Executive brief
The Linux kernel's network filesystem (netfs) layer has a bug in its write operations that can leave file pages in an inconsistent state when memory allocation fails during page writeback. This can cause data integrity issues where pending writes may not be properly retried, potentially leading to data loss or filesystem corruption if a system experiences memory pressure.
Technical details
The vulnerability is a missing error-recovery code path in the netfs writeback iteration logic (fs/netfs/write_issue.c). When netfs_write_folio() returns -ENOMEM during the writeback loop, the folio (kernel abstraction for a page) remains in a locked, dirty state without being properly redirtied and unlocked before the iteration terminates. This causes the page to become orphaned with incorrect metadata. The fix adds a conditional check that properly calls folio_redirty_for_writepage() and folio_unlock() when -ENOMEM is encountered. The vulnerability affects systems performing network filesystem write operations under memory pressure. No evidence of active exploitation is reported.
Affected products
- Linux Linux kernel various (patch backported to linux-6.x, linux-5.x, and earlier stable branches)
Timeline
- 2026-08-15: disclosed
- 2026-06-25: patched: Upstream commit b6a713fd34b9498ee2164d5d3e8460732a392efc