Re: [PATCH BUGFIX 1/1] block, bfq: add requeue-request hook

[Holger HoffstÃtte]

The plot thickens!

Just as I was about to post that I didn't have any problems - because
I didn't have any - I decided to do a second test, activated bfq on my
workstation, on a hunch typed "sync" and .. the machine locked up, hard.

Rebooted, activated bfq, typed sync..sync hangs. Luckily this time
a second terminal was still alive, so I could capture a trace for
your enjoyment:

Feb  6 14:28:17 ragnarok kernel: io scheduler bfq registered
Feb  6 14:28:20 ragnarok kernel: BUG: unable to handle kernel NULL pointer dereference at 0000000000000030
Feb  6 14:28:20 ragnarok kernel: IP: bfq_put_queue+0x10b/0x130 [bfq]
Feb  6 14:28:20 ragnarok kernel: PGD 0 P4D 0 
Feb  6 14:28:20 ragnarok kernel: Oops: 0000 [#1] SMP PTI
Feb  6 14:28:20 ragnarok kernel: Modules linked in: bfq lz4 lz4_compress lz4_decompress nfs lockd grace sunrpc autofs4 sch_fq_codel it87 hwmon_vid x86_pkg_temp_thermal snd_hda_codec_realtek coretemp radeon crc32_pclmul snd_hda_codec_generic crc32c_intel pcbc snd_hda_codec_hdmi i2c_algo_bit aesni_intel drm_kms_helper aes_x86_64 uvcvideo syscopyarea crypto_simd snd_hda_intel sysfillrect cryptd snd_usb_audio sysimgblt videobuf2_vmalloc glue_helper fb_sys_fops snd_hda_codec snd_hwdep videobuf2_memops ttm videobuf2_v4l2 snd_usbmidi_lib videobuf2_core snd_rawmidi snd_hda_core drm snd_seq_device videodev snd_pcm i2c_i801 usbhid snd_timer i2c_core snd backlight soundcore r8169 parport_pc mii parport
Feb  6 14:28:20 ragnarok kernel: CPU: 0 PID: 4 Comm: kworker/0:0H Not tainted 4.14.18 #1
Feb  6 14:28:20 ragnarok kernel: Hardware name: Gigabyte Technology Co., Ltd. P67-DS3-B3/P67-DS3-B3, BIOS F1 05/06/2011
Feb  6 14:28:20 ragnarok kernel: Workqueue: kblockd blk_mq_requeue_work
Feb  6 14:28:20 ragnarok kernel: task: ffff88060395a1c0 task.stack: ffffc90000044000
Feb  6 14:28:20 ragnarok kernel: RIP: 0010:bfq_put_queue+0x10b/0x130 [bfq]
Feb  6 14:28:20 ragnarok kernel: RSP: 0018:ffffc90000047ca0 EFLAGS: 00010286
Feb  6 14:28:20 ragnarok kernel: RAX: 0000000000000008 RBX: ffff8806023db690 RCX: 0000000000000000
Feb  6 14:28:20 ragnarok kernel: RDX: 0000000000000000 RSI: ffff880601bb39b0 RDI: ffff880601a56400
Feb  6 14:28:20 ragnarok kernel: RBP: 0000000001bb3980 R08: 0000000000000053 R09: ffff8806023db690
Feb  6 14:28:20 ragnarok kernel: R10: 000000001dd0f11e R11: 00000000080a011b R12: ffff880601a56400
Feb  6 14:28:20 ragnarok kernel: R13: ffff8806023dbed0 R14: 0000000000000053 R15: 0000000000000000
Feb  6 14:28:20 ragnarok kernel: FS:  0000000000000000(0000) GS:ffff88061f400000(0000) knlGS:0000000000000000
Feb  6 14:28:20 ragnarok kernel: CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
Feb  6 14:28:20 ragnarok kernel: CR2: 0000000000000030 CR3: 000000000200a002 CR4: 00000000000606f0
Feb  6 14:28:20 ragnarok kernel: Call Trace:
Feb  6 14:28:20 ragnarok kernel:  bfq_finish_requeue_request+0x4b/0x370 [bfq]
Feb  6 14:28:20 ragnarok kernel:  __blk_mq_requeue_request+0x57/0x130
Feb  6 14:28:20 ragnarok kernel:  blk_mq_dispatch_rq_list+0x1b3/0x510
Feb  6 14:28:20 ragnarok kernel:  ? __bfq_bfqd_reset_in_service+0x20/0x70 [bfq]
Feb  6 14:28:20 ragnarok kernel:  ? bfq_bfqq_expire+0x212/0x740 [bfq]
Feb  6 14:28:20 ragnarok kernel:  blk_mq_sched_dispatch_requests+0xf0/0x170
Feb  6 14:28:20 ragnarok kernel:  __blk_mq_run_hw_queue+0x4e/0x90
Feb  6 14:28:20 ragnarok kernel:  __blk_mq_delay_run_hw_queue+0x73/0x80
Feb  6 14:28:20 ragnarok kernel:  blk_mq_run_hw_queue+0x53/0x150
Feb  6 14:28:20 ragnarok kernel:  blk_mq_run_hw_queues+0x3a/0x50
Feb  6 14:28:20 ragnarok kernel:  blk_mq_requeue_work+0x104/0x110
Feb  6 14:28:20 ragnarok kernel:  process_one_work+0x1d4/0x3d0
Feb  6 14:28:20 ragnarok kernel:  worker_thread+0x2b/0x3c0
Feb  6 14:28:20 ragnarok kernel:  ? process_one_work+0x3d0/0x3d0
Feb  6 14:28:20 ragnarok kernel:  kthread+0x117/0x130
Feb  6 14:28:20 ragnarok kernel:  ? kthread_create_on_node+0x40/0x40
Feb  6 14:28:20 ragnarok kernel:  ret_from_fork+0x1f/0x30
Feb  6 14:28:20 ragnarok kernel: Code: c1 e8 06 83 e0 01 48 83 f8 01 45 19 f6 e8 ce 3a 00 00 41 83 e6 ee 48 89 c7 41 83 c6 53 e8 9e 3a 00 00 49 89 d9 45 89 f0 44 89 f9 <48> 8b 70 28 48 c7 c2 d8 00 25 a0 55 4c 89 ef e8 11 ba ea e0 8b 
Feb  6 14:28:20 ragnarok kernel: RIP: bfq_put_queue+0x10b/0x130 [bfq] RSP: ffffc90000047ca0
Feb  6 14:28:20 ragnarok kernel: CR2: 0000000000000030
Feb  6 14:28:20 ragnarok kernel: ---[ end trace 8b782ace30a4e7d8 ]---

Yes, this is 4.14.x but with most-of block/4.16, rock-solid otherwise (in daily use).
Looks like there is something wrong with this patch after all..

cheers
Holger