From a5f7a5bb3b7f28ba7e4fa246775b29a0e5537255 Mon Sep 17 00:00:00 2001 From: Jinke Han Date: Tue, 8 Sep 2026 15:37:42 +0800 Subject: x86/kprobes: Fix crash when probing CS CALL instructions When using eBPF to probe CS CALL instructions within a function, a crash can be triggered. The eBPF tool probes offset 257 of the __hrtimer_run_queues() function: <__hrtimer_run_queues+249>: nopl 0x0(%rax,%rax,1) <__hrtimer_run_queues+254>: mov %r14,%rdi <__hrtimer_run_queues+257>: cs call <__x86_indirect_thunk_r12> <__hrtimer_run_queues+263>: mov %eax,%r12d <__hrtimer_run_queues+266>: xchg %ax,%ax <__hrtimer_run_queues+268>: mov %r13,%rdi Which triggers this crash: BUG: unable to handle page fault for address: 00000000000f41c9 #PF: supervisor write access in kernel mode #PF: error_code(0x0002) - not-present page PGD 0 P4D 0 Oops: 0002 [#1] SMP NOPTI CPU: 1 PID: 0 Comm: swapper/1 Kdump: loaded Tainted: P RIP: 0010:__hrtimer_run_queues+0x106/0x230 Note that __hrtimer_run_queues+0x106 is __hrtimer_run_queues+262, which is at the 6th byte of the above CS CALL instruction. Since the CS CALL instruction occupies 6 bytes, the exception occurred in the middle of that call instruction. The root cause is that when using eBPF tools to probe in the middle of a function, a kprobe with INT3 is used as the underlying implementation. During single-step emulation of the original CALL instruction, int3_emulate_call() assumes that the probed CALL instruction is 5 bytes long. However, the actual CS-prefixed CALL instruction occupies 6 bytes, so it constructs an incorrect exception return address. When the CPU returns from the kprobe handler, the next instruction to be executed is at the address of the last byte of that CS CALL instruction. Coincidentally, starting from that address, the CPU fetches and decodes a completely different instruction, which ultimately triggers a kernel crash. Fix the issue by using the actual instruction length obtained from the instruction decoder when constructing the exception return address, rather than relying on the hardcoded CALL_INSN_SIZE macro. [ mingo: Refined the changelog ] Fixes: 6256e668b7af ("x86/kprobes: Use int3 instead of debug trap for single-step") Suggested-by: Masami Hiramatsu (Google) Signed-off-by: Jinke Han Signed-off-by: Ingo Molnar Reviewed-by: Masami Hiramatsu (Google) Acked-by: Yafang Shao Acked-by: Borislav Petkov Cc: Peter Zijlstra Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000 --- rust/proc-macro2/detection.rs | 77 +++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 77 insertions(+) create mode 100644 rust/proc-macro2/detection.rs (limited to 'rust/proc-macro2/detection.rs') diff --git a/rust/proc-macro2/detection.rs b/rust/proc-macro2/detection.rs new file mode 100644 index 000000000..3de448cb2 --- /dev/null +++ b/rust/proc-macro2/detection.rs @@ -0,0 +1,77 @@ +// SPDX-License-Identifier: Apache-2.0 OR MIT + +use core::sync::atomic::{AtomicUsize, Ordering}; +use std::sync::Once; + +static WORKS: AtomicUsize = AtomicUsize::new(0); +static INIT: Once = Once::new(); + +pub(crate) fn inside_proc_macro() -> bool { + match WORKS.load(Ordering::Relaxed) { + 1 => return false, + 2 => return true, + _ => {} + } + + INIT.call_once(initialize); + inside_proc_macro() +} + +pub(crate) fn force_fallback() { + WORKS.store(1, Ordering::Relaxed); +} + +pub(crate) fn unforce_fallback() { + initialize(); +} + +#[cfg(not(no_is_available))] +fn initialize() { + let available = proc_macro::is_available(); + WORKS.store(available as usize + 1, Ordering::Relaxed); +} + +// Swap in a null panic hook to avoid printing "thread panicked" to stderr, +// then use catch_unwind to determine whether the compiler's proc_macro is +// working. When proc-macro2 is used from outside of a procedural macro all +// of the proc_macro crate's APIs currently panic. +// +// The Once is to prevent the possibility of this ordering: +// +// thread 1 calls take_hook, gets the user's original hook +// thread 1 calls set_hook with the null hook +// thread 2 calls take_hook, thinks null hook is the original hook +// thread 2 calls set_hook with the null hook +// thread 1 calls set_hook with the actual original hook +// thread 2 calls set_hook with what it thinks is the original hook +// +// in which the user's hook has been lost. +// +// There is still a race condition where a panic in a different thread can +// happen during the interval that the user's original panic hook is +// unregistered such that their hook is incorrectly not called. This is +// sufficiently unlikely and less bad than printing panic messages to stderr +// on correct use of this crate. Maybe there is a libstd feature request +// here. For now, if a user needs to guarantee that this failure mode does +// not occur, they need to call e.g. `proc_macro2::Span::call_site()` from +// the main thread before launching any other threads. +#[cfg(no_is_available)] +fn initialize() { + use std::panic::{self, PanicInfo}; + + type PanicHook = dyn Fn(&PanicInfo) + Sync + Send + 'static; + + let null_hook: Box = Box::new(|_panic_info| { /* ignore */ }); + let sanity_check = &*null_hook as *const PanicHook; + let original_hook = panic::take_hook(); + panic::set_hook(null_hook); + + let works = panic::catch_unwind(proc_macro::Span::call_site).is_ok(); + WORKS.store(works as usize + 1, Ordering::Relaxed); + + let hopefully_null_hook = panic::take_hook(); + panic::set_hook(original_hook); + if sanity_check != &*hopefully_null_hook { + panic!("observed race condition in proc_macro2::inside_proc_macro"); + } +} -- cgit v1.3.1