Linux Kernel EUVD-2026-32463

OCFS2 (Oracle Cluster Filesystem 2) is a shared-disk cluster filesystem integrated into the Linux kernel, using JBD2 as its journaling layer. JBD2 uses a credit pre-allocation model where each filesystem operation must reserve a fixed number of journal block slots before proceeding; the per-journal maximum is 5449 credits. The vulnerable path begins in ocfs2_dio_end_io_write(), which handles completion of direct I/O writes by calling ocfs2_mark_extent_written() to transition extents from unwritten to written state. Under certain fragmentation patterns, ocfs2_split_extent() invokes ocfs2_try_to_merge_extent(), which in turn calls ocfs2_extend_rotate_transaction() to request additional credits mid-transaction. The cumulative credit demand can exceed the JBD2 maximum (observed at 5450 vs. the 5449 limit), causing start_this_handle() to fail with a warning. The fix batches extent processing within bounded transaction slices and repositions the orphan inode removal to occur only after a successful extent tree update, preventing stale blocks beyond EOF on crash. No CWE is formally assigned, but the root cause is a resource management defect in transaction credit accounting - analogous to CWE-400 (Uncontrolled Resource Consumption). CPE: cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*.