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, 324 insertions, 0 deletions

diff --git a/drivers/interconnect/icc-kunit.c b/drivers/interconnect/icc-kunit.c
new file mode 100644
index 000000000..bad2b5837
--- /dev/null
+++ b/drivers/interconnect/icc-kunit.c
@@ -0,0 +1,324 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * KUnit tests for the Interconnect framework.
+ *
+ * Copyright (c) 2025 Kuan-Wei Chiu <visitorckw@gmail.com>
+ *
+ * This suite verifies the behavior of the interconnect core, including
+ * topology construction, bandwidth aggregation, and path lifecycle.
+ */
+
+#include <kunit/platform_device.h>
+#include <kunit/test.h>
+#include <linux/interconnect-provider.h>
+#include <linux/interconnect.h>
+#include <linux/list.h>
+#include <linux/module.h>
+#include <linux/overflow.h>
+#include <linux/platform_device.h>
+#include <linux/slab.h>
+
+#include "internal.h"
+
+enum {
+	NODE_CPU,
+	NODE_GPU,
+	NODE_BUS,
+	NODE_DDR,
+	NODE_MAX
+};
+
+struct test_node_data {
+	int id;
+	const char *name;
+	int num_links;
+	int links[2];
+};
+
+/*
+ * Static Topology:
+ * CPU -\
+ * -> BUS -> DDR
+ * GPU -/
+ */
+static const struct test_node_data test_topology[] = {
+	{ NODE_CPU, "cpu", 1, { NODE_BUS } },
+	{ NODE_GPU, "gpu", 1, { NODE_BUS } },
+	{ NODE_BUS, "bus", 1, { NODE_DDR } },
+	{ NODE_DDR, "ddr", 0, { } },
+};
+
+struct icc_test_priv {
+	struct icc_provider provider;
+	struct platform_device *pdev;
+	struct icc_node *nodes[NODE_MAX];
+};
+
+static struct icc_node *get_node(struct icc_test_priv *priv, int id)
+{
+	int idx = id - NODE_CPU;
+
+	if (idx < 0 || idx >= ARRAY_SIZE(test_topology))
+		return NULL;
+	return priv->nodes[idx];
+}
+
+static int icc_test_set(struct icc_node *src, struct icc_node *dst)
+{
+	return 0;
+}
+
+static int icc_test_aggregate(struct icc_node *node, u32 tag, u32 avg_bw,
+			      u32 peak_bw, u32 *agg_avg, u32 *agg_peak)
+{
+	return icc_std_aggregate(node, tag, avg_bw, peak_bw, agg_avg, agg_peak);
+}
+
+static struct icc_node *icc_test_xlate(const struct of_phandle_args *spec, void *data)
+{
+	return NULL;
+}
+
+static int icc_test_get_bw(struct icc_node *node, u32 *avg, u32 *peak)
+{
+	*avg = 0;
+	*peak = 0;
+
+	return 0;
+}
+
+static int icc_test_init(struct kunit *test)
+{
+	struct icc_test_priv *priv;
+	struct icc_node *node;
+	int i, j, ret;
+
+	priv = kunit_kzalloc(test, sizeof(*priv), GFP_KERNEL);
+	KUNIT_ASSERT_NOT_ERR_OR_NULL(test, priv);
+	test->priv = priv;
+
+	priv->pdev = kunit_platform_device_alloc(test, "icc-test-dev", -1);
+	KUNIT_ASSERT_NOT_ERR_OR_NULL(test, priv->pdev);
+	KUNIT_ASSERT_EQ(test, kunit_platform_device_add(test, priv->pdev), 0);
+
+	priv->provider.set = icc_test_set;
+	priv->provider.aggregate = icc_test_aggregate;
+	priv->provider.xlate = icc_test_xlate;
+	priv->provider.get_bw = icc_test_get_bw;
+	priv->provider.dev = &priv->pdev->dev;
+	priv->provider.data = priv;
+	INIT_LIST_HEAD(&priv->provider.nodes);
+
+	ret = icc_provider_register(&priv->provider);
+	KUNIT_ASSERT_EQ(test, ret, 0);
+
+	for (i = 0; i < ARRAY_SIZE(test_topology); i++) {
+		const struct test_node_data *data = &test_topology[i];
+
+		node = icc_node_create(data->id);
+		KUNIT_ASSERT_NOT_ERR_OR_NULL(test, node);
+
+		node->name = data->name;
+		icc_node_add(node, &priv->provider);
+		priv->nodes[i] = node;
+	}
+
+	for (i = 0; i < ARRAY_SIZE(test_topology); i++) {
+		const struct test_node_data *data = &test_topology[i];
+		struct icc_node *src = get_node(priv, data->id);
+
+		for (j = 0; j < data->num_links; j++) {
+			ret = icc_link_create(src, data->links[j]);
+			KUNIT_ASSERT_EQ_MSG(test, ret, 0, "Failed to link %s->%d",
+					    src->name, data->links[j]);
+		}
+	}
+
+	icc_sync_state(&priv->pdev->dev);
+
+	return 0;
+}
+
+static void icc_test_exit(struct kunit *test)
+{
+	struct icc_test_priv *priv = test->priv;
+
+	icc_nodes_remove(&priv->provider);
+	icc_provider_deregister(&priv->provider);
+}
+
+/*
+ * Helper to construct a mock path.
+ *
+ * Because we are bypassing icc_get(), we must manually link the requests
+ * to the nodes' req_list so that icc_std_aggregate() can discover them.
+ */
+static struct icc_path *icc_test_create_path(struct kunit *test,
+					     struct icc_node **nodes, int num)
+{
+	struct icc_path *path;
+	int i;
+
+	path = kunit_kzalloc(test, struct_size(path, reqs, num), GFP_KERNEL);
+	KUNIT_ASSERT_NOT_ERR_OR_NULL(test, path);
+
+	path->num_nodes = num;
+	for (i = 0; i < num; i++) {
+		path->reqs[i].node = nodes[i];
+		hlist_add_head(&path->reqs[i].req_node, &nodes[i]->req_list);
+	}
+	path->name = "mock-path";
+
+	return path;
+}
+
+static void icc_test_destroy_path(struct kunit *test, struct icc_path *path)
+{
+	int i;
+
+	for (i = 0; i < path->num_nodes; i++)
+		hlist_del(&path->reqs[i].req_node);
+
+	kunit_kfree(test, path);
+}
+
+static void icc_test_topology_integrity(struct kunit *test)
+{
+	struct icc_test_priv *priv = test->priv;
+	struct icc_node *cpu = get_node(priv, NODE_CPU);
+	struct icc_node *bus = get_node(priv, NODE_BUS);
+
+	KUNIT_EXPECT_EQ(test, cpu->num_links, 1);
+	KUNIT_EXPECT_PTR_EQ(test, cpu->links[0], bus);
+	KUNIT_EXPECT_PTR_EQ(test, cpu->provider, &priv->provider);
+}
+
+static void icc_test_set_bw(struct kunit *test)
+{
+	struct icc_test_priv *priv = test->priv;
+	struct icc_path *path;
+	struct icc_node *path_nodes[3];
+	int ret;
+
+	/* Path: CPU -> BUS -> DDR */
+	path_nodes[0] = get_node(priv, NODE_CPU);
+	path_nodes[1] = get_node(priv, NODE_BUS);
+	path_nodes[2] = get_node(priv, NODE_DDR);
+
+	path = icc_test_create_path(test, path_nodes, 3);
+
+	ret = icc_enable(path);
+	KUNIT_ASSERT_EQ(test, ret, 0);
+
+	ret = icc_set_bw(path, 1000, 2000);
+	KUNIT_EXPECT_EQ(test, ret, 0);
+
+	KUNIT_EXPECT_EQ(test, path_nodes[0]->avg_bw, 1000);
+	KUNIT_EXPECT_EQ(test, path_nodes[0]->peak_bw, 2000);
+	KUNIT_EXPECT_EQ(test, path_nodes[1]->avg_bw, 1000);
+	KUNIT_EXPECT_EQ(test, path_nodes[1]->peak_bw, 2000);
+
+	icc_set_tag(path, 0xABC);
+	KUNIT_EXPECT_EQ(test, path->reqs[0].tag, 0xABC);
+
+	icc_disable(path);
+	KUNIT_EXPECT_EQ(test, path_nodes[0]->avg_bw, 0);
+
+	icc_test_destroy_path(test, path);
+}
+
+static void icc_test_aggregation(struct kunit *test)
+{
+	struct icc_test_priv *priv = test->priv;
+	struct icc_path *path_cpu, *path_gpu;
+	struct icc_node *nodes_cpu[3], *nodes_gpu[2];
+	struct icc_node *bus = get_node(priv, NODE_BUS);
+	int ret;
+
+	nodes_cpu[0] = get_node(priv, NODE_CPU);
+	nodes_cpu[1] = bus;
+	nodes_cpu[2] = get_node(priv, NODE_DDR);
+	path_cpu = icc_test_create_path(test, nodes_cpu, 3);
+
+	nodes_gpu[0] = get_node(priv, NODE_GPU);
+	nodes_gpu[1] = bus;
+	path_gpu = icc_test_create_path(test, nodes_gpu, 2);
+
+	icc_enable(path_cpu);
+	icc_enable(path_gpu);
+
+	ret = icc_set_bw(path_cpu, 1000, 1000);
+	KUNIT_EXPECT_EQ(test, ret, 0);
+	KUNIT_EXPECT_EQ(test, bus->avg_bw, 1000);
+
+	ret = icc_set_bw(path_gpu, 2000, 2000);
+	KUNIT_EXPECT_EQ(test, ret, 0);
+
+	/* Bus aggregates: CPU(1000) + GPU(2000) */
+	KUNIT_EXPECT_EQ(test, bus->avg_bw, 3000);
+	/* Peak aggregates: max(CPU, GPU) */
+	KUNIT_EXPECT_EQ(test, bus->peak_bw, 2000);
+
+	icc_test_destroy_path(test, path_cpu);
+	icc_test_destroy_path(test, path_gpu);
+}
+
+static void icc_test_bulk_ops(struct kunit *test)
+{
+	struct icc_test_priv *priv = test->priv;
+	struct icc_node *nodes_cpu[3], *nodes_gpu[2];
+	struct icc_bulk_data bulk[2];
+	int ret;
+
+	nodes_cpu[0] = get_node(priv, NODE_CPU);
+	nodes_cpu[1] = get_node(priv, NODE_BUS);
+	nodes_cpu[2] = get_node(priv, NODE_DDR);
+
+	nodes_gpu[0] = get_node(priv, NODE_GPU);
+	nodes_gpu[1] = get_node(priv, NODE_BUS);
+
+	bulk[0].path = icc_test_create_path(test, nodes_cpu, 3);
+	bulk[0].avg_bw = 500;
+	bulk[0].peak_bw = 500;
+
+	bulk[1].path = icc_test_create_path(test, nodes_gpu, 2);
+	bulk[1].avg_bw = 600;
+	bulk[1].peak_bw = 600;
+
+	ret = icc_bulk_set_bw(2, bulk);
+	KUNIT_EXPECT_EQ(test, ret, 0);
+	/* Paths disabled, bandwidth should be 0 */
+	KUNIT_EXPECT_EQ(test, get_node(priv, NODE_BUS)->avg_bw, 0);
+
+	ret = icc_bulk_enable(2, bulk);
+	KUNIT_EXPECT_EQ(test, ret, 0);
+	/* Paths enabled, aggregation applies */
+	KUNIT_EXPECT_EQ(test, get_node(priv, NODE_BUS)->avg_bw, 1100);
+
+	icc_bulk_disable(2, bulk);
+	KUNIT_EXPECT_EQ(test, get_node(priv, NODE_BUS)->avg_bw, 0);
+
+	icc_test_destroy_path(test, bulk[0].path);
+	icc_test_destroy_path(test, bulk[1].path);
+}
+
+static struct kunit_case icc_test_cases[] = {
+	KUNIT_CASE(icc_test_topology_integrity),
+	KUNIT_CASE(icc_test_set_bw),
+	KUNIT_CASE(icc_test_aggregation),
+	KUNIT_CASE(icc_test_bulk_ops),
+	{}
+};
+
+static struct kunit_suite icc_test_suite = {
+	.name = "interconnect",
+	.init = icc_test_init,
+	.exit = icc_test_exit,
+	.test_cases = icc_test_cases,
+};
+
+kunit_test_suite(icc_test_suite);
+
+MODULE_AUTHOR("Kuan-Wei Chiu <visitorckw@gmail.com>");
+MODULE_DESCRIPTION("KUnit tests for the Interconnect framework");
+MODULE_LICENSE("GPL");