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, 213 insertions, 0 deletions

diff --git a/drivers/acpi/nvs.c b/drivers/acpi/nvs.c
new file mode 100644
index 000000000..12aee4102
--- /dev/null
+++ b/drivers/acpi/nvs.c
@@ -0,0 +1,213 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * nvs.c - Routines for saving and restoring ACPI NVS memory region
+ *
+ * Copyright (C) 2008-2011 Rafael J. Wysocki <rjw@sisk.pl>, Novell Inc.
+ */
+
+#define pr_fmt(fmt) "ACPI: PM: " fmt
+
+#include <linux/io.h>
+#include <linux/kernel.h>
+#include <linux/list.h>
+#include <linux/mm.h>
+#include <linux/slab.h>
+#include <linux/acpi.h>
+
+#include "internal.h"
+
+/* ACPI NVS regions, APEI may use it */
+
+struct nvs_region {
+	__u64 phys_start;
+	__u64 size;
+	struct list_head node;
+};
+
+static LIST_HEAD(nvs_region_list);
+
+#ifdef CONFIG_ACPI_SLEEP
+static int suspend_nvs_register(unsigned long start, unsigned long size);
+#else
+static inline int suspend_nvs_register(unsigned long a, unsigned long b)
+{
+	return 0;
+}
+#endif
+
+int acpi_nvs_register(__u64 start, __u64 size)
+{
+	struct nvs_region *region;
+
+	region = kmalloc_obj(*region);
+	if (!region)
+		return -ENOMEM;
+	region->phys_start = start;
+	region->size = size;
+	list_add_tail(&region->node, &nvs_region_list);
+
+	return suspend_nvs_register(start, size);
+}
+
+int acpi_nvs_for_each_region(int (*func)(__u64 start, __u64 size, void *data),
+			     void *data)
+{
+	int rc;
+	struct nvs_region *region;
+
+	list_for_each_entry(region, &nvs_region_list, node) {
+		rc = func(region->phys_start, region->size, data);
+		if (rc)
+			return rc;
+	}
+
+	return 0;
+}
+
+
+#ifdef CONFIG_ACPI_SLEEP
+/*
+ * Platforms, like ACPI, may want us to save some memory used by them during
+ * suspend and to restore the contents of this memory during the subsequent
+ * resume.  The code below implements a mechanism allowing us to do that.
+ */
+
+struct nvs_page {
+	unsigned long phys_start;
+	unsigned int size;
+	void *kaddr;
+	void *data;
+	bool unmap;
+	struct list_head node;
+};
+
+static LIST_HEAD(nvs_list);
+
+/**
+ * suspend_nvs_register - register platform NVS memory region to save
+ * @start: Physical address of the region.
+ * @size: Size of the region.
+ *
+ * The NVS region need not be page-aligned (both ends) and we arrange
+ * things so that the data from page-aligned addresses in this region will
+ * be copied into separate RAM pages.
+ */
+static int suspend_nvs_register(unsigned long start, unsigned long size)
+{
+	struct nvs_page *entry, *next;
+
+	pr_info("Registering ACPI NVS region [mem %#010lx-%#010lx] (%ld bytes)\n",
+		start, start + size - 1, size);
+
+	while (size > 0) {
+		unsigned int nr_bytes;
+
+		entry = kzalloc_obj(struct nvs_page);
+		if (!entry)
+			goto Error;
+
+		list_add_tail(&entry->node, &nvs_list);
+		entry->phys_start = start;
+		nr_bytes = PAGE_SIZE - (start & ~PAGE_MASK);
+		entry->size = (size < nr_bytes) ? size : nr_bytes;
+
+		start += entry->size;
+		size -= entry->size;
+	}
+	return 0;
+
+ Error:
+	list_for_each_entry_safe(entry, next, &nvs_list, node) {
+		list_del(&entry->node);
+		kfree(entry);
+	}
+	return -ENOMEM;
+}
+
+/**
+ * suspend_nvs_free - free data pages allocated for saving NVS regions
+ */
+void suspend_nvs_free(void)
+{
+	struct nvs_page *entry;
+
+	list_for_each_entry(entry, &nvs_list, node)
+		if (entry->data) {
+			kfree(entry->data);
+			entry->data = NULL;
+			if (entry->kaddr) {
+				if (entry->unmap) {
+					iounmap(entry->kaddr);
+					entry->unmap = false;
+				} else {
+					acpi_os_unmap_iomem(entry->kaddr,
+							    entry->size);
+				}
+				entry->kaddr = NULL;
+			}
+		}
+}
+
+/**
+ * suspend_nvs_alloc - allocate memory necessary for saving NVS regions
+ */
+int suspend_nvs_alloc(void)
+{
+	struct nvs_page *entry;
+
+	list_for_each_entry(entry, &nvs_list, node) {
+		entry->data = kmalloc(PAGE_SIZE, GFP_KERNEL);
+		if (!entry->data) {
+			suspend_nvs_free();
+			return -ENOMEM;
+		}
+	}
+	return 0;
+}
+
+/**
+ * suspend_nvs_save - save NVS memory regions
+ */
+int suspend_nvs_save(void)
+{
+	struct nvs_page *entry;
+
+	pr_info("Saving platform NVS memory\n");
+
+	list_for_each_entry(entry, &nvs_list, node)
+		if (entry->data) {
+			unsigned long phys = entry->phys_start;
+			unsigned int size = entry->size;
+
+			entry->kaddr = acpi_os_get_iomem(phys, size);
+			if (!entry->kaddr) {
+				entry->kaddr = acpi_os_ioremap(phys, size);
+				entry->unmap = !!entry->kaddr;
+			}
+			if (!entry->kaddr) {
+				suspend_nvs_free();
+				return -ENOMEM;
+			}
+			memcpy(entry->data, entry->kaddr, entry->size);
+		}
+
+	return 0;
+}
+
+/**
+ * suspend_nvs_restore - restore NVS memory regions
+ *
+ * This function is going to be called with interrupts disabled, so it
+ * cannot iounmap the virtual addresses used to access the NVS region.
+ */
+void suspend_nvs_restore(void)
+{
+	struct nvs_page *entry;
+
+	pr_info("Restoring platform NVS memory\n");
+
+	list_for_each_entry(entry, &nvs_list, node)
+		if (entry->data)
+			memcpy(entry->kaddr, entry->data, entry->size);
+}
+#endif