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, 105 insertions, 0 deletions

diff --git a/lib/math/polynomial.c b/lib/math/polynomial.c
new file mode 100644
index 000000000..f26677cfe
--- /dev/null
+++ b/lib/math/polynomial.c
@@ -0,0 +1,105 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Generic polynomial calculation using integer coefficients.
+ *
+ * Copyright (C) 2020 BAIKAL ELECTRONICS, JSC
+ *
+ * Authors:
+ *   Maxim Kaurkin <maxim.kaurkin@baikalelectronics.ru>
+ *   Serge Semin <Sergey.Semin@baikalelectronics.ru>
+ *
+ */
+
+#include <linux/export.h>
+#include <linux/math.h>
+#include <linux/module.h>
+#include <linux/polynomial.h>
+
+/*
+ * The following conversion is an example:
+ *
+ * The original translation formulae of the temperature (in degrees of Celsius)
+ * to PVT data and vice-versa are following:
+ *
+ * N = 1.8322e-8*(T^4) + 2.343e-5*(T^3) + 8.7018e-3*(T^2) + 3.9269*(T^1) + 1.7204e2
+ * T = -1.6743e-11*(N^4) + 8.1542e-8*(N^3) + -1.8201e-4*(N^2) + 3.1020e-1*(N^1) - 4.838e1
+ *
+ * where T = [-48.380, 147.438]C and N = [0, 1023].
+ *
+ * They must be accordingly altered to be suitable for the integer arithmetics.
+ * The technique is called 'factor redistribution', which just makes sure the
+ * multiplications and divisions are made so to have a result of the operations
+ * within the integer numbers limit. In addition we need to translate the
+ * formulae to accept millidegrees of Celsius. Here what they look like after
+ * the alterations:
+ *
+ * N = (18322e-20*(T^4) + 2343e-13*(T^3) + 87018e-9*(T^2) + 39269e-3*T + 17204e2) / 1e4
+ * T = -16743e-12*(D^4) + 81542e-9*(D^3) - 182010e-6*(D^2) + 310200e-3*D - 48380
+ *
+ * where T = [-48380, 147438] mC and N = [0, 1023].
+ *
+ * static const struct polynomial poly_temp_to_N = {
+ *         .total_divider = 10000,
+ *         .terms = {
+ *                 {4, 18322, 10000, 10000},
+ *                 {3, 2343, 10000, 10},
+ *                 {2, 87018, 10000, 10},
+ *                 {1, 39269, 1000, 1},
+ *                 {0, 1720400, 1, 1}
+ *         }
+ * };
+ *
+ * static const struct polynomial poly_N_to_temp = {
+ *         .total_divider = 1,
+ *         .terms = {
+ *                 {4, -16743, 1000, 1},
+ *                 {3, 81542, 1000, 1},
+ *                 {2, -182010, 1000, 1},
+ *                 {1, 310200, 1000, 1},
+ *                 {0, -48380, 1, 1}
+ *         }
+ * };
+ */
+
+/**
+ * polynomial_calc - calculate a polynomial using integer arithmetic
+ *
+ * @poly: pointer to the descriptor of the polynomial
+ * @data: input value of the polynomial
+ *
+ * Calculate the result of a polynomial using only integer arithmetic. For
+ * this to work without too much loss of precision the coefficients has to
+ * be altered. This is called factor redistribution.
+ *
+ * Return: the result of the polynomial calculation.
+ */
+long polynomial_calc(const struct polynomial *poly, long data)
+{
+	const struct polynomial_term *term = poly->terms;
+	long total_divider = poly->total_divider ?: 1;
+	long tmp, ret = 0;
+	int deg;
+
+	/*
+	 * Here is the polynomial calculation function, which performs the
+	 * redistributed terms calculations. It's pretty straightforward.
+	 * We walk over each degree term up to the free one, and perform
+	 * the redistributed multiplication of the term coefficient, its
+	 * divider (as for the rationale fraction representation), data
+	 * power and the rational fraction divider leftover. Then all of
+	 * this is collected in a total sum variable, which value is
+	 * normalized by the total divider before being returned.
+	 */
+	do {
+		tmp = term->coef;
+		for (deg = 0; deg < term->deg; ++deg)
+			tmp = mult_frac(tmp, data, term->divider);
+		ret += tmp / term->divider_leftover;
+	} while ((term++)->deg);
+
+	return ret / total_divider;
+}
+EXPORT_SYMBOL_GPL(polynomial_calc);
+
+MODULE_DESCRIPTION("Generic polynomial calculations");
+MODULE_LICENSE("GPL");