CVE-2026-90211

*** Pendiente de traducción *** In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> bpf, s390: Clear fetch destination on faulting arena atomic<br /> <br /> Same missing register clear as on riscv64. A RMW atomic on an arena pointer<br /> is converted to BPF_PROBE_ATOMIC and gets an exception table entry, but<br /> bpf_jit_probe_atomic_pre() only fills in the arena base and the probe<br /> offset, leaving probe-&gt;reg at the -1 that bpf_jit_probe_init() set, which<br /> bpf_jit_probe_post() writes into the entry and ex_handler_bpf() then reads<br /> back as "there is nothing to clear".<br /> <br /> That is right for a plain BPF_{ADD,AND,OR,XOR}, which only writes memory,<br /> but an RMW carrying BPF_FETCH also reads the old value into a register:<br /> src_reg for BPF_{ADD,AND,OR,XOR} | BPF_FETCH and BPF_XCHG, and r0 for<br /> BPF_CMPXCHG. So on a fault over an unmapped arena page the program resumes<br /> at the landing pad with whatever that register held before the atomic<br /> instead of the 0 that every other BPF_PROBE_* access delivers.<br /> <br /> Fill probe-&gt;reg in from bpf_atomic_load_reg(). Unlike x86-64 and arm64,<br /> s390x does not report arena violations from its exception handler, so there<br /> is no access direction to correct here, only the missing register clear.