linux-kernel - [Bisected] Disconnecting pendrive with opened files causes trouble on 6.0.0-rc3

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Date:   Sun, 4 Sep 2022 20:28:47 +0200
From:   Mateusz Jończyk <mat.jonczyk@...pl>
To:     "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        linux-scsi@...r.kernel.org, adrian.hunter@...el.com,
        martin.petersen@...cle.com, bvanassche@....org, jaegeuk@...nel.org,
        michael.christie@...cle.com, hch@....de, ming.lei@...hat.com,
        hare@...e.de, john.garry@...wei.com
Subject: [Bisected] Disconnecting pendrive with opened files causes trouble on
 6.0.0-rc3

Hello,

On Linux 6.0.0-rc3, disconnecting a mounted pendrive with opened files causes these symptoms:

- "task kworker/X blocked for more than Y seconds" warnings (see below) are repeatedly printed in dmesg,
- after reconnecting the pendrive (or possibly other USB devices), it is not detected,
- the login screen (Gnome on Ubuntu 20.04) freezes for some time (around 20s), due to a hang in colord-sane.

These symptoms disappear after closing the opened files (and perhaps also directories) on the pendrive.

I am aware that disconnecting a pendrive with a mounted filesystem is unwise, but many people do so
nonetheless.

I have bisected this down to
commit 16728aaba62e ("scsi: core: Make sure that hosts outlive targets")

With the previous
commit fe442604199e ("scsi: core: Make sure that targets outlive devices")
the problem does not happen.

To reproduce:

1. Plug a pendrive into a USB port (checked with a pendrive formatted with a FAT32 filesystem).
2. Mount it, open a file on this pendrive with the "less" command-line program.
3. Optionally suspend the system to RAM.
4. Disconnect the pendrive.
5. Resume the system, if it was suspended.

Warnings from dmesg:

[   76.212690] usb 1-3.4.3: USB disconnect, device number 6
[   76.219765] device offline error, dev sdc, sector 2049 op 0x1:(WRITE) flags 0x100000 phys_seg 1 prio class 2
[   76.219791] Buffer I/O error on dev sdc1, logical block 1, lost async page write
[  242.764446] INFO: task kworker/0:7:1701 blocked for more than 120 seconds.
[  242.764463]       Tainted: G            E      6.0.0-rc3mj6 #239
[  242.764470] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[  242.764474] task:kworker/0:7     state:D stack:    0 pid: 1701 ppid:     2 flags:0x00004000
[  242.764488] Workqueue: usb_hub_wq hub_event
[  242.764504] Call Trace:
[  242.764507]  <TASK>
[  242.764514]  __schedule+0x2c5/0xcd0
[  242.764527]  schedule+0x5d/0xf0
[  242.764535]  scsi_remove_host+0x163/0x1b0
[  242.764547]  ? var_wake_function+0x60/0x60
[  242.764559]  usb_stor_disconnect+0x50/0xd0 [usb_storage]
[  242.764571]  usb_unbind_interface+0x8c/0x240
[  242.764581]  device_remove+0x64/0x70
[  242.764590]  device_release_driver_internal+0xd1/0x160
[  242.764599]  device_release_driver+0x12/0x20
[  242.764608]  bus_remove_device+0xde/0x150
[  242.764616]  device_del+0x192/0x3e0
[  242.764621]  ? usb_remove_ep_devs+0x1f/0x30
[  242.764632]  usb_disable_device+0xab/0x240
[  242.764639]  usb_disconnect+0xc7/0x260
[  242.764647]  ? set_port_feature+0x37/0x40
[  242.764656]  hub_event+0xeb0/0x17c0
[  242.764668]  ? __schedule+0x2cd/0xcd0
[  242.764675]  ? queue_rcu_work+0x2c/0x40
[  242.764684]  process_one_work+0x21c/0x3c0
[  242.764693]  worker_thread+0x4a/0x3b0
[  242.764703]  ? process_one_work+0x3c0/0x3c0
[  242.764711]  kthread+0xcf/0xf0
[  242.764719]  ? kthread_complete_and_exit+0x20/0x20
[  242.764728]  ret_from_fork+0x22/0x30
[  242.764737]  </TASK>

The kernel is tainted AFAIK because of the following:

[    0.442486] backlight: module verification failed: signature and/or required key missing - tainting kernel
[...]
[    0.446462] Loading compiled-in X.509 certificates
[    0.447431] Loaded X.509 cert 'Build time autogenerated kernel key: cbf6b5d58385db4124f3ddd5fa0457b112415dcb'

The module was loaded before the signing keys were processed - which is an unrelated bug, present since a long time.


System information:
- Ubuntu 20.04 amd64,
- HP 17-by0001nw laptop,
- problem happens also when using an external USB 3.0 UNITEK hub.

Greetings,

Mateusz Jończyk