nfunction call clears the TOWRITE tag of folio X, whose size = 8K and\nthe block size = 4K. It is following state.\n\n 0 4K 8K\n |/////|/////| (flag: DIRTY, tag: DIRTY)\n <-----> Process A will write this range.\n\nNow suppose process B concurrently calls writepages() with WB_SYNC_ALL. It\ncalls tag_pages_for_writeback() to tag dirty folios with\nPAGECACHE_TAG_TOWRITE. Since folio X is still dirty, it gets tagged. Then,\nB collects tagged folios using filemap_get_folios_tag() and must wait for\nfolio X to be written before returning from writepages().\n\n 0 4K 8K\n |/////|/////| (flag: DIRTY, tag: DIRTY|TOWRITE)\n\nHowever, between tagging and collecting, process A may call\nbtrfs_subpage_set_writeback() and clear folio X's TOWRITE tag.\n 0 4K 8K\n | |/////| (flag: DIRTY|WRITEBACK, tag: DIRTY)\n\nAs a result, process B won't see folio X in its batch, and returns without\nwaiting for it. This breaks the WB_SYNC_ALL ordering requirement.\n\nFix this by using btrfs_subpage_set_writeback_keepwrite(), which retains\nthe TOWRITE tag. We now manually clear the tag only after the folio becomes\nclean, via the xas operation." } ], "metrics": [ { "cvssV3_1": { "version": "3.1", "vectorString": "CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H", "baseScore": 7.8, "baseSeverity": "HIGH" }, "scenarios": [ { "lang": "en", "value": "AV:L - The vulnerable path is btrfs buffered-writeback reached only through local file syscalls (write/mmap dirtying, sync_file_range, fsync, ftruncate) on a mounted btrfs filesystem; there is no network or remote-peer input path into btrfs_subpage_set_writeback(). Exploitation requires the ability to run code on the machine.\nAC:L - The attacker drives both sides of the race himself — one thread issuing sync_file_range(SYNC_FILE_RANGE_WRITE) (WB_SYNC_NONE) over a partial range of a folio while another loops on fsync()/ftruncate() (WB_SYNC_ALL) — and can retry indefinitely until the window between tag_pages_for_writeback() and filemap_get_folios_tag() is hit. Subpage mode (sectorsize < folio size) is a standard shipping deployment on 16K/64K-page arm64 and ppc64 kernels, not a rare debug option, and the compression/zoned delalloc behaviour that leaves partial dirty folios comes from ordinary mount options.\nPR:L - Only an ordinary unprivileged user with write access to any file on the btrfs filesystem (e.g. /home or /tmp) is needed — no capabilities, namespaces, or root, since dirtying pages and calling fsync/sync_file_range/ftruncate on one's own file is enough. The attacker owns the file involved in the race.\nUI:N - The attacker performs every step on a file it creates itself; no victim must mount a filesystem, open a file, or take any action. The already-mounted btrfs is normal system state.\nS:U - The broken writeback ordering, the empty-list dereference, and the resulting filesystem metadata damage are all handled by the same kernel security authority. No VM, IOMMU, or sandbox boundary is crossed.\nC:H - With CONFIG_BTRFS_ASSERT off, btrfs_finish_ordered_zoned() performs list_first_entry() on an empty list head, an out-of-bounds read of kernel memory preceding btrfs_ordered_extent::list, and that garbage value is written into ordered->disk_bytenr / em->disk_bytenr and persisted to the file extent item, remapping the attacker's file onto an arbitrary filesystem logical address whose contents (other files' data or freed blocks) can then be read back. The same ordering breakage lets truncate proceed without flushing, the exact condition btrfs's truncated_len machinery exists to prevent from exposing stale on-disk data.\nI:H - The corrupted logical address is committed into the extent map and the on-disk file extent item, an uncontrolled modification of filesystem metadata that can also misdirect extent-tree reference accounting onto blocks owned by other files. Independently, WB_SYNC_ALL writepages returning before the folio is written means fsync()/sync() acknowledge data that was never persisted, violating the durability and transaction-commit ordering guarantees btrfs_wait_ordered_extents() provides.\nA:H - The bug was reproduced as a hard kernel BUG — \"assertion failed: !list_empty(&ordered->list)\" / \"Oops: invalid opcode\" in btrfs_finish_ordered_zoned() from a btrfs-endio-write worker — which panics the machine. Even without CONFIG_BTRFS_ASSERT, the fabricated btrfs_ordered_sum pointer and the possibly-NULL extent map in btrfs_rewrite_logical_zoned() readily oops the kernel." } ] } ], "affected": [ { "product": "Linux", "vendor": "Linux", "defaultStatus": "unaffected", "repo": "https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git", "programFiles": [ "fs/btrfs/subpage.c" ], "versions": [ { "version": "3470da3b7d87fde4b9729d4e29c2dc074cd59c10", "lessThan": "3d61136945a7008fc90d013c3c67007ce0c96131", "status": "affected", "versionType": "git" }, { "version": "3470da3b7d87fde4b9729d4e29c2dc074cd59c10", "lessThan": "bce7a5c77a1e7a759e227b7713dde18c52da4759", "status": "affected", "versionType": "git" }, { "version": "3470da3b7d87fde4b9729d4e29c2dc074cd59c10", "lessThan": "b1511360c8ac882b0c52caa263620538e8d73220", "status": "affected", "versionType": "git" } ] }, { "product": "Linux", "vendor": "Linux", "defaultStatus": "affected", "repo": "https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git", "programFiles": [ "fs/btrfs/subpage.c" ], "versions": [ { "version": "5.13", "status": "affected" }, { "version": "0", "lessThan": "5.13", "status": "unaffected", "versionType": "semver" }, { "version": "6.12.44", "lessThanOrEqual": "6.12.*", "status": "unaffected", "versionType": "semver" }, { "version": "6.16.4", "lessThanOrEqual": "6.16.*", "status": "unaffected", "versionType": "semver" }, { "version": "6.17", "lessThanOrEqual": "*", "status": "unaffected", "versionType": "original_commit_for_fix" } ] } ], "cpeApplicability": [ { "nodes": [ { "operator": "OR", "negate": false, "cpeMatch": [ { "vulnerable": true, "criteria": "cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*", "versionStartIncluding": "5.13", "versionEndExcluding": "6.12.44" }, { "vulnerable": true, "criteria": "cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*", "versionStartIncluding": "5.13", "versionEndExcluding": "6.16.4" }, { "vulnerable": true, "criteria": "cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*", "versionStartIncluding": "5.13", "versionEndExcluding": "6.17" } ] } ] } ], "references": [ { "url": "https://git.kernel.org/stable/c/3d61136945a7008fc90d013c3c67007ce0c96131" }, { "url": "https://git.kernel.org/stable/c/bce7a5c77a1e7a759e227b7713dde18c52da4759" }, { "url": "https://git.kernel.org/stable/c/b1511360c8ac882b0c52caa263620538e8d73220" } ], "title": "btrfs: subpage: keep TOWRITE tag until folio is cleaned", "x_generator": { "engine": "bippy-1.2.0" } } } }