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, 142 insertions, 0 deletions

diff --git a/drivers/clk/tegra/cvb.c b/drivers/clk/tegra/cvb.c
new file mode 100644
index 000000000..a7fdc7622
--- /dev/null
+++ b/drivers/clk/tegra/cvb.c
@@ -0,0 +1,142 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Utility functions for parsing Tegra CVB voltage tables
+ *
+ * Copyright (C) 2012-2019 NVIDIA Corporation.  All rights reserved.
+ */
+#include <linux/err.h>
+#include <linux/kernel.h>
+#include <linux/pm_opp.h>
+
+#include "cvb.h"
+
+/* cvb_mv = ((c2 * speedo / s_scale + c1) * speedo / s_scale + c0) */
+static inline int get_cvb_voltage(int speedo, int s_scale,
+				  const struct cvb_coefficients *cvb)
+{
+	int mv;
+
+	/* apply only speedo scale: output mv = cvb_mv * v_scale */
+	mv = DIV_ROUND_CLOSEST(cvb->c2 * speedo, s_scale);
+	mv = DIV_ROUND_CLOSEST((mv + cvb->c1) * speedo, s_scale) + cvb->c0;
+	return mv;
+}
+
+static int round_cvb_voltage(int mv, int v_scale,
+			     const struct rail_alignment *align)
+{
+	/* combined: apply voltage scale and round to cvb alignment step */
+	int uv;
+	int step = (align->step_uv ? : 1000) * v_scale;
+	int offset = align->offset_uv * v_scale;
+
+	uv = max(mv * 1000, offset) - offset;
+	uv = DIV_ROUND_UP(uv, step) * align->step_uv + align->offset_uv;
+	return uv / 1000;
+}
+
+enum {
+	DOWN,
+	UP
+};
+
+static int round_voltage(int mv, const struct rail_alignment *align, int up)
+{
+	if (align->step_uv) {
+		int uv;
+
+		uv = max(mv * 1000, align->offset_uv) - align->offset_uv;
+		uv = (uv + (up ? align->step_uv - 1 : 0)) / align->step_uv;
+		return (uv * align->step_uv + align->offset_uv) / 1000;
+	}
+	return mv;
+}
+
+static int build_opp_table(struct device *dev, const struct cvb_table *table,
+			   struct rail_alignment *align,
+			   int speedo_value, unsigned long max_freq)
+{
+	int i, ret, dfll_mv, min_mv, max_mv;
+
+	min_mv = round_voltage(table->min_millivolts, align, UP);
+	max_mv = round_voltage(table->max_millivolts, align, DOWN);
+
+	for (i = 0; i < MAX_DVFS_FREQS; i++) {
+		const struct cvb_table_freq_entry *entry = &table->entries[i];
+
+		if (!entry->freq || (entry->freq > max_freq))
+			break;
+
+		dfll_mv = get_cvb_voltage(speedo_value, table->speedo_scale,
+					  &entry->coefficients);
+		dfll_mv = round_cvb_voltage(dfll_mv, table->voltage_scale,
+					    align);
+		dfll_mv = clamp(dfll_mv, min_mv, max_mv);
+
+		ret = dev_pm_opp_add(dev, entry->freq, dfll_mv * 1000);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
+/**
+ * tegra_cvb_add_opp_table - build OPP table from Tegra CVB tables
+ * @dev: the struct device * for which the OPP table is built
+ * @tables: array of CVB tables
+ * @count: size of the previously mentioned array
+ * @align: parameters of the regulator step and offset
+ * @process_id: process id of the HW module
+ * @speedo_id: speedo id of the HW module
+ * @speedo_value: speedo value of the HW module
+ * @max_freq: highest safe clock rate
+ *
+ * On Tegra, a CVB table encodes the relationship between operating voltage
+ * and safe maximal frequency for a given module (e.g. GPU or CPU). This
+ * function calculates the optimal voltage-frequency operating points
+ * for the given arguments and exports them via the OPP library for the
+ * given @dev. Returns a pointer to the struct cvb_table that matched
+ * or an ERR_PTR on failure.
+ */
+const struct cvb_table *
+tegra_cvb_add_opp_table(struct device *dev, const struct cvb_table *tables,
+			size_t count, struct rail_alignment *align,
+			int process_id, int speedo_id, int speedo_value,
+			unsigned long max_freq)
+{
+	size_t i;
+	int ret;
+
+	for (i = 0; i < count; i++) {
+		const struct cvb_table *table = &tables[i];
+
+		if (table->speedo_id != -1 && table->speedo_id != speedo_id)
+			continue;
+
+		if (table->process_id != -1 && table->process_id != process_id)
+			continue;
+
+		ret = build_opp_table(dev, table, align, speedo_value,
+				      max_freq);
+		return ret ? ERR_PTR(ret) : table;
+	}
+
+	return ERR_PTR(-EINVAL);
+}
+
+void tegra_cvb_remove_opp_table(struct device *dev,
+				const struct cvb_table *table,
+				unsigned long max_freq)
+{
+	unsigned int i;
+
+	for (i = 0; i < MAX_DVFS_FREQS; i++) {
+		const struct cvb_table_freq_entry *entry = &table->entries[i];
+
+		if (!entry->freq || (entry->freq > max_freq))
+			break;
+
+		dev_pm_opp_remove(dev, entry->freq);
+	}
+}