内核6.5.0启动时出现大量bcache相关UBSAN数组索引越界报错的咨询

问题背景 #

我搭建了基于两个设备的btrfs RAID1文件系统，这两个设备是bcache设备（NVMe+HDD）通过cryptsetup luksOpen映射出来的。这个btrfs文件系统运行起来完全正常，但每次启动都会弹出大量类似下面的内核报错信息，打断了图形启动界面，看起来可能很严重，想问问这是什么情况。

报错日志 #

2023-12-19T23:21:41.595579+01:00 HostName kernel: [    4.042319] ================================================================================

2023-12-19T23:21:41.595580+01:00 HostName kernel: [    4.042327] UBSAN: array-index-out-of-bounds in /build/linux-SXblTa/linux-6.5.0/drivers/md/bcache/bset.c:1098:3

2023-12-19T23:21:41.595580+01:00 HostName kernel: [    4.042338] index 4 is out of range for type 'btree_iter_set [4]'

2023-12-19T23:21:41.595580+01:00 HostName kernel: [    4.042344] CPU: 6 PID: 133 Comm: kworker/6:1 Not tainted 6.5.0-14-generic #14-Ubuntu

2023-12-19T23:21:41.595581+01:00 HostName kernel: [    4.042346] Hardware name: ASUS System Product Name/PRIME X670E-PRO WIFI, BIOS 1813 10/13/2023

2023-12-19T23:21:41.595581+01:00 HostName kernel: [    4.042347] Workqueue: events register_cache_worker [bcache]

2023-12-19T23:21:41.595581+01:00 HostName kernel: [    4.042355] Call Trace:

2023-12-19T23:21:41.595581+01:00 HostName kernel: [    4.042356]  <TASK>

2023-12-19T23:21:41.595582+01:00 HostName kernel: [    4.042357]  dump_stack_lvl+0x48/0x70

2023-12-19T23:21:41.595582+01:00 HostName kernel: [    4.042359]  dump_stack+0x10/0x20

2023-12-19T23:21:41.595582+01:00 HostName kernel: [    4.042361]  __ubsan_handle_out_of_bounds+0xc6/0x110

2023-12-19T23:21:41.595582+01:00 HostName kernel: [    4.042364]  bch_btree_iter_push+0x2d3/0x4f0 [bcache]

2023-12-19T23:21:41.595583+01:00 HostName kernel: [    4.042373]  bch_btree_node_read_done+0xcb/0x410 [bcache]

2023-12-19T23:21:41.595583+01:00 HostName kernel: [    4.042382]  bch_btree_node_read+0xf8/0x1e0 [bcache]

2023-12-19T23:21:41.595583+01:00 HostName kernel: [    4.042391]  ? __pfx_closure_sync_fn+0x10/0x10 [bcache]

2023-12-19T23:21:41.595584+01:00 HostName kernel: [    4.042400]  bch_btree_node_get.part.0+0x15c/0x330 [bcache]

2023-12-19T23:21:41.595584+01:00 HostName kernel: [    4.042408]  ? __bch_btree_ptr_invalid+0x66/0xe0 [bcache]

2023-12-19T23:21:41.595584+01:00 HostName kernel: [    4.042416]  ? __pfx_up_write+0x10/0x10

2023-12-19T23:21:41.595584+01:00 HostName kernel: [    4.042419]  bch_btree_node_get+0x16/0x30 [bcache]

2023-12-19T23:21:41.595585+01:00 HostName kernel: [    4.042427]  run_cache_set+0x596/0x850 [bcache]

2023-12-19T23:21:41.595585+01:00 HostName kernel: [    4.042435]  ? srso_alias_return_thunk+0x5/0x7f

2023-12-19T23:21:41.595586+01:00 HostName kernel: [    4.042439]  register_cache_set+0x1a2/0x210 [bcache]

2023-12-19T23:21:41.595587+01:00 HostName kernel: [    4.042448]  register_cache+0x11a/0x1a0 [bcache]

2023-12-19T23:21:41.595587+01:00 HostName kernel: [    4.042456]  register_cache_worker+0x22/0x80 [bcache]

2023-12-19T23:21:41.595587+01:00 HostName kernel: [    4.042464]  process_one_work+0x220/0x440

2023-12-19T23:21:41.595587+01:00 HostName kernel: [    4.042467]  worker_thread+0x4d/0x3f0

2023-12-19T23:21:41.595588+01:00 HostName kernel: [    4.042468]  ? srso_alias_return_thunk+0x5/0x7f

2023-12-19T23:21:41.595588+01:00 HostName kernel: [    4.042470]  ? _raw_spin_lock_irqsave+0xe/0x20

2023-12-19T23:21:41.595588+01:00 HostName kernel: [    4.042472]  ? __pfx_worker_thread+0x10/0x10

2023-12-19T23:21:41.595588+01:00 HostName kernel: [    4.042474]  kthread+0xef/0x120

2023-12-19T23:21:41.595589+01:00 HostName kernel: [    4.042476]  ? __pfx_kthread+0x10/0x10

2023-12-19T23:21:41.595589+01:00 HostName kernel: [    4.042478]  ret_from_fork+0x44/0x70

2023-12-19T23:21:41.595590+01:00 HostName kernel: [    4.042480]  ? __pfx_kthread+0x10/0x10

2023-12-19T23:21:41.595590+01:00 HostName kernel: [    4.042482]  ret_from_fork_asm+0x1b/0x30

2023-12-19T23:21:41.595590+01:00 HostName kernel: [    4.042485]  </TASK>

2023-12-19T23:21:41.595590+01:00 HostName kernel: [    4.042486] ================================================================================

2023-12-19T23:21:41.598663+01:00 HostName kernel: [    4.091709] bcache: bch_journal_replay() journal replay done, 0 keys in 2 entries, seq 140818

问题分析与建议 #

首先别太慌，咱们拆解一下这个报错：

• 这是**UBSAN（未定义行为检查器）**抛出的数组索引越界错误，来自bcache驱动的bset.c文件第1098行——简单说就是代码里访问了一个长度为4的数组的第5个元素（索引从0开始，所以索引4超出范围）。
• 但你提到btrfs文件系统运行完全正常，而且日志最后显示bch_journal_replay()已经完成了日志重放，说明bcache初始化其实是成功的，这个报错大概率是驱动代码里一个不影响实际功能的小瑕疵触发了UBSAN的严格检查，而非会导致数据损坏的严重bug。
• 这个问题出在启动阶段的register_cache_worker工作队列里，启动完成后驱动已经正常工作，所以日常使用不会受影响。

给你几个处理方案：

1. 暂时忽略，等待内核更新：Ubuntu的6.5.x内核后续版本大概率会修复这个bcache的UBSAN触发问题，毕竟这是一个明显的代码检查告警，官方维护者会跟进的。
2. 临时屏蔽报错（仅为美观）：如果觉得启动时的报错太碍眼，可以尝试在GRUB启动参数中添加ubsan_ignore=array-index-out-of-bounds，这样UBSAN会跳过这类越界检查。不过这只是临时方案，不推荐长期用，还是等官方修复更稳妥。
3. 检查bcache配置：可以确认一下你的bcache设备是否正常初始化，比如用bcache-super-show查看设备元数据，但从日志看元数据重放已经完成，大概率没问题。

备注：内容来源于stack exchange，提问作者Fade