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, 288 insertions, 0 deletions

diff --git a/lib/test_ida.c b/lib/test_ida.c
new file mode 100644
index 000000000..63078f8dc
--- /dev/null
+++ b/lib/test_ida.c
@@ -0,0 +1,288 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * test_ida.c: Test the IDA API
+ * Copyright (c) 2016-2018 Microsoft Corporation
+ * Copyright (c) 2018 Oracle Corporation
+ * Author: Matthew Wilcox <willy@infradead.org>
+ */
+
+#include <linux/idr.h>
+#include <linux/module.h>
+
+static unsigned int tests_run;
+static unsigned int tests_passed;
+
+#ifdef __KERNEL__
+static void ida_dump(struct ida *ida) { }
+#endif
+#define IDA_BUG_ON(ida, x) do {						\
+	tests_run++;							\
+	if (x) {							\
+		ida_dump(ida);						\
+		dump_stack();						\
+	} else {							\
+		tests_passed++;						\
+	}								\
+} while (0)
+
+/*
+ * Straightforward checks that allocating and freeing IDs work.
+ */
+static void ida_check_alloc(struct ida *ida)
+{
+	int i, id;
+
+	for (i = 0; i < 10000; i++)
+		IDA_BUG_ON(ida, ida_alloc(ida, GFP_KERNEL) != i);
+
+	ida_free(ida, 20);
+	ida_free(ida, 21);
+	for (i = 0; i < 3; i++) {
+		id = ida_alloc(ida, GFP_KERNEL);
+		IDA_BUG_ON(ida, id < 0);
+		if (i == 2)
+			IDA_BUG_ON(ida, id != 10000);
+	}
+
+	for (i = 0; i < 5000; i++)
+		ida_free(ida, i);
+
+	IDA_BUG_ON(ida, ida_alloc_min(ida, 5000, GFP_KERNEL) != 10001);
+	ida_destroy(ida);
+
+	IDA_BUG_ON(ida, !ida_is_empty(ida));
+}
+
+/* Destroy an IDA with a single entry at @base */
+static void ida_check_destroy_1(struct ida *ida, unsigned int base)
+{
+	IDA_BUG_ON(ida, ida_alloc_min(ida, base, GFP_KERNEL) != base);
+	IDA_BUG_ON(ida, ida_is_empty(ida));
+	ida_destroy(ida);
+	IDA_BUG_ON(ida, !ida_is_empty(ida));
+}
+
+/* Check that ida_destroy and ida_is_empty work */
+static void ida_check_destroy(struct ida *ida)
+{
+	/* Destroy an already-empty IDA */
+	IDA_BUG_ON(ida, !ida_is_empty(ida));
+	ida_destroy(ida);
+	IDA_BUG_ON(ida, !ida_is_empty(ida));
+
+	ida_check_destroy_1(ida, 0);
+	ida_check_destroy_1(ida, 1);
+	ida_check_destroy_1(ida, 1023);
+	ida_check_destroy_1(ida, 1024);
+	ida_check_destroy_1(ida, 12345678);
+}
+
+/*
+ * Check what happens when we fill a leaf and then delete it.  This may
+ * discover mishandling of IDR_FREE.
+ */
+static void ida_check_leaf(struct ida *ida, unsigned int base)
+{
+	unsigned long i;
+
+	for (i = 0; i < IDA_BITMAP_BITS; i++) {
+		IDA_BUG_ON(ida, ida_alloc_min(ida, base, GFP_KERNEL) !=
+				base + i);
+	}
+
+	ida_destroy(ida);
+	IDA_BUG_ON(ida, !ida_is_empty(ida));
+
+	IDA_BUG_ON(ida, ida_alloc(ida, GFP_KERNEL) != 0);
+	IDA_BUG_ON(ida, ida_is_empty(ida));
+	ida_free(ida, 0);
+	IDA_BUG_ON(ida, !ida_is_empty(ida));
+}
+
+/*
+ * Check allocations up to and slightly above the maximum allowed (2^31-1) ID.
+ * Allocating up to 2^31-1 should succeed, and then allocating the next one
+ * should fail.
+ */
+static void ida_check_max(struct ida *ida)
+{
+	unsigned long i, j;
+
+	for (j = 1; j < 65537; j *= 2) {
+		unsigned long base = (1UL << 31) - j;
+		for (i = 0; i < j; i++) {
+			IDA_BUG_ON(ida, ida_alloc_min(ida, base, GFP_KERNEL) !=
+					base + i);
+		}
+		IDA_BUG_ON(ida, ida_alloc_min(ida, base, GFP_KERNEL) !=
+				-ENOSPC);
+		ida_destroy(ida);
+		IDA_BUG_ON(ida, !ida_is_empty(ida));
+	}
+}
+
+/*
+ * Check handling of conversions between exceptional entries and full bitmaps.
+ */
+static void ida_check_conv(struct ida *ida)
+{
+	unsigned long i;
+
+	for (i = 0; i < IDA_BITMAP_BITS * 2; i += IDA_BITMAP_BITS) {
+		IDA_BUG_ON(ida, ida_alloc_min(ida, i + 1, GFP_KERNEL) != i + 1);
+		IDA_BUG_ON(ida, ida_alloc_min(ida, i + BITS_PER_LONG,
+					GFP_KERNEL) != i + BITS_PER_LONG);
+		ida_free(ida, i + 1);
+		ida_free(ida, i + BITS_PER_LONG);
+		IDA_BUG_ON(ida, !ida_is_empty(ida));
+	}
+
+	for (i = 0; i < IDA_BITMAP_BITS * 2; i++)
+		IDA_BUG_ON(ida, ida_alloc(ida, GFP_KERNEL) != i);
+	for (i = IDA_BITMAP_BITS * 2; i > 0; i--)
+		ida_free(ida, i - 1);
+	IDA_BUG_ON(ida, !ida_is_empty(ida));
+
+	for (i = 0; i < IDA_BITMAP_BITS + BITS_PER_LONG - 4; i++)
+		IDA_BUG_ON(ida, ida_alloc(ida, GFP_KERNEL) != i);
+	for (i = IDA_BITMAP_BITS + BITS_PER_LONG - 4; i > 0; i--)
+		ida_free(ida, i - 1);
+	IDA_BUG_ON(ida, !ida_is_empty(ida));
+}
+
+/*
+ * Check various situations where we attempt to free an ID we don't own.
+ */
+static void ida_check_bad_free(struct ida *ida)
+{
+	unsigned long i;
+
+	printk("vvv Ignore \"not allocated\" warnings\n");
+	/* IDA is empty; all of these will fail */
+	ida_free(ida, 0);
+	for (i = 0; i < 31; i++)
+		ida_free(ida, 1 << i);
+
+	/* IDA contains a single value entry */
+	IDA_BUG_ON(ida, ida_alloc_min(ida, 3, GFP_KERNEL) != 3);
+	ida_free(ida, 0);
+	for (i = 0; i < 31; i++)
+		ida_free(ida, 1 << i);
+
+	/* IDA contains a single bitmap */
+	IDA_BUG_ON(ida, ida_alloc_min(ida, 1023, GFP_KERNEL) != 1023);
+	ida_free(ida, 0);
+	for (i = 0; i < 31; i++)
+		ida_free(ida, 1 << i);
+
+	/* IDA contains a tree */
+	IDA_BUG_ON(ida, ida_alloc_min(ida, (1 << 20) - 1, GFP_KERNEL) != (1 << 20) - 1);
+	ida_free(ida, 0);
+	for (i = 0; i < 31; i++)
+		ida_free(ida, 1 << i);
+	printk("^^^ \"not allocated\" warnings over\n");
+
+	ida_free(ida, 3);
+	ida_free(ida, 1023);
+	ida_free(ida, (1 << 20) - 1);
+
+	IDA_BUG_ON(ida, !ida_is_empty(ida));
+}
+
+/*
+ * Check ida_find_first_range() and varriants.
+ */
+static void ida_check_find_first(struct ida *ida)
+{
+	/* IDA is empty; all of the below should be not exist */
+	IDA_BUG_ON(ida, ida_exists(ida, 0));
+	IDA_BUG_ON(ida, ida_exists(ida, 3));
+	IDA_BUG_ON(ida, ida_exists(ida, 63));
+	IDA_BUG_ON(ida, ida_exists(ida, 1023));
+	IDA_BUG_ON(ida, ida_exists(ida, (1 << 20) - 1));
+
+	/* IDA contains a single value entry */
+	IDA_BUG_ON(ida, ida_alloc_min(ida, 3, GFP_KERNEL) != 3);
+	IDA_BUG_ON(ida, ida_exists(ida, 0));
+	IDA_BUG_ON(ida, !ida_exists(ida, 3));
+	IDA_BUG_ON(ida, ida_exists(ida, 63));
+	IDA_BUG_ON(ida, ida_exists(ida, 1023));
+	IDA_BUG_ON(ida, ida_exists(ida, (1 << 20) - 1));
+
+	IDA_BUG_ON(ida, ida_alloc_min(ida, 63, GFP_KERNEL) != 63);
+	IDA_BUG_ON(ida, ida_exists(ida, 0));
+	IDA_BUG_ON(ida, !ida_exists(ida, 3));
+	IDA_BUG_ON(ida, !ida_exists(ida, 63));
+	IDA_BUG_ON(ida, ida_exists(ida, 1023));
+	IDA_BUG_ON(ida, ida_exists(ida, (1 << 20) - 1));
+
+	/* IDA contains a single bitmap */
+	IDA_BUG_ON(ida, ida_alloc_min(ida, 1023, GFP_KERNEL) != 1023);
+	IDA_BUG_ON(ida, ida_exists(ida, 0));
+	IDA_BUG_ON(ida, !ida_exists(ida, 3));
+	IDA_BUG_ON(ida, !ida_exists(ida, 63));
+	IDA_BUG_ON(ida, !ida_exists(ida, 1023));
+	IDA_BUG_ON(ida, ida_exists(ida, (1 << 20) - 1));
+
+	/* IDA contains a tree */
+	IDA_BUG_ON(ida, ida_alloc_min(ida, (1 << 20) - 1, GFP_KERNEL) != (1 << 20) - 1);
+	IDA_BUG_ON(ida, ida_exists(ida, 0));
+	IDA_BUG_ON(ida, !ida_exists(ida, 3));
+	IDA_BUG_ON(ida, !ida_exists(ida, 63));
+	IDA_BUG_ON(ida, !ida_exists(ida, 1023));
+	IDA_BUG_ON(ida, !ida_exists(ida, (1 << 20) - 1));
+
+	/* Now try to find first */
+	IDA_BUG_ON(ida, ida_find_first(ida) != 3);
+	IDA_BUG_ON(ida, ida_find_first_range(ida, -1, 2) != -EINVAL);
+	IDA_BUG_ON(ida, ida_find_first_range(ida, 0, 2) != -ENOENT); // no used ID
+	IDA_BUG_ON(ida, ida_find_first_range(ida, 0, 3) != 3);
+	IDA_BUG_ON(ida, ida_find_first_range(ida, 1, 3) != 3);
+	IDA_BUG_ON(ida, ida_find_first_range(ida, 3, 3) != 3);
+	IDA_BUG_ON(ida, ida_find_first_range(ida, 2, 4) != 3);
+	IDA_BUG_ON(ida, ida_find_first_range(ida, 4, 3) != -ENOENT); // min > max, fail
+	IDA_BUG_ON(ida, ida_find_first_range(ida, 4, 60) != -ENOENT); // no used ID
+	IDA_BUG_ON(ida, ida_find_first_range(ida, 4, 64) != 63);
+	IDA_BUG_ON(ida, ida_find_first_range(ida, 63, 63) != 63);
+	IDA_BUG_ON(ida, ida_find_first_range(ida, 64, 1026) != 1023);
+	IDA_BUG_ON(ida, ida_find_first_range(ida, 1023, 1023) != 1023);
+	IDA_BUG_ON(ida, ida_find_first_range(ida, 1023, (1 << 20) - 1) != 1023);
+	IDA_BUG_ON(ida, ida_find_first_range(ida, 1024, (1 << 20) - 1) != (1 << 20) - 1);
+	IDA_BUG_ON(ida, ida_find_first_range(ida, (1 << 20), INT_MAX) != -ENOENT);
+
+	ida_free(ida, 3);
+	ida_free(ida, 63);
+	ida_free(ida, 1023);
+	ida_free(ida, (1 << 20) - 1);
+
+	IDA_BUG_ON(ida, !ida_is_empty(ida));
+}
+
+static DEFINE_IDA(ida);
+
+static int ida_checks(void)
+{
+	IDA_BUG_ON(&ida, !ida_is_empty(&ida));
+	ida_check_alloc(&ida);
+	ida_check_destroy(&ida);
+	ida_check_leaf(&ida, 0);
+	ida_check_leaf(&ida, 1024);
+	ida_check_leaf(&ida, 1024 * 64);
+	ida_check_max(&ida);
+	ida_check_conv(&ida);
+	ida_check_bad_free(&ida);
+	ida_check_find_first(&ida);
+
+	printk("IDA: %u of %u tests passed\n", tests_passed, tests_run);
+	return (tests_run != tests_passed) ? 0 : -EINVAL;
+}
+
+static void ida_exit(void)
+{
+}
+
+module_init(ida_checks);
+module_exit(ida_exit);
+MODULE_AUTHOR("Matthew Wilcox <willy@infradead.org>");
+MODULE_DESCRIPTION("Test the IDA API");
+MODULE_LICENSE("GPL");