Created attachment 282667 [details] detailed_configuration_and_traceback This kernel module BUG is triggered on manual cached dev detach. Tested kernels 4.14.101 - 4.14.114 Linux distro - Gentoo To test create bcached backing and caching device attach and attempt to detach. make-bcache -B /dev/sda make-bcache -C /dev/sdb echo cdev.uuid > /sys/block/bcache0/bcache/attach echo cdev.uuid > /sys/block/bcache0/bcache/detach To improve logs, enabled dynamic_debug by writing echo 'options bcache dyndbg=+pt' > /etc/modprobe.d/bcache.conf in dmesg following BUG will appear: kernel: [10279] bcache: __write_super() ver 1, flags 2305843009213693952, seq 0 kernel: ------------[ cut here ]------------ kernel: Kernel BUG at ffffffffa0784f25 [verbose debug info unavailable] kernel: invalid opcode: 0000 [#1] SMP PTI kernel: Modules linked in: drbd lru_cache iptable_filter ip_tables nfsd auth_rpcgss oid_registry nfs_acl lockd grace sunrpc mlx4_ib ib_core mlx4_en intel_rapl x86_pkg_temp_thermal coretemp kvm_intel kvm irqbypass intel_cstate iTCO_wdt iTCO_vendor_support mlx4_core igb input_leds bcac he devlink intel_rapl_perf pcspkr led_class i2c_i801 ptp lpc_ich joydev ioatdma pps_core ipmi_si dca ipmi_devintf rtc_cmos ipmi_msghandler pcc_cpufreq acpi_pad sch_fq_codel btrfs xor zstd_decompress zstd_compress xxhash lzo_compress raid6_pq dm_crypt usbhid mxm_wmi crc32c_intel arcmsr ahci ast xhci_pci libahci ehci _pci ttm ehci_hcd xhci_hcd libata wmi button dm_mirror dm_region_hash dm_log dm_mod kernel: CPU: 27 PID: 295 Comm: kworker/27:1 Not tainted 4.14.114-gentoo-r1-201905 #1 kernel: Hardware name: Supermicro X10DRH LN4/X10DRH-ILN4, BIOS 2.0 01/30/2016 kernel: Workqueue: events cached_dev_detach_finish [bcache] kernel: task: ffff88885cf84440 task.stack: ffffc90008a48000 kernel: RIP: 0010:cached_dev_detach_finish+0x55/0x1b0 [bcache] kernel: RSP: 0018:ffffc90008a4bdf8 EFLAGS: 00010286 kernel: RAX: 00000000ffffffff RBX: ffff88905c200af0 RCX: 0000000000000000 kernel: RDX: 0000000000000001 RSI: ffff88905c200af8 RDI: ffffc90008a4be48 kernel: RBP: ffff88905c200000 R08: 0000000000000001 R09: 0000000000000001 kernel: R10: ffffc90008a4be30 R11: 00273bdd233be0d4 R12: ffffc90008a4bdf8 kernel: R13: 0000000000000000 R14: ffff88885fc60880 R15: 0000000000000000 kernel: FS: 0000000000000000(0000) GS:ffff88885fc40000(0000) knlGS:0000000000000000 kernel: CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 kernel: CR2: 000055f9fae819f8 CR3: 000000000220a004 CR4: 00000000003606e0 kernel: DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 kernel: DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 kernel: Call Trace: kernel: process_one_work+0x1cf/0x370 kernel: worker_thread+0x208/0x380 kernel: ? process_one_work+0x370/0x370 kernel: kthread+0x11d/0x130 kernel: ? kthread_destroy_worker+0x40/0x40 kernel: ret_from_fork+0x35/0x40 kernel: Code: 89 44 24 70 31 c0 49 89 e4 4c 89 e7 f3 48 ab c7 44 24 28 01 00 00 a0 48 8b 83 d0 f5 ff ff a8 02 75 02 0f 0b 8b 43 f8 85 c0 74 02 <0f> 0b 48 c7 c7 00 1d 7a a0 e8 7d 37 0f e1 48 8d 7b 68 e8 54 91 kernel: RIP: cached_dev_detach_finish+0x55/0x1b0 [bcache] RSP: ffffc90008a4bdf8 kernel: ---[ end trace e213d6c6632e8c72 ]--- In attachment included traceback, GDB_disassemble, kernel config
Created attachment 282763 [details] bcache patch bundle Including bundle of 15 patches to fix bcache in kernel based on version 4.14.114/115/116/117/118/119 ├── 1001.bcache_avoid_nested_58f913dce281.patch ├── 1002.bcache_fix_comment_type_b1e8139e48b5.patch ├── 1003.bcache_rewrite_multipart_1dbe32ad0a82.patch ├── 1007.bcache_dont_writeback_failed_io_5fa89fb9a86b.patch ├── 1015.bcache_convert_cacheed_dev_atomi_to_refcount_3b304d24a718.patch ├── 1016.bcache_update_bucket_real_time_d44c2f9e7cc0.patch ├── 1019.bcache_4.15_backport_writeback.patch ├── 1020.bcache_convert_timers_8376d3c1f989.patch ├── 1021.bcache_journal_comment_bb22cafd7568.patch ├── 1025.bcache_use_PTR_ERR_OR_ZERO_9d13411784e2.patch ├── 1028.bcache_fix_wrong_return_bch_debug_init_539d39eb2708.patch ├── 1031.bcache_improve_efficiency_closure_e4bf791937d8.patch ├── 1033.bcache_reduce_cache_set_device_iteration_2831231d4c3f.patch └── 1036.bcache_closure_move_control_bits_3609c471a1b8.patch patches where created by back-porting from linux kernel git tree. Current state of bcache in kernel 4.14.114/115/116/117/118/119 Create backing and cache device and attach works but we can't detach at all. I/O errors are not detected which might cause unexpected problems. Detach functionality could be fixed by applying patch 1019.bcache_4.15_backport_writeback.patch After that we can detach but first need to write none into cache_module echo none > /sys/block/bcacheX/bcache/cache_modue Following to that we need to fix I/O error detection otherwise when we have active I/O operations on caching device and we attempt to detach, then will find broken symlink in /sys/block/bcacheX/bcache So to sensible fix bcache in kernel 4.14.X we need to apply all above patches. Above patches where tested on kernel 4.14.115. After applying we can create backing, cache device, attach, detach, disable, partition with fixed implementation, I/O error detection works and cache device being detach with following messages in dmesg """ bcache: bch_count_io_errors() sdc: IO error on writing data to cache, recovering bcache: error on ed80d9c1-5fc2-4390-bd9e-f633dd6a3a4f: journal io error , disabling caching bcache: cached_dev_detach_finish() Caching disabled for dm-1 bcache: bch_count_io_errors() sdc: IO error on writing btree, recovering bcache: cache_set_free() Cache set ed80d9c1-5fc2-4390-bd9e-f633dd6a3a4f unregistered """ Tested performance witch file copy and benchmark with FIO tool.
Does this issue still show up in Linux v5.5 or v5.6-rc kernel ? Thanks. Coly Li