CVE-2026-64359

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> nilfs2: reject CLEAN_SEGMENTS ioctl with out-of-range segment numbers<br /> <br /> Syzbot reported a hung task in nilfs_transaction_begin() where multiple<br /> tasks performing chmod() on a nilfs2 mount blocked for over 143 seconds<br /> waiting to acquire ns_segctor_sem for read:<br /> <br /> INFO: task syz.0.17:5918 blocked for more than 143 seconds.<br /> Call Trace:<br /> schedule+0x164/0x360<br /> rwsem_down_read_slowpath+0x6d9/0x940<br /> down_read+0x99/0x2e0<br /> nilfs_transaction_begin+0x364/0x710 fs/nilfs2/segment.c:221<br /> nilfs_setattr+0x124/0x2c0 fs/nilfs2/inode.c:921<br /> notify_change+0xc1a/0xf40<br /> chmod_common+0x273/0x4a0<br /> do_fchmodat+0x12d/0x230<br /> <br /> The writer holding ns_segctor_sem was a concurrent<br /> NILFS_IOCTL_CLEAN_SEGMENTS caller, stuck inside printk while emitting<br /> per-element warnings from nilfs_sufile_updatev():<br /> <br /> __nilfs_msg+0x373/0x450 fs/nilfs2/super.c:78<br /> nilfs_sufile_updatev+0x21c/0x6d0 fs/nilfs2/sufile.c:186<br /> nilfs_sufile_freev fs/nilfs2/sufile.h:93 [inline]<br /> nilfs_free_segments fs/nilfs2/segment.c:1140 [inline]<br /> nilfs_segctor_collect_blocks fs/nilfs2/segment.c:1261 [inline]<br /> nilfs_segctor_do_construct+0x1f55/0x76c0<br /> nilfs_clean_segments+0x3bd/0xa50<br /> nilfs_ioctl_clean_segments fs/nilfs2/ioctl.c:922 [inline]<br /> nilfs_ioctl+0x261f/0x2780<br /> <br /> The root cause is that user-supplied segment numbers are not validated<br /> before nilfs_clean_segments() begins doing work; the range check on<br /> each segnum is performed deep inside the call chain by<br /> nilfs_sufile_updatev(), which emits a nilfs_warn() per invalid entry<br /> while still holding the segctor lock and the sufile mi_sem. Under load<br /> (repeated invocations across multiple mounts saturating the global<br /> printk path), the cumulative printk latency keeps ns_segctor_sem held<br /> long enough to trip the hung_task watchdog, blocking concurrent<br /> operations such as chmod() that need ns_segctor_sem for read.<br /> <br /> Fix by validating the contents of kbufs[4] in nilfs_clean_segments()<br /> immediately after acquiring ns_segctor_sem via nilfs_transaction_lock().<br /> Holding ns_segctor_sem serializes the check against<br /> nilfs_ioctl_resize(), which can modify ns_nsegments, so the validation<br /> uses a consistent value. Out-of-range segment numbers are rejected<br /> with -EINVAL before any segment-cleaning work begins, so the bad<br /> entries never reach the per-element diagnostic path inside<br /> nilfs_sufile_updatev().