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 --- include/xen/interface/sched.h | 166 ++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 166 insertions(+) create mode 100644 include/xen/interface/sched.h (limited to 'include/xen/interface/sched.h') diff --git a/include/xen/interface/sched.h b/include/xen/interface/sched.h new file mode 100644 index 000000000..4dac0634f --- /dev/null +++ b/include/xen/interface/sched.h @@ -0,0 +1,166 @@ +/* SPDX-License-Identifier: MIT */ +/****************************************************************************** + * sched.h + * + * Scheduler state interactions + * + * Copyright (c) 2005, Keir Fraser + */ + +#ifndef __XEN_PUBLIC_SCHED_H__ +#define __XEN_PUBLIC_SCHED_H__ + +#include + +/* + * Guest Scheduler Operations + * + * The SCHEDOP interface provides mechanisms for a guest to interact + * with the scheduler, including yield, blocking and shutting itself + * down. + */ + +/* + * The prototype for this hypercall is: + * long HYPERVISOR_sched_op(enum sched_op cmd, void *arg, ...) + * + * @cmd == SCHEDOP_??? (scheduler operation). + * @arg == Operation-specific extra argument(s), as described below. + * ... == Additional Operation-specific extra arguments, described below. + * + * Versions of Xen prior to 3.0.2 provided only the following legacy version + * of this hypercall, supporting only the commands yield, block and shutdown: + * long sched_op(int cmd, unsigned long arg) + * @cmd == SCHEDOP_??? (scheduler operation). + * @arg == 0 (SCHEDOP_yield and SCHEDOP_block) + * == SHUTDOWN_* code (SCHEDOP_shutdown) + * + * This legacy version is available to new guests as: + * long HYPERVISOR_sched_op_compat(enum sched_op cmd, unsigned long arg) + */ + +/* + * Voluntarily yield the CPU. + * @arg == NULL. + */ +#define SCHEDOP_yield 0 + +/* + * Block execution of this VCPU until an event is received for processing. + * If called with event upcalls masked, this operation will atomically + * reenable event delivery and check for pending events before blocking the + * VCPU. This avoids a "wakeup waiting" race. + * @arg == NULL. + */ +#define SCHEDOP_block 1 + +/* + * Halt execution of this domain (all VCPUs) and notify the system controller. + * @arg == pointer to sched_shutdown structure. + * + * If the sched_shutdown_t reason is SHUTDOWN_suspend then + * x86 PV guests must also set RDX (EDX for 32-bit guests) to the MFN + * of the guest's start info page. RDX/EDX is the third hypercall + * argument. + * + * In addition, which reason is SHUTDOWN_suspend this hypercall + * returns 1 if suspend was cancelled or the domain was merely + * checkpointed, and 0 if it is resuming in a new domain. + */ +#define SCHEDOP_shutdown 2 + +/* + * Poll a set of event-channel ports. Return when one or more are pending. An + * optional timeout may be specified. + * @arg == pointer to sched_poll structure. + */ +#define SCHEDOP_poll 3 + +/* + * Declare a shutdown for another domain. The main use of this function is + * in interpreting shutdown requests and reasons for fully-virtualized + * domains. A para-virtualized domain may use SCHEDOP_shutdown directly. + * @arg == pointer to sched_remote_shutdown structure. + */ +#define SCHEDOP_remote_shutdown 4 + +/* + * Latch a shutdown code, so that when the domain later shuts down it + * reports this code to the control tools. + * @arg == sched_shutdown, as for SCHEDOP_shutdown. + */ +#define SCHEDOP_shutdown_code 5 + +/* + * Setup, poke and destroy a domain watchdog timer. + * @arg == pointer to sched_watchdog structure. + * With id == 0, setup a domain watchdog timer to cause domain shutdown + * after timeout, returns watchdog id. + * With id != 0 and timeout == 0, destroy domain watchdog timer. + * With id != 0 and timeout != 0, poke watchdog timer and set new timeout. + */ +#define SCHEDOP_watchdog 6 + +/* + * Override the current vcpu affinity by pinning it to one physical cpu or + * undo this override restoring the previous affinity. + * @arg == pointer to sched_pin_override structure. + * + * A negative pcpu value will undo a previous pin override and restore the + * previous cpu affinity. + * This call is allowed for the hardware domain only and requires the cpu + * to be part of the domain's cpupool. + */ +#define SCHEDOP_pin_override 7 + +struct sched_shutdown { + unsigned int reason; /* SHUTDOWN_* => shutdown reason */ +}; +DEFINE_GUEST_HANDLE_STRUCT(sched_shutdown); + +struct sched_poll { + GUEST_HANDLE(evtchn_port_t) ports; + unsigned int nr_ports; + uint64_t timeout; +}; +DEFINE_GUEST_HANDLE_STRUCT(sched_poll); + +struct sched_remote_shutdown { + domid_t domain_id; /* Remote domain ID */ + unsigned int reason; /* SHUTDOWN_* => shutdown reason */ +}; +DEFINE_GUEST_HANDLE_STRUCT(sched_remote_shutdown); + +struct sched_watchdog { + uint32_t id; /* watchdog ID */ + uint32_t timeout; /* timeout */ +}; +DEFINE_GUEST_HANDLE_STRUCT(sched_watchdog); + +struct sched_pin_override { + int32_t pcpu; +}; +DEFINE_GUEST_HANDLE_STRUCT(sched_pin_override); + +/* + * Reason codes for SCHEDOP_shutdown. These may be interpreted by control + * software to determine the appropriate action. For the most part, Xen does + * not care about the shutdown code. + */ +#define SHUTDOWN_poweroff 0 /* Domain exited normally. Clean up and kill. */ +#define SHUTDOWN_reboot 1 /* Clean up, kill, and then restart. */ +#define SHUTDOWN_suspend 2 /* Clean up, save suspend info, kill. */ +#define SHUTDOWN_crash 3 /* Tell controller we've crashed. */ +#define SHUTDOWN_watchdog 4 /* Restart because watchdog time expired. */ + +/* + * Domain asked to perform 'soft reset' for it. The expected behavior is to + * reset internal Xen state for the domain returning it to the point where it + * was created but leaving the domain's memory contents and vCPU contexts + * intact. This will allow the domain to start over and set up all Xen specific + * interfaces again. + */ +#define SHUTDOWN_soft_reset 5 +#define SHUTDOWN_MAX 5 /* Maximum valid shutdown reason. */ + +#endif /* __XEN_PUBLIC_SCHED_H__ */ -- cgit v1.3.1