{ "schema_version": "1.6.1", "id": "RPI-CVE-2023-54173", "modified": "2026-10-09T11:39:44Z", "upstream": [ "CVE-2023-54173" ], "published": "2025-12-30T12:09:43Z", "summary": "bpf: Disable preemption in bpf_event_output", "details": "In the Linux kernel, the following vulnerability has been resolved:\n\nbpf: Disable preemption in bpf_event_output\n\nWe received report [1] of kernel crash, which is caused by\nusing nesting protection without disabled preemption.\n\nThe bpf_event_output can be called by programs executed by\nbpf_prog_run_array_cg function that disabled migration but\nkeeps preemption enabled.\n\nThis can cause task to be preempted by another one inside the\nnesting protection and lead eventually to two tasks using same\nperf_sample_data buffer and cause crashes like:\n\n BUG: kernel NULL pointer dereference, address: 0000000000000001\n #PF: supervisor instruction fetch in kernel mode\n #PF: error_code(0x0010) - not-present page\n ...\n ? perf_output_sample+0x12a/0x9a0\n ? finish_task_switch.isra.0+0x81/0x280\n ? perf_event_output+0x66/0xa0\n ? bpf_event_output+0x13a/0x190\n ? bpf_event_output_data+0x22/0x40\n ? bpf_prog_dfc84bbde731b257_cil_sock4_connect+0x40a/0xacb\n ? xa_load+0x87/0xe0\n ? __cgroup_bpf_run_filter_sock_addr+0xc1/0x1a0\n ? release_sock+0x3e/0x90\n ? sk_setsockopt+0x1a1/0x12f0\n ? udp_pre_connect+0x36/0x50\n ? inet_dgram_connect+0x93/0xa0\n ? __sys_connect+0xb4/0xe0\n ? udp_setsockopt+0x27/0x40\n ? __pfx_udp_push_pending_frames+0x10/0x10\n ? __sys_setsockopt+0xdf/0x1a0\n ? __x64_sys_connect+0xf/0x20\n ? do_syscall_64+0x3a/0x90\n ? entry_SYSCALL_64_after_hwframe+0x72/0xdc\n\nFixing this by disabling preemption in bpf_event_output.\n\n[1] https://github.com/cilium/cilium/issues/26756", "severity": [ { "type": "CVSS_V3", "score": "CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H" } ], "affected": [ { "package": { "ecosystem": "Raspberry Pi OS:12", "name": "linux" }, "versions": [ "1:6.1.21-1+rpt1" ], "ecosystem_specific": { "binaries": [ { "binary_name": "linux-image-6.1.0-rpi1-rpi-v6", "binary_version": "1:6.1.21-1+rpt1" }, { "binary_name": "linux-image-6.1.0-rpi1-rpi-v7", "binary_version": "1:6.1.21-1+rpt1" }, { "binary_name": "linux-image-6.1.0-rpi1-rpi-v7l", "binary_version": "1:6.1.21-1+rpt1" }, { "binary_name": "linux-image-6.1.0-rpi1-rpi-v8", "binary_version": "1:6.1.21-1+rpt1" } ] }, "database_specific": { "per_version": { "1:6.1.21-1+rpt1": { "linux_commit": "0afb5e98488aed7017b9bf321b575d0177feb7ed", "upstream_version": "6.1.21", "assessment": "commit ancestry (strak rule)", "upstream_fixed": "6.1.45" } } } } ], "references": [ { "type": "ADVISORY", "url": "https://www.cve.org/CVERecord?id=CVE-2023-54173" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/3048cb0dc0cc9dc74ed93690dffef00733bcad5b" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/c81bdf8f9f2b002d217c3d5357cdea9f2b82ff90" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/36dd8ca330b76585640ed32255a3c99f901e1502" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/063c9ce8e74e07bf94f99cd13146f42867875e8b" }, { "type": "FIX", "url": "https://git.kernel.org/stable/c/d62cc390c2e99ae267ffe4b8d7e2e08b6c758c32" } ], "database_specific": { "source": "https://git.kernel.org/pub/scm/linux/security/vulns.git", "upstream_modified": "2026-08-05T11:24:39Z", "severity": { "source": "kernel CNA", "cvss_v3_score": 7.8, "cvss_v3_severity": "High", "scenario": "AV:L - The vulnerable window is entered from local syscalls (connect/sendmsg/setsockopt/close) that invoke cgroup BPF programs via bpf_prog_run_array_cg in preemptible process context. Remote packets only reach this code through softirq paths where preemption is already disabled, so exploitation requires local execution.\nAC:L - The attacker controls both sides of the race \u2014 multiple threads pinned to the same CPU repeatedly invoking the helper, plus self-generated scheduling pressure to force preemption inside the unprotected nest-level window. The required preemptible kernel (CONFIG_PREEMPT/PREEMPT_DYNAMIC/PREEMPT_RT) and a perf-emitting eBPF agent are the default in Android, container, and cloud deployments.\nPR:L - Any unprivileged local process \u2014 including code confined inside a container or pod \u2014 triggers the cgroup BPF program merely by calling connect(), sendmsg(), or setsockopt(); no capability, CAP_BPF, or namespace privilege is required on the triggering side.\nUI:N - No victim action is needed; the attacker's own syscalls drive both racing invocations, and the presence of an attached eBPF monitoring program is ambient system configuration rather than user interaction.\nS:U - The corruption and its consequences are confined to the kernel's own security authority with no crossing of a hypervisor, IOMMU, or sandbox boundary. This is standard in-kernel memory corruption / privilege escalation.\nC:H - A hijacked perf_sample_data leaves sd->raw pointing at a dead kernel stack frame, so __output_copy() streams frag->size bytes from an unvalidated frag->data pointer into the perf ring buffer that userspace mmaps \u2014 an arbitrary kernel memory read primitive into attacker-readable memory.\nI:H - perf_output_sample() performs an indirect call through frag->copy read from the stale stack frame (__output_custom), which the mainline reports observed as executing an NX-protected page and fetching instructions from address 0x1. Since the stale frame belongs to an attacker-controlled task, this is a groomable control-flow hijack primitive.\nA:H - The bug was reported from production as a hard kernel crash \u2014 NULL pointer dereference / supervisor instruction fetch oops in perf_output_sample \u2014 and can be triggered repeatedly by any local user, producing a reliable denial of service." } } }