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, 164 insertions, 0 deletions

diff --git a/drivers/net/pse-pd/realtek-pse-mcu-uart.c b/drivers/net/pse-pd/realtek-pse-mcu-uart.c
new file mode 100644
index 000000000..9baa17d8d
--- /dev/null
+++ b/drivers/net/pse-pd/realtek-pse-mcu-uart.c
@@ -0,0 +1,164 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include <linux/cleanup.h>
+#include <linux/completion.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/pse-pd/pse.h>
+#include <linux/serdev.h>
+#include <linux/spinlock.h>
+#include <linux/string.h>
+
+#include "realtek-pse-mcu.h"
+
+#define RTPSE_MCU_UART_BAUD_DEFAULT	19200
+#define RTPSE_MCU_UART_TX_TIMEOUT	msecs_to_jiffies(100)
+#define RTPSE_MCU_UART_RX_TIMEOUT	msecs_to_jiffies(RTPSE_MCU_RESPONSE_MAX_MS)
+
+struct rtpse_mcu_uart {
+	struct rtpse_mcu_ctrl pse;
+	struct serdev_device *serdev;
+	struct completion rx_done;
+	spinlock_t rx_lock;		/* protects rx_buf and rx_len */
+	size_t rx_len;
+	u8 rx_buf[RTPSE_MCU_MSG_SIZE];
+};
+
+#define to_rtpse_mcu_uart(p)  container_of(p, struct rtpse_mcu_uart, pse)
+
+/*
+ * No framing is done here: a glitched frame costs one transaction, then
+ * the next _send re-frames from rx_len 0. Resync works by returning count
+ * (not take), dropping any overflow so serdev keeps no leftover to bleed
+ * into the next frame.
+ */
+static size_t rtpse_mcu_uart_receive(struct serdev_device *serdev,
+				     const u8 *buf, size_t count)
+{
+	struct rtpse_mcu_uart *ctx = serdev_device_get_drvdata(serdev);
+	size_t take;
+
+	scoped_guard(spinlock_irqsave, &ctx->rx_lock) {
+		take = min(count, sizeof(ctx->rx_buf) - ctx->rx_len);
+		if (take) {
+			memcpy(ctx->rx_buf + ctx->rx_len, buf, take);
+			ctx->rx_len += take;
+			if (ctx->rx_len == sizeof(ctx->rx_buf))
+				complete(&ctx->rx_done);
+		}
+	}
+
+	/* consume all to avoid desync/misalignment */
+	return count;
+}
+
+static const struct serdev_device_ops rtpse_mcu_uart_serdev_ops = {
+	.receive_buf = rtpse_mcu_uart_receive,
+	.write_wakeup = serdev_device_write_wakeup,
+};
+
+static int rtpse_mcu_uart_send(struct rtpse_mcu_ctrl *pse, const struct rtpse_mcu_msg *req)
+{
+	struct rtpse_mcu_uart *ctx = to_rtpse_mcu_uart(pse);
+	int written;
+
+	/* clear any leftover rx state before transmitting */
+	scoped_guard(spinlock_irqsave, &ctx->rx_lock) {
+		reinit_completion(&ctx->rx_done);
+		ctx->rx_len = 0;
+	}
+
+	written = serdev_device_write(ctx->serdev, (const u8 *)req, sizeof(*req),
+				      RTPSE_MCU_UART_TX_TIMEOUT);
+	if (written < 0)
+		return written;
+	if (written != sizeof(*req))
+		return -EIO;
+
+	return 0;
+}
+
+static int rtpse_mcu_uart_recv(struct rtpse_mcu_ctrl *pse,
+			       const struct rtpse_mcu_msg *req,
+			       struct rtpse_mcu_msg *resp)
+{
+	struct rtpse_mcu_uart *ctx = to_rtpse_mcu_uart(pse);
+
+	if (!wait_for_completion_timeout(&ctx->rx_done, RTPSE_MCU_UART_RX_TIMEOUT))
+		return -ETIMEDOUT;
+
+	scoped_guard(spinlock_irqsave, &ctx->rx_lock) {
+		if (ctx->rx_len != sizeof(*resp))
+			return -EIO;
+
+		memcpy(resp, ctx->rx_buf, sizeof(*resp));
+	}
+	return 0;
+}
+
+static const struct rtpse_mcu_transport_ops rtpse_mcu_uart_transport_ops = {
+	.send = rtpse_mcu_uart_send,
+	.recv = rtpse_mcu_uart_recv,
+};
+
+static int rtpse_mcu_uart_probe(struct serdev_device *serdev)
+{
+	u32 speed = RTPSE_MCU_UART_BAUD_DEFAULT;
+	struct device *dev = &serdev->dev;
+	struct rtpse_mcu_uart *ctx;
+	unsigned int baud;
+	int ret;
+
+	ctx = devm_kzalloc(dev, sizeof(*ctx), GFP_KERNEL);
+	if (!ctx)
+		return -ENOMEM;
+
+	ctx->serdev = serdev;
+	ctx->pse.dev = dev;
+	ctx->pse.pcdev.owner = THIS_MODULE;
+	ctx->pse.transport = &rtpse_mcu_uart_transport_ops;
+	init_completion(&ctx->rx_done);
+	spin_lock_init(&ctx->rx_lock);
+
+	serdev_device_set_drvdata(serdev, ctx);
+	serdev_device_set_client_ops(serdev, &rtpse_mcu_uart_serdev_ops);
+
+	ret = devm_serdev_device_open(dev, serdev);
+	if (ret)
+		return dev_err_probe(dev, ret, "failed to open serdev\n");
+
+	fwnode_property_read_u32(dev_fwnode(dev), "current-speed", &speed);
+
+	baud = serdev_device_set_baudrate(serdev, speed);
+	if (baud != speed)
+		dev_warn(dev, "could not set baudrate %u, controller uses %u\n",
+			 speed, baud);
+
+	serdev_device_set_flow_control(serdev, false);
+
+	ret = serdev_device_set_parity(serdev, SERDEV_PARITY_NONE);
+	if (ret)
+		dev_warn(dev, "could not set parity to none: %d\n", ret);
+
+	return rtpse_mcu_register(&ctx->pse);
+}
+
+static const struct of_device_id rtpse_mcu_uart_of_match[] = {
+	{ .compatible = "realtek,pse-mcu-gen1", .data = &rtpse_mcu_gen1_data },
+	{ .compatible = "realtek,pse-mcu-gen2", .data = &rtpse_mcu_gen2_data },
+	{ /* sentinel */ }
+};
+MODULE_DEVICE_TABLE(of, rtpse_mcu_uart_of_match);
+
+static struct serdev_device_driver rtpse_mcu_uart_driver = {
+	.driver = {
+		.name = "realtek-pse-mcu-uart",
+		.of_match_table = rtpse_mcu_uart_of_match,
+	},
+	.probe  = rtpse_mcu_uart_probe,
+};
+module_serdev_device_driver(rtpse_mcu_uart_driver);
+
+MODULE_AUTHOR("Jonas Jelonek <jelonek.jonas@gmail.com>");
+MODULE_DESCRIPTION("Realtek PSE MCU driver (UART transport)");
+MODULE_LICENSE("GPL");