From a5f7a5bb3b7f28ba7e4fa246775b29a0e5537255 Mon Sep 17 00:00:00 2001 From: Jinke Han Date: Tue, 8 Sep 2026 15:37:42 +0800 Subject: x86/kprobes: Fix crash when probing CS CALL instructions 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) Signed-off-by: Jinke Han Signed-off-by: Ingo Molnar Reviewed-by: Masami Hiramatsu (Google) Acked-by: Yafang Shao Acked-by: Borislav Petkov Cc: Peter Zijlstra Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000 --- lib/find_bit_benchmark.c | 218 +++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 218 insertions(+) create mode 100644 lib/find_bit_benchmark.c (limited to 'lib/find_bit_benchmark.c') diff --git a/lib/find_bit_benchmark.c b/lib/find_bit_benchmark.c new file mode 100644 index 000000000..05305e655 --- /dev/null +++ b/lib/find_bit_benchmark.c @@ -0,0 +1,218 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * Test for find_*_bit functions. + * + * Copyright (c) 2017 Cavium. + */ + +/* + * find_bit functions are widely used in kernel, so the successful boot + * is good enough test for correctness. + * + * This test is focused on performance of traversing bitmaps. Two typical + * scenarios are reproduced: + * - randomly filled bitmap with approximately equal number of set and + * cleared bits; + * - sparse bitmap with few set bits at random positions. + */ + +#include +#include +#include +#include +#include +#include + +#define BITMAP_LEN (4096UL * 8 * 10) +#define SPARSE 500 + +static DECLARE_BITMAP(bitmap, BITMAP_LEN) __initdata; +static DECLARE_BITMAP(bitmap2, BITMAP_LEN) __initdata; + +/* + * This is Schlemiel the Painter's algorithm. + */ +static int __init test_find_first_bit(const void *bitmap, unsigned long len) +{ + static DECLARE_BITMAP(cp, BITMAP_LEN) __initdata; + unsigned long i, cnt; + ktime_t time; + + bitmap_copy(cp, bitmap, BITMAP_LEN); + + time = ktime_get(); + for (cnt = i = 0; i < len; cnt++) { + i = find_first_bit(cp, len); + __clear_bit(i, cp); + } + time = ktime_get() - time; + pr_err("find_first_bit: %18llu ns, %6ld iterations\n", time, cnt); + + return 0; +} + +static int __init test_find_first_and_bit(const void *bitmap, const void *bitmap2, + unsigned long len) +{ + static DECLARE_BITMAP(cp, BITMAP_LEN) __initdata; + unsigned long i, cnt; + ktime_t time; + + bitmap_copy(cp, bitmap, BITMAP_LEN); + + time = ktime_get(); + for (cnt = i = 0; i < len; cnt++) { + i = find_first_and_bit(cp, bitmap2, len); + __clear_bit(i, cp); + } + time = ktime_get() - time; + pr_err("find_first_and_bit: %18llu ns, %6ld iterations\n", time, cnt); + + return 0; +} + +static int __init test_find_next_bit(const void *bitmap, unsigned long len) +{ + unsigned long i, cnt; + ktime_t time; + + time = ktime_get(); + for (cnt = i = 0; i < BITMAP_LEN; cnt++) + i = find_next_bit(bitmap, BITMAP_LEN, i) + 1; + time = ktime_get() - time; + pr_err("find_next_bit: %18llu ns, %6ld iterations\n", time, cnt); + + return 0; +} + +static int __init test_find_next_zero_bit(const void *bitmap, unsigned long len) +{ + unsigned long i, cnt; + ktime_t time; + + time = ktime_get(); + for (cnt = i = 0; i < BITMAP_LEN; cnt++) + i = find_next_zero_bit(bitmap, len, i) + 1; + time = ktime_get() - time; + pr_err("find_next_zero_bit: %18llu ns, %6ld iterations\n", time, cnt); + + return 0; +} + +static int __init test_find_last_bit(const void *bitmap, unsigned long len) +{ + unsigned long l, cnt = 0; + ktime_t time; + + time = ktime_get(); + do { + cnt++; + l = find_last_bit(bitmap, len); + if (l >= len) + break; + len = l; + } while (len); + time = ktime_get() - time; + pr_err("find_last_bit: %18llu ns, %6ld iterations\n", time, cnt); + + return 0; +} + +static int __init test_find_nth_bit(const unsigned long *bitmap, unsigned long len) +{ + unsigned long l, n, w = bitmap_weight(bitmap, len); + ktime_t time; + + time = ktime_get(); + for (n = 0; n < w; n++) { + l = find_nth_bit(bitmap, len, n); + WARN_ON(l >= len); + } + time = ktime_get() - time; + pr_err("find_nth_bit: %18llu ns, %6ld iterations\n", time, w); + + return 0; +} + +static int __init test_find_next_and_bit(const void *bitmap, + const void *bitmap2, unsigned long len) +{ + unsigned long i, cnt; + ktime_t time; + + time = ktime_get(); + for (cnt = i = 0; i < BITMAP_LEN; cnt++) + i = find_next_and_bit(bitmap, bitmap2, BITMAP_LEN, i + 1); + time = ktime_get() - time; + pr_err("find_next_and_bit: %18llu ns, %6ld iterations\n", time, cnt); + + return 0; +} + +static int __init +test_bitmap_find_next_zero_area_off(unsigned long *bitmap, unsigned long len) +{ + unsigned long i, cnt; + ktime_t time; + + time = ktime_get(); + for (cnt = i = 0; i < BITMAP_LEN; cnt++) + i = bitmap_find_next_zero_area_off(bitmap, BITMAP_LEN, i, 8, 0, 0) + 1; + time = ktime_get() - time; + pr_err("bitmap_find_next_zero_area_off:%7llu ns, %6ld iterations\n", time, cnt); + + return 0; +} + +static int __init find_bit_test(void) +{ + unsigned long nbits = BITMAP_LEN / SPARSE; + + pr_err("\nStart testing find_bit() with random-filled bitmap\n"); + + get_random_bytes(bitmap, sizeof(bitmap)); + get_random_bytes(bitmap2, sizeof(bitmap2)); + + test_bitmap_find_next_zero_area_off(bitmap, BITMAP_LEN); + test_find_next_bit(bitmap, BITMAP_LEN); + test_find_next_zero_bit(bitmap, BITMAP_LEN); + test_find_last_bit(bitmap, BITMAP_LEN); + test_find_nth_bit(bitmap, BITMAP_LEN / 10); + + /* + * test_find_first_bit() may take some time, so + * traverse only part of bitmap to avoid soft lockup. + */ + test_find_first_bit(bitmap, BITMAP_LEN / 10); + test_find_first_and_bit(bitmap, bitmap2, BITMAP_LEN / 2); + test_find_next_and_bit(bitmap, bitmap2, BITMAP_LEN); + + pr_err("\nStart testing find_bit() with sparse bitmap\n"); + + bitmap_zero(bitmap, BITMAP_LEN); + bitmap_zero(bitmap2, BITMAP_LEN); + + while (nbits--) { + __set_bit(get_random_u32_below(BITMAP_LEN), bitmap); + __set_bit(get_random_u32_below(BITMAP_LEN), bitmap2); + } + + test_bitmap_find_next_zero_area_off(bitmap, BITMAP_LEN); + test_find_next_bit(bitmap, BITMAP_LEN); + test_find_next_zero_bit(bitmap, BITMAP_LEN); + test_find_last_bit(bitmap, BITMAP_LEN); + test_find_nth_bit(bitmap, BITMAP_LEN); + test_find_first_bit(bitmap, BITMAP_LEN); + test_find_first_and_bit(bitmap, bitmap2, BITMAP_LEN); + test_find_next_and_bit(bitmap, bitmap2, BITMAP_LEN); + + /* + * Everything is OK. Return error just to let user run benchmark + * again without annoying rmmod. + */ + return -EINVAL; +} +module_init(find_bit_test); + +MODULE_DESCRIPTION("Test for find_*_bit functions"); +MODULE_LICENSE("GPL"); -- cgit v1.3.1