CVE-2026-89501

Quoted from the kernel CNA's own scoring file, word for word. A highlighted metric carries the most severe value CVSS v3.1 defines for it, which is what pushes the score up. source

AV:LAttack vector, local

The bug is reached only via local tracefs syscalls: write() to buffer_subbuf_size_kb (ring_buffer_subbuf_order_set) racing with read()/close() on per_cpu/*/trace_pipe_raw (ring_buffer_alloc_read_page and ring_buffer_free_read_page). No network, adjacent-radio, or physical path reaches this free_page steal.

AC:LAttack complexity, low

The attacker controls both sides of the race: one thread toggles buffer_subbuf_size_kb to copy and NULL cpu_buffer->free_page without cpu_buffer->lock, while another reads or closes trace_pipe_raw, which takes or caches that page under the lock. Success does not depend on an uninfluenced victim state.

PR:LPrivileges required, low

tracing_check_open_get_tr() enforces only LOCKDOWN_TRACEFS and DAC with no capable() check. buffer_subbuf_size_kb and trace_pipe_raw are mode 0640 and are routinely delegated via tracefs gid=/mode= to tracing-group members on Android/Perfetto, ChromeOS, and developer kernels, matching prior CNA tracefs UAF scores.

UI:NUser interaction, none

Exploitation uses the attacker’s own tracefs file descriptors and threads (open/read/close trace_pipe_raw while writing buffer_subbuf_size_kb). No separate victim mount, click, or configuration action is required.

S:UScope, unchanged

The use-after-free and double-free corrupt kernel ring-buffer pages and metadata within the same OS security authority. This enables local privilege escalation but does not cross VM, IOMMU, container-hypervisor, or hardware trust boundaries.

C:HConfidentiality impact, high

ring_buffer_subbuf_order_set() can free the cached free_page after alloc_read_page has already claimed it, including after ring_buffer_read_page swaps it into the live reader page. Recycled heap contents are then copied to userspace via tracing_buffers_read, giving a high-impact disclosure primitive.

I:HIntegrity impact, high

The same race double-frees the page or installs a freed page as the live ring-buffer reader page, so later tracing writes and free_read_page hit attacker-reclaimable heap. This UAF/double-free is exploitable for arbitrary kernel writes and control-flow hijack, not merely a crash.

A:HAvailability impact, high

Concurrent free of a still-in-use reader or spare page corrupts ring-buffer lists and causes kernel oops/panic on UAF or double-free. The attacker can retrigger the race at will through repeated subbuf resizes and raw-buffer reads.