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, 178 insertions, 0 deletions

diff --git a/drivers/nvmem/rmem.c b/drivers/nvmem/rmem.c
new file mode 100644
index 000000000..b39d628cb
--- /dev/null
+++ b/drivers/nvmem/rmem.c
@@ -0,0 +1,178 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Copyright (C) 2020 Nicolas Saenz Julienne <nsaenzjulienne@suse.de>
+ */
+
+#include <linux/crc32.h>
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/nvmem-provider.h>
+#include <linux/of_reserved_mem.h>
+#include <linux/platform_device.h>
+#include <linux/slab.h>
+
+struct rmem {
+	struct device *dev;
+	struct nvmem_device *nvmem;
+	struct reserved_mem *mem;
+};
+
+struct rmem_match_data {
+	int (*checksum)(struct rmem *priv);
+};
+
+struct __packed rmem_eyeq5_header {
+	u32 magic;
+	u32 version;
+	u32 size;
+};
+
+#define RMEM_EYEQ5_MAGIC	((u32)0xDABBAD00)
+
+static int rmem_read(void *context, unsigned int offset,
+		     void *val, size_t bytes)
+{
+	struct rmem *priv = context;
+	void *addr;
+
+	if ((phys_addr_t)offset + bytes > priv->mem->size)
+		return -EIO;
+
+	/*
+	 * Only map the reserved memory at this point to avoid potential rogue
+	 * kernel threads inadvertently modifying it. Based on the current
+	 * uses-cases for this driver, the performance hit isn't a concern.
+	 * Nor is likely to be, given the nature of the subsystem. Most nvmem
+	 * devices operate over slow buses to begin with.
+	 *
+	 * An alternative would be setting the memory as RO, set_memory_ro(),
+	 * but as of Dec 2020 this isn't possible on arm64.
+	 */
+	addr = memremap(priv->mem->base, priv->mem->size, MEMREMAP_WB);
+	if (!addr) {
+		dev_err(priv->dev, "Failed to remap memory region\n");
+		return -ENOMEM;
+	}
+
+	memcpy(val, addr + offset, bytes);
+
+	memunmap(addr);
+
+	return 0;
+}
+
+static int rmem_eyeq5_checksum(struct rmem *priv)
+{
+	void *buf __free(kfree) = NULL;
+	struct rmem_eyeq5_header header;
+	u32 computed_crc, *target_crc;
+	size_t data_size;
+	int ret;
+
+	ret = rmem_read(priv, 0, &header, sizeof(header));
+	if (ret)
+		return ret;
+
+	if (header.magic != RMEM_EYEQ5_MAGIC)
+		return -EINVAL;
+
+	/*
+	 * Avoid massive kmalloc() if header read is invalid;
+	 * the check would be done by the next rmem_read() anyway.
+	 */
+	if (header.size > priv->mem->size)
+		return -EINVAL;
+
+	/*
+	 *           0 +-------------------+
+	 *             | Header (12 bytes) | \
+	 *             +-------------------+ |
+	 *             |                   | | data to be CRCed
+	 *             |        ...        | |
+	 *             |                   | /
+	 *   data_size +-------------------+
+	 *             |   CRC (4 bytes)   |
+	 * header.size +-------------------+
+	 */
+
+	buf = kmalloc(header.size, GFP_KERNEL);
+	if (!buf)
+		return -ENOMEM;
+
+	ret = rmem_read(priv, 0, buf, header.size);
+	if (ret)
+		return ret;
+
+	data_size = header.size - sizeof(*target_crc);
+	target_crc = buf + data_size;
+	computed_crc = crc32(U32_MAX, buf, data_size) ^ U32_MAX;
+
+	if (computed_crc == *target_crc)
+		return 0;
+
+	dev_err(priv->dev,
+		"checksum failed: computed %#x, expected %#x, header (%#x, %#x, %#x)\n",
+		computed_crc, *target_crc, header.magic, header.version, header.size);
+	return -EINVAL;
+}
+
+static int rmem_probe(struct platform_device *pdev)
+{
+	struct nvmem_config config = { };
+	struct device *dev = &pdev->dev;
+	const struct rmem_match_data *match_data = device_get_match_data(dev);
+	struct reserved_mem *mem;
+	struct rmem *priv;
+
+	priv = devm_kzalloc(dev, sizeof(*priv), GFP_KERNEL);
+	if (!priv)
+		return -ENOMEM;
+	priv->dev = dev;
+
+	mem = of_reserved_mem_lookup(dev->of_node);
+	if (!mem) {
+		dev_err(dev, "Failed to lookup reserved memory\n");
+		return -EINVAL;
+	}
+	priv->mem = mem;
+
+	config.dev = dev;
+	config.priv = priv;
+	config.name = "rmem";
+	config.id = NVMEM_DEVID_AUTO;
+	config.size = mem->size;
+	config.reg_read = rmem_read;
+
+	if (match_data && match_data->checksum) {
+		int ret = match_data->checksum(priv);
+
+		if (ret)
+			return ret;
+	}
+
+	return PTR_ERR_OR_ZERO(devm_nvmem_register(dev, &config));
+}
+
+static const struct rmem_match_data rmem_eyeq5_match_data = {
+	.checksum = rmem_eyeq5_checksum,
+};
+
+static const struct of_device_id rmem_match[] = {
+	{ .compatible = "mobileye,eyeq5-bootloader-config", .data = &rmem_eyeq5_match_data },
+	{ .compatible = "nvmem-rmem", },
+	{ /* sentinel */ },
+};
+MODULE_DEVICE_TABLE(of, rmem_match);
+
+static struct platform_driver rmem_driver = {
+	.probe = rmem_probe,
+	.driver = {
+		.name = "rmem",
+		.of_match_table = rmem_match,
+	},
+};
+module_platform_driver(rmem_driver);
+
+MODULE_AUTHOR("Nicolas Saenz Julienne <nsaenzjulienne@suse.de>");
+MODULE_DESCRIPTION("Reserved Memory Based nvmem Driver");
+MODULE_LICENSE("GPL");