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, 121 insertions, 0 deletions

diff --git a/drivers/misc/sram-exec.c b/drivers/misc/sram-exec.c
new file mode 100644
index 000000000..b71dbbd73
--- /dev/null
+++ b/drivers/misc/sram-exec.c
@@ -0,0 +1,121 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * SRAM protect-exec region helper functions
+ *
+ * Copyright (C) 2017 Texas Instruments Incorporated - https://www.ti.com/
+ *	Dave Gerlach
+ */
+
+#include <linux/device.h>
+#include <linux/genalloc.h>
+#include <linux/mm.h>
+#include <linux/sram.h>
+#include <linux/set_memory.h>
+
+#include <asm/fncpy.h>
+
+#include "sram.h"
+
+static DEFINE_MUTEX(exec_pool_list_mutex);
+static LIST_HEAD(exec_pool_list);
+
+int sram_check_protect_exec(struct sram_dev *sram, struct sram_reserve *block,
+			    struct sram_partition *part)
+{
+	unsigned long base = (unsigned long)part->base;
+	unsigned long end = base + block->size;
+
+	if (!PAGE_ALIGNED(base) || !PAGE_ALIGNED(end)) {
+		dev_err(sram->dev,
+			"SRAM pool marked with 'protect-exec' is not page aligned and will not be created.\n");
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+int sram_add_protect_exec(struct sram_partition *part)
+{
+	mutex_lock(&exec_pool_list_mutex);
+	list_add_tail(&part->list, &exec_pool_list);
+	mutex_unlock(&exec_pool_list_mutex);
+
+	return 0;
+}
+
+/**
+ * sram_exec_copy - copy data to a protected executable region of sram
+ *
+ * @pool: struct gen_pool retrieved that is part of this sram
+ * @dst: Destination address for the copy, that must be inside pool
+ * @src: Source address for the data to copy
+ * @size: Size of copy to perform, which starting from dst, must reside in pool
+ *
+ * Return: Address for copied data that can safely be called through function
+ *	   pointer, or NULL if problem.
+ *
+ * This helper function allows sram driver to act as central control location
+ * of 'protect-exec' pools which are normal sram pools but are always set
+ * read-only and executable except when copying data to them, at which point
+ * they are set to read-write non-executable, to make sure no memory is
+ * writeable and executable at the same time. This region must be page-aligned
+ * and is checked during probe, otherwise page attribute manipulation would
+ * not be possible. Care must be taken to only call the returned address as
+ * dst address is not guaranteed to be safely callable.
+ *
+ * NOTE: This function uses the fncpy macro to move code to the executable
+ * region. Some architectures have strict requirements for relocating
+ * executable code, so fncpy is a macro that must be defined by any arch
+ * making use of this functionality that guarantees a safe copy of exec
+ * data and returns a safe address that can be called as a C function
+ * pointer.
+ */
+void *sram_exec_copy(struct gen_pool *pool, void *dst, void *src,
+		     size_t size)
+{
+	struct sram_partition *part = NULL, *p;
+	unsigned long base;
+	int pages;
+	void *dst_cpy;
+	int ret;
+
+	mutex_lock(&exec_pool_list_mutex);
+	list_for_each_entry(p, &exec_pool_list, list) {
+		if (p->pool == pool)
+			part = p;
+	}
+	mutex_unlock(&exec_pool_list_mutex);
+
+	if (!part)
+		return NULL;
+
+	if (!gen_pool_has_addr(pool, (unsigned long)dst, size))
+		return NULL;
+
+	base = (unsigned long)part->base;
+	pages = PAGE_ALIGN(size) / PAGE_SIZE;
+
+	mutex_lock(&part->lock);
+
+	ret = set_memory_nx((unsigned long)base, pages);
+	if (ret)
+		goto error_out;
+	ret = set_memory_rw((unsigned long)base, pages);
+	if (ret)
+		goto error_out;
+
+	dst_cpy = fncpy(dst, src, size);
+
+	ret = set_memory_rox((unsigned long)base, pages);
+	if (ret)
+		goto error_out;
+
+	mutex_unlock(&part->lock);
+
+	return dst_cpy;
+
+error_out:
+	mutex_unlock(&part->lock);
+	return NULL;
+}
+EXPORT_SYMBOL_GPL(sram_exec_copy);