{"id":"CVE-2026-89902","published":"2026-09-16T11:16:59.143","lastModified":"2026-09-16T15:18:16.307","description":"In the Linux kernel, the following vulnerability has been resolved:\n\nLoongArch: Avoid preempt count underflow without probe\n\nLoongArch uses break 11 for the breakpoint placed after an instruction\nthat Kprobes executes out of line. Since userspace can issue the same\nbreak instruction, do_bp() can reach kprobe_singlestep_handler() when\nthere is no current probe.\n\nThe handler actually returns false in this case, but it first calls\npreempt_enable_no_resched(). The corresponding preempt_disable() is done\nby kprobe_breakpoint_handler() on a real Kprobe hit, so it has not run\nhere. As a result, an ordinary userspace breakpoint (code 11) underflows\nthe current task's preempt count.\n\nThis also makes in_interrupt() return true until the task schedules. One\nvisible consequence is the socket cgroup attribution: cgroup_sk_alloc()\ntreats the allocation as interrupt context and assigns the socket to the\nroot cgroup. A socket opened from the SIGTRAP handler can then avoid a\nBPF_CGROUP_INET_SOCK_CREATE policy attached to the task's own cgroup.\n\nReturn as soon as kprobe_running() reports no active probe.\n\nThe same check has appeared in [PATCH v10 2/4] of the original LoongArch\nKprobes series, but was dropped before the feature reached mainline.","cvssScore":7.8,"cvssSeverity":"HIGH","cvssVector":"CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H","cwes":[],"vendors":[],"products":[],"references":[{"url":"https://git.kernel.org/stable/c/72ce4b24676e8b3b75376c4c559dd81c1ac52d5a","tags":[]},{"url":"https://git.kernel.org/stable/c/76395109a051b53860faaec96522ef08ae9e8e27","tags":[]},{"url":"https://git.kernel.org/stable/c/a1cb727fa27c2d9c7859729cd90db7bc0f46957f","tags":[]},{"url":"https://git.kernel.org/stable/c/c4357ebeee63cb8da1446fc936573a34e68b0f71","tags":[]},{"url":"https://git.kernel.org/stable/c/ff40e597d5d98c0c45a4d09df8177202fac9520e","tags":[]}],"exploitRefs":[],"hasPoc":false,"ai":null}