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

1 files changed, 270 insertions, 0 deletions

diff --git a/drivers/s390/char/vmcp.c b/drivers/s390/char/vmcp.c
new file mode 100644
index 000000000..b277f9d0c
--- /dev/null
+++ b/drivers/s390/char/vmcp.c
@@ -0,0 +1,270 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright IBM Corp. 2004, 2010
+ * Interface implementation for communication with the z/VM control program
+ *
+ * Author(s): Christian Borntraeger <borntraeger@de.ibm.com>
+ *
+ * z/VMs CP offers the possibility to issue commands via the diagnose code 8
+ * this driver implements a character device that issues these commands and
+ * returns the answer of CP.
+ *
+ * The idea of this driver is based on cpint from Neale Ferguson and #CP in CMS
+ */
+
+#include <linux/fs.h>
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/miscdevice.h>
+#include <linux/slab.h>
+#include <linux/uaccess.h>
+#include <linux/mutex.h>
+#include <linux/cma.h>
+#include <linux/mm.h>
+#include <asm/machine.h>
+#include <asm/cpcmd.h>
+#include <asm/debug.h>
+#include <asm/vmcp.h>
+
+struct vmcp_session {
+	char *response;
+	unsigned int bufsize;
+	unsigned int cma_alloc : 1;
+	int resp_size;
+	int resp_code;
+	struct mutex mutex;
+};
+
+static debug_info_t *vmcp_debug;
+
+static unsigned long vmcp_cma_size __initdata = CONFIG_VMCP_CMA_SIZE * 1024 * 1024;
+static struct cma *vmcp_cma;
+
+static int __init early_parse_vmcp_cma(char *p)
+{
+	if (!p)
+		return 1;
+	vmcp_cma_size = ALIGN(memparse(p, NULL), PAGE_SIZE);
+	return 0;
+}
+early_param("vmcp_cma", early_parse_vmcp_cma);
+
+void __init vmcp_cma_reserve(void)
+{
+	if (!machine_is_vm())
+		return;
+	cma_declare_contiguous(0, vmcp_cma_size, 0, 0, 0, false, "vmcp", &vmcp_cma);
+}
+
+static void vmcp_response_alloc(struct vmcp_session *session)
+{
+	struct page *page = NULL;
+	int nr_pages, order;
+
+	order = get_order(session->bufsize);
+	nr_pages = ALIGN(session->bufsize, PAGE_SIZE) >> PAGE_SHIFT;
+	/*
+	 * For anything below order 3 allocations rely on the buddy
+	 * allocator. If such low-order allocations can't be handled
+	 * anymore the system won't work anyway.
+	 */
+	if (order > 2)
+		page = cma_alloc(vmcp_cma, nr_pages, 0, false);
+	if (page) {
+		session->response = (char *)page_to_virt(page);
+		session->cma_alloc = 1;
+		return;
+	}
+	session->response = (char *)__get_free_pages(GFP_KERNEL | __GFP_RETRY_MAYFAIL, order);
+}
+
+static void vmcp_response_free(struct vmcp_session *session)
+{
+	int nr_pages, order;
+	struct page *page;
+
+	if (!session->response)
+		return;
+	order = get_order(session->bufsize);
+	nr_pages = ALIGN(session->bufsize, PAGE_SIZE) >> PAGE_SHIFT;
+	if (session->cma_alloc) {
+		page = virt_to_page(session->response);
+		cma_release(vmcp_cma, page, nr_pages);
+		session->cma_alloc = 0;
+	} else {
+		free_pages((unsigned long)session->response, order);
+	}
+	session->response = NULL;
+}
+
+static int vmcp_open(struct inode *inode, struct file *file)
+{
+	struct vmcp_session *session;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	session = kmalloc_obj(*session);
+	if (!session)
+		return -ENOMEM;
+
+	session->bufsize = PAGE_SIZE;
+	session->response = NULL;
+	session->resp_size = 0;
+	mutex_init(&session->mutex);
+	file->private_data = session;
+	return nonseekable_open(inode, file);
+}
+
+static int vmcp_release(struct inode *inode, struct file *file)
+{
+	struct vmcp_session *session;
+
+	session = file->private_data;
+	file->private_data = NULL;
+	vmcp_response_free(session);
+	kfree(session);
+	return 0;
+}
+
+static ssize_t
+vmcp_read(struct file *file, char __user *buff, size_t count, loff_t *ppos)
+{
+	ssize_t ret;
+	size_t size;
+	struct vmcp_session *session;
+
+	session = file->private_data;
+	if (mutex_lock_interruptible(&session->mutex))
+		return -ERESTARTSYS;
+	if (!session->response) {
+		mutex_unlock(&session->mutex);
+		return 0;
+	}
+	size = min_t(size_t, session->resp_size, session->bufsize);
+	ret = simple_read_from_buffer(buff, count, ppos,
+					session->response, size);
+
+	mutex_unlock(&session->mutex);
+
+	return ret;
+}
+
+static ssize_t
+vmcp_write(struct file *file, const char __user *buff, size_t count,
+	   loff_t *ppos)
+{
+	char *cmd;
+	struct vmcp_session *session;
+
+	if (count > 240)
+		return -EINVAL;
+	cmd = memdup_user_nul(buff, count);
+	if (IS_ERR(cmd))
+		return PTR_ERR(cmd);
+	session = file->private_data;
+	if (mutex_lock_interruptible(&session->mutex)) {
+		kfree(cmd);
+		return -ERESTARTSYS;
+	}
+	if (!session->response)
+		vmcp_response_alloc(session);
+	if (!session->response) {
+		mutex_unlock(&session->mutex);
+		kfree(cmd);
+		return -ENOMEM;
+	}
+	debug_text_event(vmcp_debug, 1, cmd);
+	session->resp_size = cpcmd(cmd, session->response, session->bufsize,
+				   &session->resp_code);
+	mutex_unlock(&session->mutex);
+	kfree(cmd);
+	*ppos = 0;		/* reset the file pointer after a command */
+	return count;
+}
+
+
+/*
+ * These ioctls are available, as the semantics of the diagnose 8 call
+ * does not fit very well into a Linux call. Diagnose X'08' is described in
+ * CP Programming Services SC24-6084-00
+ *
+ * VMCP_GETCODE: gives the CP return code back to user space
+ * VMCP_SETBUF: sets the response buffer for the next write call. diagnose 8
+ * expects adjacent pages in real storage and to make matters worse, we
+ * dont know the size of the response. Therefore we default to PAGESIZE and
+ * let userspace to change the response size, if userspace expects a bigger
+ * response
+ */
+static long vmcp_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
+{
+	struct vmcp_session *session;
+	int ret = -ENOTTY;
+	int __user *argp;
+
+	session = file->private_data;
+	argp = (int __user *)arg;
+	if (mutex_lock_interruptible(&session->mutex))
+		return -ERESTARTSYS;
+	switch (cmd) {
+	case VMCP_GETCODE:
+		ret = put_user(session->resp_code, argp);
+		break;
+	case VMCP_SETBUF:
+		vmcp_response_free(session);
+		ret = get_user(session->bufsize, argp);
+		if (ret)
+			session->bufsize = PAGE_SIZE;
+		if (!session->bufsize || get_order(session->bufsize) > 8) {
+			session->bufsize = PAGE_SIZE;
+			ret = -EINVAL;
+		}
+		break;
+	case VMCP_GETSIZE:
+		ret = put_user(session->resp_size, argp);
+		break;
+	default:
+		break;
+	}
+	mutex_unlock(&session->mutex);
+	return ret;
+}
+
+static const struct file_operations vmcp_fops = {
+	.owner		= THIS_MODULE,
+	.open		= vmcp_open,
+	.release	= vmcp_release,
+	.read		= vmcp_read,
+	.write		= vmcp_write,
+	.unlocked_ioctl	= vmcp_ioctl,
+};
+
+static struct miscdevice vmcp_dev = {
+	.name	= "vmcp",
+	.minor	= MISC_DYNAMIC_MINOR,
+	.fops	= &vmcp_fops,
+};
+
+static int __init vmcp_init(void)
+{
+	int ret;
+
+	if (!machine_is_vm())
+		return 0;
+
+	vmcp_debug = debug_register("vmcp", 1, 1, 240);
+	if (!vmcp_debug)
+		return -ENOMEM;
+
+	ret = debug_register_view(vmcp_debug, &debug_hex_ascii_view);
+	if (ret) {
+		debug_unregister(vmcp_debug);
+		return ret;
+	}
+
+	ret = misc_register(&vmcp_dev);
+	if (ret)
+		debug_unregister(vmcp_debug);
+	return ret;
+}
+device_initcall(vmcp_init);