x86/kprobes: Fix crash when probing CS CALL instructionsgrafted

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) <mhiramat@kernel.org>
Signed-off-by: Jinke Han <jinkehan@didiglobal.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Acked-by: Yafang Shao <laoar.shao@gmail.com>
Acked-by: Borislav Petkov <bp@alien8.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000

4 files changed, 162 insertions, 0 deletions

diff --git a/samples/acrn/Makefile b/samples/acrn/Makefile
new file mode 100644
index 000000000..c8e3ed978
--- /dev/null
+++ b/samples/acrn/Makefile
@@ -0,0 +1,12 @@
+# SPDX-License-Identifier: GPL-2.0
+
+.PHONY: vm-sample
+
+vm-sample: vm-sample.o payload.o
+	$(CC) $^ -o $@
+
+payload.o: payload.ld guest16.o
+	$(LD) -T $< -o $@
+
+clean:
+	rm *.o vm-sample
diff --git a/samples/acrn/guest.ld b/samples/acrn/guest.ld
new file mode 100644
index 000000000..5127c682b
--- /dev/null
+++ b/samples/acrn/guest.ld
@@ -0,0 +1,9 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+OUTPUT_FORMAT(binary)
+SECTIONS
+{
+        .start : { *(.start) }
+        .text : { *(.text*) }
+        .rodata : { *(.rodata) }
+        .data : { *(.data) }
+}
diff --git a/samples/acrn/payload.ld b/samples/acrn/payload.ld
new file mode 100644
index 000000000..e8d9a498a
--- /dev/null
+++ b/samples/acrn/payload.ld
@@ -0,0 +1,9 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+SECTIONS
+{
+        .payload16 0 : {
+                guest16 = .;
+                guest16.o(.text)
+                guest16_end = .;
+        }
+}
diff --git a/samples/acrn/vm-sample.c b/samples/acrn/vm-sample.c
new file mode 100644
index 000000000..c61e0f914
--- /dev/null
+++ b/samples/acrn/vm-sample.c
@@ -0,0 +1,132 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * A sample program to run a User VM on the ACRN hypervisor
+ *
+ * This sample runs in a Service VM, which is a privileged VM of ACRN.
+ * CONFIG_ACRN_HSM needs to be enabled in the Service VM.
+ *
+ * Guest VM code in guest16.s will be executed after the VM launched.
+ *
+ * Copyright (C) 2020 Intel Corporation. All rights reserved.
+ */
+#include <stdio.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <signal.h>
+#include <sys/ioctl.h>
+#include <linux/acrn.h>
+
+#define GUEST_MEMORY_SIZE	(1024*1024)
+void *guest_memory;
+
+extern const unsigned char guest16[], guest16_end[];
+static char io_request_page[4096] __attribute__((aligned(4096)));
+static struct acrn_io_request *io_req_buf = (struct acrn_io_request *)io_request_page;
+
+__u16 vcpu_num;
+__u16 vmid;
+
+int hsm_fd;
+int is_running = 1;
+
+void vm_exit(int sig)
+{
+	sig = sig;
+
+	is_running = 0;
+	ioctl(hsm_fd, ACRN_IOCTL_PAUSE_VM, vmid);
+	ioctl(hsm_fd, ACRN_IOCTL_DESTROY_IOREQ_CLIENT, 0);
+}
+
+int main(int argc, char **argv)
+{
+	int vcpu_id, ret;
+	struct acrn_vm_creation create_vm = {0};
+	struct acrn_vm_memmap ram_map = {0};
+	struct acrn_vcpu_regs regs;
+	struct acrn_io_request *io_req;
+	struct acrn_ioreq_notify __attribute__((aligned(8))) notify;
+
+	argc = argc;
+	argv = argv;
+
+	ret = posix_memalign(&guest_memory, 4096, GUEST_MEMORY_SIZE);
+	if (ret < 0) {
+		printf("Not enough memory!\n");
+		return -1;
+	}
+	hsm_fd = open("/dev/acrn_hsm", O_RDWR|O_CLOEXEC);
+
+	create_vm.ioreq_buf = (__u64)io_req_buf;
+	ret = ioctl(hsm_fd, ACRN_IOCTL_CREATE_VM, &create_vm);
+	printf("Created VM! [%d]\n", ret);
+	vcpu_num = create_vm.vcpu_num;
+	vmid = create_vm.vmid;
+
+	/* setup guest memory */
+	ram_map.type = ACRN_MEMMAP_RAM;
+	ram_map.vma_base = (__u64)guest_memory;
+	ram_map.len = GUEST_MEMORY_SIZE;
+	ram_map.user_vm_pa = 0;
+	ram_map.attr = ACRN_MEM_ACCESS_RWX;
+	ret = ioctl(hsm_fd, ACRN_IOCTL_SET_MEMSEG, &ram_map);
+	printf("Set up VM memory! [%d]\n", ret);
+
+	memcpy(guest_memory, guest16, guest16_end-guest16);
+
+	/* setup vcpu registers */
+	memset(&regs, 0, sizeof(regs));
+	regs.vcpu_id = 0;
+	regs.vcpu_regs.rip = 0;
+
+	/* CR0_ET | CR0_NE */
+	regs.vcpu_regs.cr0 = 0x30U;
+	regs.vcpu_regs.cs_ar = 0x009FU;
+	regs.vcpu_regs.cs_sel = 0xF000U;
+	regs.vcpu_regs.cs_limit = 0xFFFFU;
+	regs.vcpu_regs.cs_base = 0 & 0xFFFF0000UL;
+	regs.vcpu_regs.rip = 0 & 0xFFFFUL;
+
+	ret = ioctl(hsm_fd, ACRN_IOCTL_SET_VCPU_REGS, &regs);
+	printf("Set up VM BSP registers! [%d]\n", ret);
+
+	/* create an ioreq client for this VM */
+	ret = ioctl(hsm_fd, ACRN_IOCTL_CREATE_IOREQ_CLIENT, 0);
+	printf("Created IO request client! [%d]\n", ret);
+
+	/* run vm */
+	ret = ioctl(hsm_fd, ACRN_IOCTL_START_VM, vmid);
+	printf("Start VM! [%d]\n", ret);
+
+	signal(SIGINT, vm_exit);
+	while (is_running) {
+		ret = ioctl(hsm_fd, ACRN_IOCTL_ATTACH_IOREQ_CLIENT, 0);
+
+		for (vcpu_id = 0; vcpu_id < vcpu_num; vcpu_id++) {
+			io_req = &io_req_buf[vcpu_id];
+			if ((__sync_add_and_fetch(&io_req->processed, 0) == ACRN_IOREQ_STATE_PROCESSING)
+					&& (!io_req->kernel_handled))
+				if (io_req->type == ACRN_IOREQ_TYPE_PORTIO) {
+					int bytes, port, in;
+
+					port = io_req->reqs.pio_request.address;
+					bytes = io_req->reqs.pio_request.size;
+					in = (io_req->reqs.pio_request.direction == ACRN_IOREQ_DIR_READ);
+					printf("Guest VM %s PIO[%x] with size[%x]\n", in ? "read" : "write", port, bytes);
+
+					notify.vmid = vmid;
+					notify.vcpu = vcpu_id;
+					ioctl(hsm_fd, ACRN_IOCTL_NOTIFY_REQUEST_FINISH, &notify);
+				}
+		}
+	}
+
+	ret = ioctl(hsm_fd, ACRN_IOCTL_DESTROY_VM, NULL);
+	printf("Destroy VM! [%d]\n", ret);
+	close(hsm_fd);
+	free(guest_memory);
+	return 0;
+}