CVE-2024-35895

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> bpf, sockmap: Prevent lock inversion deadlock in map delete elem<br /> <br /> syzkaller started using corpuses where a BPF tracing program deletes<br /> elements from a sockmap/sockhash map. Because BPF tracing programs can be<br /> invoked from any interrupt context, locks taken during a map_delete_elem<br /> operation must be hardirq-safe. Otherwise a deadlock due to lock inversion<br /> is possible, as reported by lockdep:<br /> <br /> CPU0 CPU1<br /> ---- ----<br /> lock(&amp;htab-&gt;buckets[i].lock);<br /> local_irq_disable();<br /> lock(&amp;host-&gt;lock);<br /> lock(&amp;htab-&gt;buckets[i].lock);<br /> <br /> lock(&amp;host-&gt;lock);<br /> <br /> Locks in sockmap are hardirq-unsafe by design. We expects elements to be<br /> deleted from sockmap/sockhash only in task (normal) context with interrupts<br /> enabled, or in softirq context.<br /> <br /> Detect when map_delete_elem operation is invoked from a context which is<br /> _not_ hardirq-unsafe, that is interrupts are disabled, and bail out with an<br /> error.<br /> <br /> Note that map updates are not affected by this issue. BPF verifier does not<br /> allow updating sockmap/sockhash from a BPF tracing program today.