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, 172 insertions, 0 deletions

diff --git a/drivers/clk/ti/clockdomain.c b/drivers/clk/ti/clockdomain.c
new file mode 100644
index 000000000..c897ad7e6
--- /dev/null
+++ b/drivers/clk/ti/clockdomain.c
@@ -0,0 +1,172 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * OMAP clockdomain support
+ *
+ * Copyright (C) 2013 Texas Instruments, Inc.
+ *
+ * Tero Kristo <t-kristo@ti.com>
+ */
+
+#include <linux/clk.h>
+#include <linux/clk-provider.h>
+#include <linux/slab.h>
+#include <linux/of.h>
+#include <linux/of_address.h>
+#include <linux/clk/ti.h>
+
+#include "clock.h"
+
+#undef pr_fmt
+#define pr_fmt(fmt) "%s: " fmt, __func__
+
+/**
+ * omap2_clkops_enable_clkdm - increment usecount on clkdm of @hw
+ * @hw: struct clk_hw * of the clock being enabled
+ *
+ * Increment the usecount of the clockdomain of the clock pointed to
+ * by @hw; if the usecount is 1, the clockdomain will be "enabled."
+ * Only needed for clocks that don't use omap2_dflt_clk_enable() as
+ * their enable function pointer.  Passes along the return value of
+ * clkdm_clk_enable(), -EINVAL if @hw is not associated with a
+ * clockdomain, or 0 if clock framework-based clockdomain control is
+ * not implemented.
+ */
+int omap2_clkops_enable_clkdm(struct clk_hw *hw)
+{
+	struct clk_hw_omap *clk;
+	int ret = 0;
+
+	clk = to_clk_hw_omap(hw);
+
+	if (unlikely(!clk->clkdm)) {
+		pr_err("%s: %s: no clkdm set ?!\n", __func__,
+		       clk_hw_get_name(hw));
+		return -EINVAL;
+	}
+
+	if (ti_clk_get_features()->flags & TI_CLK_DISABLE_CLKDM_CONTROL) {
+		pr_err("%s: %s: clkfw-based clockdomain control disabled ?!\n",
+		       __func__, clk_hw_get_name(hw));
+		return 0;
+	}
+
+	ret = ti_clk_ll_ops->clkdm_clk_enable(clk->clkdm, hw->clk);
+	WARN(ret, "%s: could not enable %s's clockdomain %s: %d\n",
+	     __func__, clk_hw_get_name(hw), clk->clkdm_name, ret);
+
+	return ret;
+}
+
+/**
+ * omap2_clkops_disable_clkdm - decrement usecount on clkdm of @hw
+ * @hw: struct clk_hw * of the clock being disabled
+ *
+ * Decrement the usecount of the clockdomain of the clock pointed to
+ * by @hw; if the usecount is 0, the clockdomain will be "disabled."
+ * Only needed for clocks that don't use omap2_dflt_clk_disable() as their
+ * disable function pointer.  No return value.
+ */
+void omap2_clkops_disable_clkdm(struct clk_hw *hw)
+{
+	struct clk_hw_omap *clk;
+
+	clk = to_clk_hw_omap(hw);
+
+	if (unlikely(!clk->clkdm)) {
+		pr_err("%s: %s: no clkdm set ?!\n", __func__,
+		       clk_hw_get_name(hw));
+		return;
+	}
+
+	if (ti_clk_get_features()->flags & TI_CLK_DISABLE_CLKDM_CONTROL) {
+		pr_err("%s: %s: clkfw-based clockdomain control disabled ?!\n",
+		       __func__, clk_hw_get_name(hw));
+		return;
+	}
+
+	ti_clk_ll_ops->clkdm_clk_disable(clk->clkdm, hw->clk);
+}
+
+/**
+ * omap2_init_clk_clkdm - look up a clockdomain name, store pointer in clk
+ * @hw: Pointer to clk_hw_omap used to obtain OMAP clock struct ptr to use
+ *
+ * Convert a clockdomain name stored in a struct clk 'clk' into a
+ * clockdomain pointer, and save it into the struct clk.  Intended to be
+ * called during clk_register(). Returns 0 on success, -EERROR otherwise.
+ */
+int omap2_init_clk_clkdm(struct clk_hw *hw)
+{
+	struct clk_hw_omap *clk = to_clk_hw_omap(hw);
+	struct clockdomain *clkdm;
+	const char *clk_name;
+
+	if (!clk->clkdm_name)
+		return 0;
+
+	clk_name = __clk_get_name(hw->clk);
+
+	clkdm = ti_clk_ll_ops->clkdm_lookup(clk->clkdm_name);
+	if (clkdm) {
+		pr_debug("clock: associated clk %s to clkdm %s\n",
+			 clk_name, clk->clkdm_name);
+		clk->clkdm = clkdm;
+	} else {
+		pr_debug("clock: could not associate clk %s to clkdm %s\n",
+			 clk_name, clk->clkdm_name);
+	}
+
+	return 0;
+}
+
+static void __init of_ti_clockdomain_setup(struct device_node *node)
+{
+	struct clk *clk;
+	struct clk_hw *clk_hw;
+	const char *clkdm_name = ti_dt_clk_name(node);
+	int i;
+	unsigned int num_clks;
+
+	num_clks = of_clk_get_parent_count(node);
+
+	for (i = 0; i < num_clks; i++) {
+		clk = of_clk_get(node, i);
+		if (IS_ERR(clk)) {
+			pr_err("%s: Failed get %pOF' clock nr %d (%ld)\n",
+			       __func__, node, i, PTR_ERR(clk));
+			continue;
+		}
+		clk_hw = __clk_get_hw(clk);
+		if (!omap2_clk_is_hw_omap(clk_hw)) {
+			pr_warn("can't setup clkdm for basic clk %s\n",
+				__clk_get_name(clk));
+			clk_put(clk);
+			continue;
+		}
+		to_clk_hw_omap(clk_hw)->clkdm_name = clkdm_name;
+		omap2_init_clk_clkdm(clk_hw);
+		clk_put(clk);
+	}
+}
+
+static const struct of_device_id ti_clkdm_match_table[] __initconst = {
+	{ .compatible = "ti,clockdomain" },
+	{ }
+};
+
+/**
+ * ti_dt_clockdomains_setup - setup device tree clockdomains
+ *
+ * Initializes clockdomain nodes for a SoC. This parses through all the
+ * nodes with compatible = "ti,clockdomain", and add the clockdomain
+ * info for all the clocks listed under these. This function shall be
+ * called after rest of the DT clock init has completed and all
+ * clock nodes have been registered.
+ */
+void __init ti_dt_clockdomains_setup(void)
+{
+	struct device_node *np;
+	for_each_matching_node(np, ti_clkdm_match_table) {
+		of_ti_clockdomain_setup(np);
+	}
+}