{"id":"CVE-2026-89923","published":"2026-09-16T11:17:01.397","lastModified":"2026-09-16T11:17:01.397","description":"In the Linux kernel, the following vulnerability has been resolved:\n\nKVM: s390: Free guest debug data on vcpu destroy\n\nkvm_s390_clear_bp_data() is only called from\nkvm_arch_vcpu_ioctl_set_guest_debug(), i.e. when user space changes or\ndisables debugging. A vCPU that is destroyed while hardware breakpoints\nare still armed - the normal case when the VMM just exits or crashes -\nleaks hw_bp_info, hw_wp_info and all old_data buffers, since generic KVM\nfrees the vCPU right after kvm_arch_vcpu_destroy().\n\nThat is bounded by MAX_BP_COUNT entries, so roughly 8 KiB per vCPU, but\nit is unbounded over VM lifetimes. The allocations are\nGFP_KERNEL_ACCOUNT, so the charge also outlives the exiting process and\npins dying memcgs.\n\nFix by clearing the debug data on vCPU destruction. Calling it\nunconditionally is fine: struct kvm_vcpu is zero allocated, so for a vCPU\nthat never enabled debugging the counters are 0 and the pointers NULL.","cvssScore":null,"cvssSeverity":null,"cvssVector":null,"cwes":[],"vendors":[],"products":[],"references":[{"url":"https://git.kernel.org/stable/c/053f2068cf9b0654b651fed5ad7649d3deb982bf","tags":[]},{"url":"https://git.kernel.org/stable/c/544a5a665019cef394ce8bda97fa7a2cab253781","tags":[]},{"url":"https://git.kernel.org/stable/c/6c996b12bbe4a6023076e8eee3cfaf2cc99db76c","tags":[]},{"url":"https://git.kernel.org/stable/c/9a32c4c6a6c23b7f5bf765eb5b83ba7f64a8ec63","tags":[]},{"url":"https://git.kernel.org/stable/c/b5acacfdb8966abff91836c1c5bd7da59d8a526e","tags":[]},{"url":"https://git.kernel.org/stable/c/d6280fc540d8f4653c9b54a78cb747e128070f5e","tags":[]},{"url":"https://git.kernel.org/stable/c/e4d8b7f176b976474e0800ba71ee2dfa3ce263c9","tags":[]},{"url":"https://git.kernel.org/stable/c/e7f698b09d4a7c36b299acf680fc50fe868e2bcd","tags":[]}],"exploitRefs":[],"hasPoc":false,"ai":null}