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, 315 insertions, 0 deletions

diff --git a/drivers/mailbox/sun6i-msgbox.c b/drivers/mailbox/sun6i-msgbox.c
new file mode 100644
index 000000000..1cbc4d478
--- /dev/null
+++ b/drivers/mailbox/sun6i-msgbox.c
@@ -0,0 +1,315 @@
+// SPDX-License-Identifier: GPL-2.0
+//
+// Copyright (c) 2017-2019 Samuel Holland <samuel@sholland.org>
+
+#include <linux/bitops.h>
+#include <linux/clk.h>
+#include <linux/device.h>
+#include <linux/err.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/kernel.h>
+#include <linux/mailbox_controller.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/of_irq.h>
+#include <linux/platform_device.h>
+#include <linux/reset.h>
+#include <linux/spinlock.h>
+
+#define NUM_CHANS		8
+
+#define CTRL_REG(n)		(0x0000 + 0x4 * ((n) / 4))
+#define CTRL_RX(n)		BIT(0 + 8 * ((n) % 4))
+#define CTRL_TX(n)		BIT(4 + 8 * ((n) % 4))
+
+#define REMOTE_IRQ_EN_REG	0x0040
+#define REMOTE_IRQ_STAT_REG	0x0050
+#define LOCAL_IRQ_EN_REG	0x0060
+#define LOCAL_IRQ_STAT_REG	0x0070
+
+#define RX_IRQ(n)		BIT(0 + 2 * (n))
+#define RX_IRQ_MASK		0x5555
+#define TX_IRQ(n)		BIT(1 + 2 * (n))
+#define TX_IRQ_MASK		0xaaaa
+
+#define FIFO_STAT_REG(n)	(0x0100 + 0x4 * (n))
+#define FIFO_STAT_MASK		GENMASK(0, 0)
+
+#define MSG_STAT_REG(n)		(0x0140 + 0x4 * (n))
+#define MSG_STAT_MASK		GENMASK(2, 0)
+
+#define MSG_DATA_REG(n)		(0x0180 + 0x4 * (n))
+
+#define mbox_dbg(mbox, ...)	dev_dbg((mbox)->controller.dev, __VA_ARGS__)
+
+struct sun6i_msgbox {
+	struct mbox_controller controller;
+	struct clk *clk;
+	spinlock_t lock;
+	void __iomem *regs;
+};
+
+static bool sun6i_msgbox_last_tx_done(struct mbox_chan *chan);
+static bool sun6i_msgbox_peek_data(struct mbox_chan *chan);
+
+static inline int channel_number(struct mbox_chan *chan)
+{
+	return chan - chan->mbox->chans;
+}
+
+static inline struct sun6i_msgbox *to_sun6i_msgbox(struct mbox_chan *chan)
+{
+	return chan->con_priv;
+}
+
+static irqreturn_t sun6i_msgbox_irq(int irq, void *dev_id)
+{
+	struct sun6i_msgbox *mbox = dev_id;
+	uint32_t status;
+	int n;
+
+	/* Only examine channels that are currently enabled. */
+	status = readl(mbox->regs + LOCAL_IRQ_EN_REG) &
+		 readl(mbox->regs + LOCAL_IRQ_STAT_REG);
+
+	if (!(status & RX_IRQ_MASK))
+		return IRQ_NONE;
+
+	for (n = 0; n < NUM_CHANS; ++n) {
+		struct mbox_chan *chan = &mbox->controller.chans[n];
+
+		if (!(status & RX_IRQ(n)))
+			continue;
+
+		while (sun6i_msgbox_peek_data(chan)) {
+			uint32_t msg = readl(mbox->regs + MSG_DATA_REG(n));
+
+			mbox_dbg(mbox, "Channel %d received 0x%08x\n", n, msg);
+			mbox_chan_received_data(chan, &msg);
+		}
+
+		/* The IRQ can be cleared only once the FIFO is empty. */
+		writel(RX_IRQ(n), mbox->regs + LOCAL_IRQ_STAT_REG);
+	}
+
+	return IRQ_HANDLED;
+}
+
+static int sun6i_msgbox_send_data(struct mbox_chan *chan, void *data)
+{
+	struct sun6i_msgbox *mbox = to_sun6i_msgbox(chan);
+	int n = channel_number(chan);
+	uint32_t msg = *(uint32_t *)data;
+
+	/* Using a channel backwards gets the hardware into a bad state. */
+	if (WARN_ON_ONCE(!(readl(mbox->regs + CTRL_REG(n)) & CTRL_TX(n))))
+		return 0;
+
+	writel(msg, mbox->regs + MSG_DATA_REG(n));
+	mbox_dbg(mbox, "Channel %d sent 0x%08x\n", n, msg);
+
+	return 0;
+}
+
+static int sun6i_msgbox_startup(struct mbox_chan *chan)
+{
+	struct sun6i_msgbox *mbox = to_sun6i_msgbox(chan);
+	int n = channel_number(chan);
+
+	/* The coprocessor is responsible for setting channel directions. */
+	if (readl(mbox->regs + CTRL_REG(n)) & CTRL_RX(n)) {
+		/* Flush the receive FIFO. */
+		while (sun6i_msgbox_peek_data(chan))
+			readl(mbox->regs + MSG_DATA_REG(n));
+		writel(RX_IRQ(n), mbox->regs + LOCAL_IRQ_STAT_REG);
+
+		/* Enable the receive IRQ. */
+		spin_lock(&mbox->lock);
+		writel(readl(mbox->regs + LOCAL_IRQ_EN_REG) | RX_IRQ(n),
+		       mbox->regs + LOCAL_IRQ_EN_REG);
+		spin_unlock(&mbox->lock);
+	}
+
+	mbox_dbg(mbox, "Channel %d startup complete\n", n);
+
+	return 0;
+}
+
+static void sun6i_msgbox_shutdown(struct mbox_chan *chan)
+{
+	struct sun6i_msgbox *mbox = to_sun6i_msgbox(chan);
+	int n = channel_number(chan);
+
+	if (readl(mbox->regs + CTRL_REG(n)) & CTRL_RX(n)) {
+		/* Disable the receive IRQ. */
+		spin_lock(&mbox->lock);
+		writel(readl(mbox->regs + LOCAL_IRQ_EN_REG) & ~RX_IRQ(n),
+		       mbox->regs + LOCAL_IRQ_EN_REG);
+		spin_unlock(&mbox->lock);
+
+		/* Attempt to flush the FIFO until the IRQ is cleared. */
+		do {
+			while (sun6i_msgbox_peek_data(chan))
+				readl(mbox->regs + MSG_DATA_REG(n));
+			writel(RX_IRQ(n), mbox->regs + LOCAL_IRQ_STAT_REG);
+		} while (readl(mbox->regs + LOCAL_IRQ_STAT_REG) & RX_IRQ(n));
+	}
+
+	mbox_dbg(mbox, "Channel %d shutdown complete\n", n);
+}
+
+static bool sun6i_msgbox_last_tx_done(struct mbox_chan *chan)
+{
+	struct sun6i_msgbox *mbox = to_sun6i_msgbox(chan);
+	int n = channel_number(chan);
+
+	/*
+	 * The hardware allows snooping on the remote user's IRQ statuses.
+	 * We consider a message to be acknowledged only once the receive IRQ
+	 * for that channel is cleared. Since the receive IRQ for a channel
+	 * cannot be cleared until the FIFO for that channel is empty, this
+	 * ensures that the message has actually been read. It also gives the
+	 * recipient an opportunity to perform minimal processing before
+	 * acknowledging the message.
+	 */
+	return !(readl(mbox->regs + REMOTE_IRQ_STAT_REG) & RX_IRQ(n));
+}
+
+static bool sun6i_msgbox_peek_data(struct mbox_chan *chan)
+{
+	struct sun6i_msgbox *mbox = to_sun6i_msgbox(chan);
+	int n = channel_number(chan);
+
+	return readl(mbox->regs + MSG_STAT_REG(n)) & MSG_STAT_MASK;
+}
+
+static const struct mbox_chan_ops sun6i_msgbox_chan_ops = {
+	.send_data    = sun6i_msgbox_send_data,
+	.startup      = sun6i_msgbox_startup,
+	.shutdown     = sun6i_msgbox_shutdown,
+	.last_tx_done = sun6i_msgbox_last_tx_done,
+	.peek_data    = sun6i_msgbox_peek_data,
+};
+
+static int sun6i_msgbox_probe(struct platform_device *pdev)
+{
+	struct device *dev = &pdev->dev;
+	struct mbox_chan *chans;
+	struct reset_control *reset;
+	struct sun6i_msgbox *mbox;
+	int i, ret;
+
+	mbox = devm_kzalloc(dev, sizeof(*mbox), GFP_KERNEL);
+	if (!mbox)
+		return -ENOMEM;
+
+	chans = devm_kcalloc(dev, NUM_CHANS, sizeof(*chans), GFP_KERNEL);
+	if (!chans)
+		return -ENOMEM;
+
+	for (i = 0; i < NUM_CHANS; ++i)
+		chans[i].con_priv = mbox;
+
+	mbox->clk = devm_clk_get(dev, NULL);
+	if (IS_ERR(mbox->clk)) {
+		ret = PTR_ERR(mbox->clk);
+		dev_err(dev, "Failed to get clock: %d\n", ret);
+		return ret;
+	}
+
+	ret = clk_prepare_enable(mbox->clk);
+	if (ret) {
+		dev_err(dev, "Failed to enable clock: %d\n", ret);
+		return ret;
+	}
+
+	reset = devm_reset_control_get_exclusive(dev, NULL);
+	if (IS_ERR(reset)) {
+		ret = PTR_ERR(reset);
+		dev_err(dev, "Failed to get reset control: %d\n", ret);
+		goto err_disable_unprepare;
+	}
+
+	/*
+	 * NOTE: We rely on platform firmware to preconfigure the channel
+	 * directions, and we share this hardware block with other firmware
+	 * that runs concurrently with Linux (e.g. a trusted monitor).
+	 *
+	 * Therefore, we do *not* assert the reset line if probing fails or
+	 * when removing the device.
+	 */
+	ret = reset_control_deassert(reset);
+	if (ret) {
+		dev_err(dev, "Failed to deassert reset: %d\n", ret);
+		goto err_disable_unprepare;
+	}
+
+	mbox->regs = devm_platform_ioremap_resource(pdev, 0);
+	if (IS_ERR(mbox->regs)) {
+		ret = PTR_ERR(mbox->regs);
+		dev_err(dev, "Failed to map MMIO resource: %d\n", ret);
+		goto err_disable_unprepare;
+	}
+
+	/* Disable all IRQs for this end of the msgbox. */
+	writel(0, mbox->regs + LOCAL_IRQ_EN_REG);
+
+	ret = devm_request_irq(dev, irq_of_parse_and_map(dev->of_node, 0),
+			       sun6i_msgbox_irq, 0, dev_name(dev), mbox);
+	if (ret)
+		goto err_disable_unprepare;
+
+	mbox->controller.dev           = dev;
+	mbox->controller.ops           = &sun6i_msgbox_chan_ops;
+	mbox->controller.chans         = chans;
+	mbox->controller.num_chans     = NUM_CHANS;
+	mbox->controller.txdone_irq    = false;
+	mbox->controller.txdone_poll   = true;
+	mbox->controller.txpoll_period = 5;
+
+	spin_lock_init(&mbox->lock);
+	platform_set_drvdata(pdev, mbox);
+
+	ret = mbox_controller_register(&mbox->controller);
+	if (ret) {
+		dev_err(dev, "Failed to register controller: %d\n", ret);
+		goto err_disable_unprepare;
+	}
+
+	return 0;
+
+err_disable_unprepare:
+	clk_disable_unprepare(mbox->clk);
+
+	return ret;
+}
+
+static void sun6i_msgbox_remove(struct platform_device *pdev)
+{
+	struct sun6i_msgbox *mbox = platform_get_drvdata(pdev);
+
+	mbox_controller_unregister(&mbox->controller);
+	/* See the comment in sun6i_msgbox_probe about the reset line. */
+	clk_disable_unprepare(mbox->clk);
+}
+
+static const struct of_device_id sun6i_msgbox_of_match[] = {
+	{ .compatible = "allwinner,sun6i-a31-msgbox", },
+	{},
+};
+MODULE_DEVICE_TABLE(of, sun6i_msgbox_of_match);
+
+static struct platform_driver sun6i_msgbox_driver = {
+	.driver = {
+		.name = "sun6i-msgbox",
+		.of_match_table = sun6i_msgbox_of_match,
+	},
+	.probe = sun6i_msgbox_probe,
+	.remove = sun6i_msgbox_remove,
+};
+module_platform_driver(sun6i_msgbox_driver);
+
+MODULE_AUTHOR("Samuel Holland <samuel@sholland.org>");
+MODULE_DESCRIPTION("Allwinner sun6i/sun8i/sun9i/sun50i Message Box");
+MODULE_LICENSE("GPL v2");