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, 653 insertions, 0 deletions

diff --git a/drivers/tty/serial/sunhv.c b/drivers/tty/serial/sunhv.c
new file mode 100644
index 000000000..35301d6d5
--- /dev/null
+++ b/drivers/tty/serial/sunhv.c
@@ -0,0 +1,653 @@
+// SPDX-License-Identifier: GPL-2.0
+/* sunhv.c: Serial driver for SUN4V hypervisor console.
+ *
+ * Copyright (C) 2006, 2007 David S. Miller (davem@davemloft.net)
+ */
+
+#include <linux/kernel.h>
+#include <linux/errno.h>
+#include <linux/tty.h>
+#include <linux/tty_flip.h>
+#include <linux/major.h>
+#include <linux/circ_buf.h>
+#include <linux/serial.h>
+#include <linux/sysrq.h>
+#include <linux/console.h>
+#include <linux/spinlock.h>
+#include <linux/slab.h>
+#include <linux/delay.h>
+#include <linux/init.h>
+#include <linux/of.h>
+#include <linux/platform_device.h>
+
+#include <asm/hypervisor.h>
+#include <asm/spitfire.h>
+#include <asm/irq.h>
+#include <asm/setup.h>
+
+#include <linux/serial_core.h>
+#include <linux/sunserialcore.h>
+
+#define CON_BREAK	((long)-1)
+#define CON_HUP		((long)-2)
+
+#define IGNORE_BREAK	0x1
+#define IGNORE_ALL	0x2
+
+static char *con_write_page;
+static char *con_read_page;
+
+static int hung_up = 0;
+
+static void transmit_chars_putchar(struct uart_port *port,
+		struct tty_port *tport)
+{
+	unsigned char ch;
+
+	while (kfifo_peek(&tport->xmit_fifo, &ch)) {
+		long status = sun4v_con_putchar(ch);
+
+		if (status != HV_EOK)
+			break;
+
+		uart_xmit_advance(port, 1);
+	}
+}
+
+static void transmit_chars_write(struct uart_port *port, struct tty_port *tport)
+{
+	while (!kfifo_is_empty(&tport->xmit_fifo)) {
+		unsigned long len, ra, status, sent;
+		unsigned char *tail;
+
+		len = kfifo_out_linear_ptr(&tport->xmit_fifo, &tail,
+				UART_XMIT_SIZE);
+		ra = __pa(tail);
+
+		status = sun4v_con_write(ra, len, &sent);
+		if (status != HV_EOK)
+			break;
+		uart_xmit_advance(port, sent);
+	}
+}
+
+static int receive_chars_getchar(struct uart_port *port)
+{
+	int saw_console_brk = 0;
+	int limit = 10000;
+
+	while (limit-- > 0) {
+		long status;
+		long c = sun4v_con_getchar(&status);
+
+		if (status == HV_EWOULDBLOCK)
+			break;
+
+		if (c == CON_BREAK) {
+			if (uart_handle_break(port))
+				continue;
+			saw_console_brk = 1;
+			c = 0;
+		}
+
+		if (c == CON_HUP) {
+			hung_up = 1;
+			uart_handle_dcd_change(port, false);
+		} else if (hung_up) {
+			hung_up = 0;
+			uart_handle_dcd_change(port, true);
+		}
+
+		if (port->state == NULL) {
+			uart_handle_sysrq_char(port, c);
+			continue;
+		}
+
+		port->icount.rx++;
+
+		if (uart_handle_sysrq_char(port, c))
+			continue;
+
+		tty_insert_flip_char(&port->state->port, c, TTY_NORMAL);
+	}
+
+	return saw_console_brk;
+}
+
+static int receive_chars_read(struct uart_port *port)
+{
+	static int saw_console_brk;
+	int limit = 10000;
+
+	while (limit-- > 0) {
+		unsigned long ra = __pa(con_read_page);
+		unsigned long bytes_read, i;
+		long stat = sun4v_con_read(ra, PAGE_SIZE, &bytes_read);
+
+		if (stat != HV_EOK) {
+			bytes_read = 0;
+
+			if (stat == CON_BREAK) {
+				if (saw_console_brk)
+					sun_do_break();
+
+				if (uart_handle_break(port))
+					continue;
+				saw_console_brk = 1;
+				*con_read_page = 0;
+				bytes_read = 1;
+			} else if (stat == CON_HUP) {
+				hung_up = 1;
+				uart_handle_dcd_change(port, false);
+				continue;
+			} else {
+				/* HV_EWOULDBLOCK, etc.  */
+				break;
+			}
+		}
+
+		if (hung_up) {
+			hung_up = 0;
+			uart_handle_dcd_change(port, true);
+		}
+
+		if (port->sysrq != 0 &&  *con_read_page) {
+			for (i = 0; i < bytes_read; i++)
+				uart_handle_sysrq_char(port, con_read_page[i]);
+			saw_console_brk = 0;
+		}
+
+		if (port->state == NULL)
+			continue;
+
+		port->icount.rx += bytes_read;
+
+		tty_insert_flip_string(&port->state->port, con_read_page,
+				bytes_read);
+	}
+
+	return saw_console_brk;
+}
+
+struct sunhv_ops {
+	void (*transmit_chars)(struct uart_port *port, struct tty_port *tport);
+	int (*receive_chars)(struct uart_port *port);
+};
+
+static const struct sunhv_ops bychar_ops = {
+	.transmit_chars = transmit_chars_putchar,
+	.receive_chars = receive_chars_getchar,
+};
+
+static const struct sunhv_ops bywrite_ops = {
+	.transmit_chars = transmit_chars_write,
+	.receive_chars = receive_chars_read,
+};
+
+static const struct sunhv_ops *sunhv_ops = &bychar_ops;
+
+static struct tty_port *receive_chars(struct uart_port *port)
+{
+	struct tty_port *tport = NULL;
+
+	if (port->state != NULL)		/* Unopened serial console */
+		tport = &port->state->port;
+
+	if (sunhv_ops->receive_chars(port))
+		sun_do_break();
+
+	return tport;
+}
+
+static void transmit_chars(struct uart_port *port)
+{
+	struct tty_port *tport;
+
+	if (!port->state)
+		return;
+
+	tport = &port->state->port;
+	if (kfifo_is_empty(&tport->xmit_fifo) || uart_tx_stopped(port))
+		return;
+
+	sunhv_ops->transmit_chars(port, tport);
+
+	if (kfifo_len(&tport->xmit_fifo) < WAKEUP_CHARS)
+		uart_write_wakeup(port);
+}
+
+static irqreturn_t sunhv_interrupt(int irq, void *dev_id)
+{
+	struct uart_port *port = dev_id;
+	struct tty_port *tport;
+	unsigned long flags;
+
+	uart_port_lock_irqsave(port, &flags);
+	tport = receive_chars(port);
+	transmit_chars(port);
+	uart_port_unlock_irqrestore(port, flags);
+
+	if (tport)
+		tty_flip_buffer_push(tport);
+
+	return IRQ_HANDLED;
+}
+
+/* port->lock is not held.  */
+static unsigned int sunhv_tx_empty(struct uart_port *port)
+{
+	/* Transmitter is always empty for us.  If the circ buffer
+	 * is non-empty or there is an x_char pending, our caller
+	 * will do the right thing and ignore what we return here.
+	 */
+	return TIOCSER_TEMT;
+}
+
+/* port->lock held by caller.  */
+static void sunhv_set_mctrl(struct uart_port *port, unsigned int mctrl)
+{
+	return;
+}
+
+/* port->lock is held by caller and interrupts are disabled.  */
+static unsigned int sunhv_get_mctrl(struct uart_port *port)
+{
+	return TIOCM_DSR | TIOCM_CAR | TIOCM_CTS;
+}
+
+/* port->lock held by caller.  */
+static void sunhv_stop_tx(struct uart_port *port)
+{
+	return;
+}
+
+/* port->lock held by caller.  */
+static void sunhv_start_tx(struct uart_port *port)
+{
+	transmit_chars(port);
+}
+
+/* port->lock is not held.  */
+static void sunhv_send_xchar(struct uart_port *port, char ch)
+{
+	unsigned long flags;
+	int limit = 10000;
+
+	if (ch == __DISABLED_CHAR)
+		return;
+
+	uart_port_lock_irqsave(port, &flags);
+
+	while (limit-- > 0) {
+		long status = sun4v_con_putchar(ch);
+		if (status == HV_EOK)
+			break;
+		udelay(1);
+	}
+
+	uart_port_unlock_irqrestore(port, flags);
+}
+
+/* port->lock held by caller.  */
+static void sunhv_stop_rx(struct uart_port *port)
+{
+}
+
+/* port->lock is not held.  */
+static void sunhv_break_ctl(struct uart_port *port, int break_state)
+{
+	if (break_state) {
+		unsigned long flags;
+		int limit = 10000;
+
+		uart_port_lock_irqsave(port, &flags);
+
+		while (limit-- > 0) {
+			long status = sun4v_con_putchar(CON_BREAK);
+			if (status == HV_EOK)
+				break;
+			udelay(1);
+		}
+
+		uart_port_unlock_irqrestore(port, flags);
+	}
+}
+
+/* port->lock is not held.  */
+static int sunhv_startup(struct uart_port *port)
+{
+	return 0;
+}
+
+/* port->lock is not held.  */
+static void sunhv_shutdown(struct uart_port *port)
+{
+}
+
+/* port->lock is not held.  */
+static void sunhv_set_termios(struct uart_port *port, struct ktermios *termios,
+			      const struct ktermios *old)
+{
+	unsigned int baud = uart_get_baud_rate(port, termios, old, 0, 4000000);
+	unsigned int quot = uart_get_divisor(port, baud);
+	unsigned int iflag, cflag;
+	unsigned long flags;
+
+	uart_port_lock_irqsave(port, &flags);
+
+	iflag = termios->c_iflag;
+	cflag = termios->c_cflag;
+
+	port->ignore_status_mask = 0;
+	if (iflag & IGNBRK)
+		port->ignore_status_mask |= IGNORE_BREAK;
+	if ((cflag & CREAD) == 0)
+		port->ignore_status_mask |= IGNORE_ALL;
+
+	/* XXX */
+	uart_update_timeout(port, cflag,
+			    (port->uartclk / (16 * quot)));
+
+	uart_port_unlock_irqrestore(port, flags);
+}
+
+static const char *sunhv_type(struct uart_port *port)
+{
+	return "SUN4V HCONS";
+}
+
+static void sunhv_release_port(struct uart_port *port)
+{
+}
+
+static int sunhv_request_port(struct uart_port *port)
+{
+	return 0;
+}
+
+static void sunhv_config_port(struct uart_port *port, int flags)
+{
+}
+
+static int sunhv_verify_port(struct uart_port *port, struct serial_struct *ser)
+{
+	return -EINVAL;
+}
+
+static const struct uart_ops sunhv_pops = {
+	.tx_empty	= sunhv_tx_empty,
+	.set_mctrl	= sunhv_set_mctrl,
+	.get_mctrl	= sunhv_get_mctrl,
+	.stop_tx	= sunhv_stop_tx,
+	.start_tx	= sunhv_start_tx,
+	.send_xchar	= sunhv_send_xchar,
+	.stop_rx	= sunhv_stop_rx,
+	.break_ctl	= sunhv_break_ctl,
+	.startup	= sunhv_startup,
+	.shutdown	= sunhv_shutdown,
+	.set_termios	= sunhv_set_termios,
+	.type		= sunhv_type,
+	.release_port	= sunhv_release_port,
+	.request_port	= sunhv_request_port,
+	.config_port	= sunhv_config_port,
+	.verify_port	= sunhv_verify_port,
+};
+
+static struct uart_driver sunhv_reg = {
+	.owner			= THIS_MODULE,
+	.driver_name		= "sunhv",
+	.dev_name		= "ttyHV",
+	.major			= TTY_MAJOR,
+};
+
+static struct uart_port *sunhv_port;
+
+void sunhv_migrate_hvcons_irq(int cpu)
+{
+	/* Migrate hvcons irq to param cpu */
+	irq_force_affinity(sunhv_port->irq, cpumask_of(cpu));
+}
+
+/* Copy 's' into the con_write_page, decoding "\n" into
+ * "\r\n" along the way.  We have to return two lengths
+ * because the caller needs to know how much to advance
+ * 's' and also how many bytes to output via con_write_page.
+ */
+static int fill_con_write_page(const char *s, unsigned int n,
+			       unsigned long *page_bytes)
+{
+	const char *orig_s = s;
+	char *p = con_write_page;
+	int left = PAGE_SIZE;
+
+	while (n--) {
+		if (*s == '\n') {
+			if (left < 2)
+				break;
+			*p++ = '\r';
+			left--;
+		} else if (left < 1)
+			break;
+		*p++ = *s++;
+		left--;
+	}
+	*page_bytes = p - con_write_page;
+	return s - orig_s;
+}
+
+static void sunhv_console_write_paged(struct console *con, const char *s, unsigned n)
+{
+	struct uart_port *port = sunhv_port;
+	unsigned long flags;
+	int locked = 1;
+
+	if (port->sysrq || oops_in_progress)
+		locked = uart_port_trylock_irqsave(port, &flags);
+	else
+		uart_port_lock_irqsave(port, &flags);
+
+	while (n > 0) {
+		unsigned long ra = __pa(con_write_page);
+		unsigned long page_bytes;
+		unsigned int cpy = fill_con_write_page(s, n,
+						       &page_bytes);
+
+		n -= cpy;
+		s += cpy;
+		while (page_bytes > 0) {
+			unsigned long written;
+			int limit = 1000000;
+
+			while (limit--) {
+				unsigned long stat;
+
+				stat = sun4v_con_write(ra, page_bytes,
+						       &written);
+				if (stat == HV_EOK)
+					break;
+				udelay(1);
+			}
+			if (limit < 0)
+				break;
+			page_bytes -= written;
+			ra += written;
+		}
+	}
+
+	if (locked)
+		uart_port_unlock_irqrestore(port, flags);
+}
+
+static inline void sunhv_console_putchar(struct uart_port *port, char c)
+{
+	int limit = 1000000;
+
+	while (limit-- > 0) {
+		long status = sun4v_con_putchar(c);
+		if (status == HV_EOK)
+			break;
+		udelay(1);
+	}
+}
+
+static void sunhv_console_write_bychar(struct console *con, const char *s, unsigned n)
+{
+	struct uart_port *port = sunhv_port;
+	unsigned long flags;
+	int i, locked = 1;
+
+	if (port->sysrq || oops_in_progress)
+		locked = uart_port_trylock_irqsave(port, &flags);
+	else
+		uart_port_lock_irqsave(port, &flags);
+
+	for (i = 0; i < n; i++) {
+		if (*s == '\n')
+			sunhv_console_putchar(port, '\r');
+		sunhv_console_putchar(port, *s++);
+	}
+
+	if (locked)
+		uart_port_unlock_irqrestore(port, flags);
+}
+
+static struct console sunhv_console = {
+	.name	=	"ttyHV",
+	.write	=	sunhv_console_write_bychar,
+	.device	=	uart_console_device,
+	.flags	=	CON_PRINTBUFFER,
+	.index	=	-1,
+	.data	=	&sunhv_reg,
+};
+
+static int hv_probe(struct platform_device *op)
+{
+	struct uart_port *port;
+	unsigned long minor;
+	int err;
+
+	if (op->archdata.irqs[0] == 0xffffffff)
+		return -ENODEV;
+
+	port = kzalloc_obj(struct uart_port);
+	if (unlikely(!port))
+		return -ENOMEM;
+
+	minor = 1;
+	if (sun4v_hvapi_register(HV_GRP_CORE, 1, &minor) == 0 &&
+	    minor >= 1) {
+		err = -ENOMEM;
+		con_write_page = kzalloc(PAGE_SIZE, GFP_KERNEL);
+		if (!con_write_page)
+			goto out_free_port;
+
+		con_read_page = kzalloc(PAGE_SIZE, GFP_KERNEL);
+		if (!con_read_page)
+			goto out_free_con_write_page;
+
+		sunhv_console.write = sunhv_console_write_paged;
+		sunhv_ops = &bywrite_ops;
+	}
+
+	sunhv_port = port;
+
+	port->has_sysrq = 1;
+	port->line = 0;
+	port->ops = &sunhv_pops;
+	port->type = PORT_SUNHV;
+	port->uartclk = ( 29491200 / 16 ); /* arbitrary */
+
+	port->membase = (unsigned char __iomem *) __pa(port);
+
+	port->irq = op->archdata.irqs[0];
+
+	port->dev = &op->dev;
+
+	err = sunserial_register_minors(&sunhv_reg, 1);
+	if (err)
+		goto out_free_con_read_page;
+
+	sunserial_console_match(&sunhv_console, op->dev.of_node,
+				&sunhv_reg, port->line, false);
+
+	err = uart_add_one_port(&sunhv_reg, port);
+	if (err)
+		goto out_unregister_driver;
+
+	err = request_irq(port->irq, sunhv_interrupt, 0, "hvcons", port);
+	if (err)
+		goto out_remove_port;
+
+	platform_set_drvdata(op, port);
+
+	return 0;
+
+out_remove_port:
+	uart_remove_one_port(&sunhv_reg, port);
+
+out_unregister_driver:
+	sunserial_unregister_minors(&sunhv_reg, 1);
+
+out_free_con_read_page:
+	kfree(con_read_page);
+
+out_free_con_write_page:
+	kfree(con_write_page);
+
+out_free_port:
+	kfree(port);
+	sunhv_port = NULL;
+	return err;
+}
+
+static void hv_remove(struct platform_device *dev)
+{
+	struct uart_port *port = platform_get_drvdata(dev);
+
+	free_irq(port->irq, port);
+
+	uart_remove_one_port(&sunhv_reg, port);
+
+	sunserial_unregister_minors(&sunhv_reg, 1);
+	kfree(con_read_page);
+	kfree(con_write_page);
+	kfree(port);
+	sunhv_port = NULL;
+}
+
+static const struct of_device_id hv_match[] = {
+	{
+		.name = "console",
+		.compatible = "qcn",
+	},
+	{
+		.name = "console",
+		.compatible = "SUNW,sun4v-console",
+	},
+	{},
+};
+
+static struct platform_driver hv_driver = {
+	.driver = {
+		.name = "hv",
+		.of_match_table = hv_match,
+	},
+	.probe		= hv_probe,
+	.remove		= hv_remove,
+};
+
+static int __init sunhv_init(void)
+{
+	if (tlb_type != hypervisor)
+		return -ENODEV;
+
+	return platform_driver_register(&hv_driver);
+}
+device_initcall(sunhv_init);
+
+#if 0 /* ...def MODULE ; never supported as such */
+MODULE_AUTHOR("David S. Miller");
+MODULE_DESCRIPTION("SUN4V Hypervisor console driver");
+MODULE_VERSION("2.0");
+MODULE_LICENSE("GPL");
+#endif