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, 369 insertions, 0 deletions

diff --git a/lib/interval_tree_test.c b/lib/interval_tree_test.c
new file mode 100644
index 000000000..b0b07270c
--- /dev/null
+++ b/lib/interval_tree_test.c
@@ -0,0 +1,369 @@
+// SPDX-License-Identifier: GPL-2.0-only
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/interval_tree.h>
+#include <linux/prandom.h>
+#include <linux/slab.h>
+#include <linux/printk.h>
+#include <asm/timex.h>
+#include <linux/bitmap.h>
+#include <linux/maple_tree.h>
+
+#define __param(type, name, init, msg)		\
+	static type name = init;		\
+	module_param(name, type, 0444);		\
+	MODULE_PARM_DESC(name, msg);
+
+__param(int, nnodes, 100, "Number of nodes in the interval tree");
+__param(int, perf_loops, 1000, "Number of iterations modifying the tree");
+
+__param(int, nsearches, 100, "Number of searches to the interval tree");
+__param(int, search_loops, 1000, "Number of iterations searching the tree");
+__param(bool, search_all, false, "Searches will iterate all nodes in the tree");
+
+__param(uint, max_endpoint, ~0, "Largest value for the interval's endpoint");
+__param(ullong, seed, 3141592653589793238ULL, "Random seed");
+
+static struct rb_root_cached root = RB_ROOT_CACHED;
+static struct interval_tree_node *nodes = NULL;
+static u32 *queries = NULL;
+
+static struct rnd_state rnd;
+
+static inline unsigned long
+search(struct rb_root_cached *root, unsigned long start, unsigned long last)
+{
+	struct interval_tree_node *node;
+	unsigned long results = 0;
+
+	for (node = interval_tree_iter_first(root, start, last); node;
+	     node = interval_tree_iter_next(node, start, last))
+		results++;
+	return results;
+}
+
+static void init(void)
+{
+	int i;
+
+	for (i = 0; i < nnodes; i++) {
+		u32 b = (prandom_u32_state(&rnd) >> 4) % max_endpoint;
+		u32 a = (prandom_u32_state(&rnd) >> 4) % b;
+
+		nodes[i].start = a;
+		nodes[i].last = b;
+	}
+
+	/*
+	 * Limit the search scope to what the user defined.
+	 * Otherwise we are merely measuring empty walks,
+	 * which is pointless.
+	 */
+	for (i = 0; i < nsearches; i++)
+		queries[i] = (prandom_u32_state(&rnd) >> 4) % max_endpoint;
+}
+
+static int basic_check(void)
+{
+	int i, j;
+	cycles_t time1, time2, time;
+
+	printk(KERN_ALERT "interval tree insert/remove");
+
+	init();
+
+	time1 = get_cycles();
+
+	for (i = 0; i < perf_loops; i++) {
+		for (j = 0; j < nnodes; j++)
+			interval_tree_insert(nodes + j, &root);
+		for (j = 0; j < nnodes; j++)
+			interval_tree_remove(nodes + j, &root);
+	}
+
+	time2 = get_cycles();
+	time = time2 - time1;
+
+	time = div_u64(time, perf_loops);
+	printk(" -> %llu cycles\n", (unsigned long long)time);
+
+	return 0;
+}
+
+static int search_check(void)
+{
+	int i, j;
+	unsigned long results;
+	cycles_t time1, time2, time;
+
+	printk(KERN_ALERT "interval tree search");
+
+	init();
+
+	for (j = 0; j < nnodes; j++)
+		interval_tree_insert(nodes + j, &root);
+
+	time1 = get_cycles();
+
+	results = 0;
+	for (i = 0; i < search_loops; i++)
+		for (j = 0; j < nsearches; j++) {
+			unsigned long start = search_all ? 0 : queries[j];
+			unsigned long last = search_all ? max_endpoint : queries[j];
+
+			results += search(&root, start, last);
+		}
+
+	time2 = get_cycles();
+	time = time2 - time1;
+
+	time = div_u64(time, search_loops);
+	results = div_u64(results, search_loops);
+	printk(" -> %llu cycles (%lu results)\n",
+	       (unsigned long long)time, results);
+
+	for (j = 0; j < nnodes; j++)
+		interval_tree_remove(nodes + j, &root);
+
+	return 0;
+}
+
+static int intersection_range_check(void)
+{
+	int i, j, k;
+	unsigned long start, last;
+	struct interval_tree_node *node;
+	unsigned long *intxn1;
+	unsigned long *intxn2;
+
+	printk(KERN_ALERT "interval tree iteration\n");
+
+	intxn1 = bitmap_alloc(nnodes, GFP_KERNEL);
+	if (!intxn1) {
+		WARN_ONCE(1, "Failed to allocate intxn1\n");
+		return -ENOMEM;
+	}
+
+	intxn2 = bitmap_alloc(nnodes, GFP_KERNEL);
+	if (!intxn2) {
+		WARN_ONCE(1, "Failed to allocate intxn2\n");
+		bitmap_free(intxn1);
+		return -ENOMEM;
+	}
+
+	for (i = 0; i < search_loops; i++) {
+		/* Initialize interval tree for each round */
+		init();
+		for (j = 0; j < nnodes; j++)
+			interval_tree_insert(nodes + j, &root);
+
+		/* Let's try nsearches different ranges */
+		for (k = 0; k < nsearches; k++) {
+			/* Try whole range once */
+			if (!k) {
+				start = 0UL;
+				last = ULONG_MAX;
+			} else {
+				last = (prandom_u32_state(&rnd) >> 4) % max_endpoint;
+				start = (prandom_u32_state(&rnd) >> 4) % last;
+			}
+
+			/* Walk nodes to mark intersection nodes */
+			bitmap_zero(intxn1, nnodes);
+			for (j = 0; j < nnodes; j++) {
+				node = nodes + j;
+
+				if (start <= node->last && last >= node->start)
+					bitmap_set(intxn1, j, 1);
+			}
+
+			/* Iterate tree to clear intersection nodes */
+			bitmap_zero(intxn2, nnodes);
+			for (node = interval_tree_iter_first(&root, start, last); node;
+			     node = interval_tree_iter_next(node, start, last))
+				bitmap_set(intxn2, node - nodes, 1);
+
+			WARN_ON_ONCE(!bitmap_equal(intxn1, intxn2, nnodes));
+		}
+
+		for (j = 0; j < nnodes; j++)
+			interval_tree_remove(nodes + j, &root);
+	}
+
+	bitmap_free(intxn1);
+	bitmap_free(intxn2);
+	return 0;
+}
+
+#ifdef CONFIG_INTERVAL_TREE_SPAN_ITER
+/*
+ * Helper function to get span of current position from maple tree point of
+ * view.
+ */
+static void mas_cur_span(struct ma_state *mas, struct interval_tree_span_iter *state)
+{
+	unsigned long cur_start;
+	unsigned long cur_last;
+	int is_hole;
+
+	if (mas->status == ma_overflow)
+		return;
+
+	/* walk to current position */
+	state->is_hole = mas_walk(mas) ? 0 : 1;
+
+	cur_start = mas->index < state->first_index ?
+			state->first_index : mas->index;
+
+	/* whether we have followers */
+	do {
+
+		cur_last = mas->last > state->last_index ?
+				state->last_index : mas->last;
+
+		is_hole = mas_next_range(mas, state->last_index) ? 0 : 1;
+
+	} while (mas->status != ma_overflow && is_hole == state->is_hole);
+
+	if (state->is_hole) {
+		state->start_hole = cur_start;
+		state->last_hole = cur_last;
+	} else {
+		state->start_used = cur_start;
+		state->last_used = cur_last;
+	}
+
+	/* advance position for next round */
+	if (mas->status != ma_overflow)
+		mas_set(mas, cur_last + 1);
+}
+
+static int span_iteration_check(void)
+{
+	int i, j, k;
+	unsigned long start, last;
+	struct interval_tree_span_iter span, mas_span;
+
+	DEFINE_MTREE(tree);
+
+	MA_STATE(mas, &tree, 0, 0);
+
+	printk(KERN_ALERT "interval tree span iteration\n");
+
+	for (i = 0; i < search_loops; i++) {
+		/* Initialize interval tree for each round */
+		init();
+		for (j = 0; j < nnodes; j++)
+			interval_tree_insert(nodes + j, &root);
+
+		/* Put all the range into maple tree */
+		mt_init_flags(&tree, MT_FLAGS_ALLOC_RANGE);
+		mt_set_in_rcu(&tree);
+
+		for (j = 0; j < nnodes; j++)
+			WARN_ON_ONCE(mtree_store_range(&tree, nodes[j].start,
+					nodes[j].last, nodes + j, GFP_KERNEL));
+
+		/* Let's try nsearches different ranges */
+		for (k = 0; k < nsearches; k++) {
+			/* Try whole range once */
+			if (!k) {
+				start = 0UL;
+				last = ULONG_MAX;
+			} else {
+				last = (prandom_u32_state(&rnd) >> 4) % max_endpoint;
+				start = (prandom_u32_state(&rnd) >> 4) % last;
+			}
+
+			mas_span.first_index = start;
+			mas_span.last_index = last;
+			mas_span.is_hole = -1;
+			mas_set(&mas, start);
+
+			interval_tree_for_each_span(&span, &root, start, last) {
+				mas_cur_span(&mas, &mas_span);
+
+				WARN_ON_ONCE(span.is_hole != mas_span.is_hole);
+
+				if (span.is_hole) {
+					WARN_ON_ONCE(span.start_hole != mas_span.start_hole);
+					WARN_ON_ONCE(span.last_hole != mas_span.last_hole);
+				} else {
+					WARN_ON_ONCE(span.start_used != mas_span.start_used);
+					WARN_ON_ONCE(span.last_used != mas_span.last_used);
+				}
+			}
+
+		}
+
+		WARN_ON_ONCE(mas.status != ma_overflow);
+
+		/* Cleanup maple tree for each round */
+		mtree_destroy(&tree);
+		/* Cleanup interval tree for each round */
+		for (j = 0; j < nnodes; j++)
+			interval_tree_remove(nodes + j, &root);
+	}
+	return 0;
+}
+#else
+static inline int span_iteration_check(void) {return 0; }
+#endif
+
+static int interval_tree_test_init(void)
+{
+	if (nnodes <= 0) {
+		pr_warn("nnodes must be positive\n");
+		return -EINVAL;
+	}
+	if (nsearches <= 0) {
+		pr_warn("nsearches must be positive\n");
+		return -EINVAL;
+	}
+	if (perf_loops <= 0) {
+		pr_warn("perf_loops must be positive\n");
+		return -EINVAL;
+	}
+	if (search_loops <= 0) {
+		pr_warn("search_loops must be positive\n");
+		return -EINVAL;
+	}
+	if (max_endpoint < 2) {
+		pr_warn("max_endpoint must be at least 2\n");
+		return -EINVAL;
+	}
+
+	nodes = kmalloc_objs(struct interval_tree_node, nnodes);
+	if (!nodes)
+		return -ENOMEM;
+
+	queries = kmalloc_array(nsearches, sizeof(int), GFP_KERNEL);
+	if (!queries) {
+		kfree(nodes);
+		return -ENOMEM;
+	}
+
+	prandom_seed_state(&rnd, seed);
+
+	basic_check();
+	search_check();
+	intersection_range_check();
+	span_iteration_check();
+
+	kfree(queries);
+	kfree(nodes);
+
+	return -EAGAIN; /* Fail will directly unload the module */
+}
+
+static void interval_tree_test_exit(void)
+{
+	printk(KERN_ALERT "test exit\n");
+}
+
+module_init(interval_tree_test_init)
+module_exit(interval_tree_test_exit)
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Michel Lespinasse");
+MODULE_DESCRIPTION("Interval Tree test");