CVE-2025-38553

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> net/sched: Restrict conditions for adding duplicating netems to qdisc tree<br /> <br /> netem_enqueue&amp;#39;s duplication prevention logic breaks when a netem<br /> resides in a qdisc tree with other netems - this can lead to a<br /> soft lockup and OOM loop in netem_dequeue, as seen in [1].<br /> Ensure that a duplicating netem cannot exist in a tree with other<br /> netems.<br /> <br /> Previous approaches suggested in discussions in chronological order:<br /> <br /> 1) Track duplication status or ttl in the sk_buff struct. Considered<br /> too specific a use case to extend such a struct, though this would<br /> be a resilient fix and address other previous and potential future<br /> DOS bugs like the one described in loopy fun [2].<br /> <br /> 2) Restrict netem_enqueue recursion depth like in act_mirred with a<br /> per cpu variable. However, netem_dequeue can call enqueue on its<br /> child, and the depth restriction could be bypassed if the child is a<br /> netem.<br /> <br /> 3) Use the same approach as in 2, but add metadata in netem_skb_cb<br /> to handle the netem_dequeue case and track a packet&amp;#39;s involvement<br /> in duplication. This is an overly complex approach, and Jamal<br /> notes that the skb cb can be overwritten to circumvent this<br /> safeguard.<br /> <br /> 4) Prevent the addition of a netem to a qdisc tree if its ancestral<br /> path contains a netem. However, filters and actions can cause a<br /> packet to change paths when re-enqueued to the root from netem<br /> duplication, leading us to the current solution: prevent a<br /> duplicating netem from inhabiting the same tree as other netems.<br /> <br /> [1] https://lore.kernel.org/netdev/8DuRWwfqjoRDLDmBMlIfbrsZg9Gx50DHJc1ilxsEBNe2D6NMoigR_eIRIG0LOjMc3r10nUUZtArXx4oZBIdUfZQrwjcQhdinnMis_0G7VEk=@willsroot.io/<br /> [2] https://lwn.net/Articles/719297/