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, 239 insertions, 0 deletions

diff --git a/drivers/firmware/efi/apple-properties.c b/drivers/firmware/efi/apple-properties.c
new file mode 100644
index 000000000..f401a04ed
--- /dev/null
+++ b/drivers/firmware/efi/apple-properties.c
@@ -0,0 +1,239 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * apple-properties.c - EFI device properties on Macs
+ * Copyright (C) 2016 Lukas Wunner <lukas@wunner.de>
+ *
+ * Properties are stored either as:
+ * u8 arrays which can be retrieved with device_property_read_u8_array() or
+ * booleans which can be queried with device_property_present().
+ */
+
+#define pr_fmt(fmt) "apple-properties: " fmt
+
+#include <linux/memblock.h>
+#include <linux/efi.h>
+#include <linux/io.h>
+#include <linux/platform_data/x86/apple.h>
+#include <linux/property.h>
+#include <linux/slab.h>
+#include <linux/ucs2_string.h>
+#include <asm/setup.h>
+
+static bool dump_properties __initdata;
+
+static int __init dump_properties_enable(char *arg)
+{
+	dump_properties = true;
+	return 1;
+}
+
+__setup("dump_apple_properties", dump_properties_enable);
+
+struct dev_header {
+	u32 len;
+	u32 prop_count;
+	struct efi_dev_path path[];
+	/*
+	 * followed by key/value pairs, each key and value preceded by u32 len,
+	 * len includes itself, value may be empty (in which case its len is 4)
+	 */
+};
+
+struct properties_header {
+	u32 len;
+	u32 version;
+	u32 dev_count;
+	struct dev_header dev_header[];
+};
+
+static void __init unmarshal_key_value_pairs(struct dev_header *dev_header,
+					     struct device *dev, const void *ptr,
+					     struct property_entry entry[])
+{
+	int i;
+
+	for (i = 0; i < dev_header->prop_count; i++) {
+		int remaining = dev_header->len - (ptr - (void *)dev_header);
+		u32 key_len, val_len, entry_len;
+		const u8 *entry_data;
+		char *key;
+
+		if (sizeof(key_len) > remaining)
+			break;
+
+		key_len = *(typeof(key_len) *)ptr;
+		if (key_len + sizeof(val_len) > remaining ||
+		    key_len < sizeof(key_len) + sizeof(efi_char16_t) ||
+		    *(efi_char16_t *)(ptr + sizeof(key_len)) == 0) {
+			dev_err(dev, "invalid property name len at %#zx\n",
+				ptr - (void *)dev_header);
+			break;
+		}
+
+		val_len = *(typeof(val_len) *)(ptr + key_len);
+		if (key_len + val_len > remaining ||
+		    val_len < sizeof(val_len)) {
+			dev_err(dev, "invalid property val len at %#zx\n",
+				ptr - (void *)dev_header + key_len);
+			break;
+		}
+
+		/* 4 bytes to accommodate UTF-8 code points + null byte */
+		key = kzalloc((key_len - sizeof(key_len)) * 4 + 1, GFP_KERNEL);
+		if (!key) {
+			dev_err(dev, "cannot allocate property name\n");
+			break;
+		}
+		ucs2_as_utf8(key, ptr + sizeof(key_len),
+			     key_len - sizeof(key_len));
+
+		entry_data = ptr + key_len + sizeof(val_len);
+		entry_len = val_len - sizeof(val_len);
+		if (entry_len)
+			entry[i] = PROPERTY_ENTRY_U8_ARRAY_LEN(key, entry_data,
+							       entry_len);
+		else
+			entry[i] = PROPERTY_ENTRY_BOOL(key);
+
+		if (dump_properties) {
+			dev_info(dev, "property: %s\n", key);
+			print_hex_dump(KERN_INFO, pr_fmt(), DUMP_PREFIX_OFFSET,
+				16, 1, entry_data, entry_len, true);
+		}
+
+		ptr += key_len + val_len;
+	}
+
+	if (i != dev_header->prop_count) {
+		dev_err(dev, "got %d device properties, expected %u\n", i,
+			dev_header->prop_count);
+		print_hex_dump(KERN_ERR, pr_fmt(), DUMP_PREFIX_OFFSET,
+			16, 1, dev_header, dev_header->len, true);
+		return;
+	}
+
+	dev_info(dev, "assigning %d device properties\n", i);
+}
+
+static int __init unmarshal_devices(struct properties_header *properties)
+{
+	size_t offset = offsetof(struct properties_header, dev_header[0]);
+
+	while (offset + sizeof(struct dev_header) < properties->len) {
+		struct dev_header *dev_header = (void *)properties + offset;
+		struct property_entry *entry = NULL;
+		const struct efi_dev_path *ptr;
+		struct device *dev;
+		size_t len;
+		int ret, i;
+
+		if (offset + dev_header->len > properties->len ||
+		    dev_header->len <= sizeof(*dev_header)) {
+			pr_err("invalid len in dev_header at %#zx\n", offset);
+			return -EINVAL;
+		}
+
+		ptr = dev_header->path;
+		len = dev_header->len - sizeof(*dev_header);
+
+		dev = efi_get_device_by_path(&ptr, &len);
+		if (IS_ERR(dev)) {
+			pr_err("device path parse error %ld at %#zx:\n",
+			       PTR_ERR(dev), (void *)ptr - (void *)dev_header);
+			print_hex_dump(KERN_ERR, pr_fmt(), DUMP_PREFIX_OFFSET,
+			       16, 1, dev_header, dev_header->len, true);
+			dev = NULL;
+			goto skip_device;
+		}
+
+		entry = kzalloc_objs(*entry, dev_header->prop_count + 1);
+		if (!entry) {
+			dev_err(dev, "cannot allocate properties\n");
+			goto skip_device;
+		}
+
+		unmarshal_key_value_pairs(dev_header, dev, ptr, entry);
+		if (!entry[0].name)
+			goto skip_device;
+
+		ret = device_create_managed_software_node(dev, entry, NULL);
+		if (ret)
+			dev_err(dev, "error %d assigning properties\n", ret);
+
+		for (i = 0; entry[i].name; i++)
+			kfree(entry[i].name);
+
+skip_device:
+		kfree(entry);
+		put_device(dev);
+		offset += dev_header->len;
+	}
+
+	return 0;
+}
+
+static int __init map_properties(void)
+{
+	struct properties_header *properties;
+	struct setup_data *data;
+	u32 data_len;
+	u64 pa_data;
+	int ret;
+
+	if (!x86_apple_machine)
+		return 0;
+
+	pa_data = boot_params.hdr.setup_data;
+	while (pa_data) {
+		data = memremap(pa_data, sizeof(*data), MEMREMAP_WB);
+		if (!data) {
+			pr_err("cannot map setup_data header\n");
+			return -ENOMEM;
+		}
+
+		if (data->type != SETUP_APPLE_PROPERTIES) {
+			pa_data = data->next;
+			memunmap(data);
+			continue;
+		}
+
+		data_len = data->len;
+		memunmap(data);
+
+		data = memremap(pa_data, sizeof(*data) + data_len, MEMREMAP_WB);
+		if (!data) {
+			pr_err("cannot map setup_data payload\n");
+			return -ENOMEM;
+		}
+
+		properties = (struct properties_header *)data->data;
+		if (data_len < sizeof(*properties)) {
+			pr_err("truncated properties header\n");
+			ret = -EINVAL;
+		} else if (properties->version != 1) {
+			pr_err("unsupported version:\n");
+			print_hex_dump(KERN_ERR, pr_fmt(), DUMP_PREFIX_OFFSET,
+			       16, 1, properties, data_len, true);
+			ret = -ENOTSUPP;
+		} else if (properties->len != data_len) {
+			pr_err("length mismatch, expected %u\n", data_len);
+			print_hex_dump(KERN_ERR, pr_fmt(), DUMP_PREFIX_OFFSET,
+			       16, 1, properties, data_len, true);
+			ret = -EINVAL;
+		} else
+			ret = unmarshal_devices(properties);
+
+		/*
+		 * Can only free the setup_data payload but not its header
+		 * to avoid breaking the chain of ->next pointers.
+		 */
+		data->len = 0;
+		memunmap(data);
+		memblock_phys_free(pa_data + sizeof(*data), data_len);
+
+		return ret;
+	}
+	return 0;
+}
+
+fs_initcall(map_properties);