[uml-user] UML not shutting down due to "task killall5 blocked for more than 120 seconds"

[Riccardo M. <ric...@gm...>]

Hello,

we're having issues with a UML machine that refuses to shut down: the
halt sequence commences and then we get the following kernel error
message and backtrace:

    Asking all remaining processes to terminate...
    INFO: task killall5:5269 blocked for more than 120 seconds.
    [ 1080.320000] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs"
disables this message.
    [ 1080.320000] killall5        D 00000000405070c3     0  5269
5266 0x00000000
    [ 1080.320000] 93d84c08 9354d288 9354ce38 6063e550 9200bc28
6001a7f3 93d84780 92008000
    [ 1080.320000]        92008000 93d84780 9200bc88 604345ad 9200bc48
9354ce00 9200bc88 92008000
    [ 1080.320000]        00000001 93d84780 93588060 00000001
ffffffffffffffff 00000000 9200bce8 604357cd
    [ 1080.320000] Call Trace:
    [ 1080.320000] 9200bc00:  [<6001a7f3>] _switch_to+0x5e/0xae
    [ 1080.320000] 9200bc30:  [<604345ad>] schedule+0x240/0x27a
    [ 1080.320000] 9200bc90:  [<604357cd>] rwsem_down_failed_common+0xb8/0xd8
    [ 1080.320000] 9200bcf0:  [<60435814>] rwsem_down_read_failed+0x12/0x14
    [ 1080.320000] 9200bd00:  [<6002ecec>] call_rwsem_down_read_failed+0x14/0x24
    [ 1080.320000] 9200bd48:  [<60434fc2>] down_read+0x11/0x13
    [ 1080.320000] 9200bd58:  [<6007a6ec>] access_process_vm+0x45/0x147
    [ 1080.320000] 9200bdc8:  [<600cdd13>] proc_pid_cmdline+0x65/0xf9
    [ 1080.320000] 9200be18:  [<600cf298>] proc_info_read+0x68/0xcb
    [ 1080.320000] 9200be68:  [<6008f2c7>] vfs_read+0xa7/0x155
    [ 1080.320000] 9200bea8:  [<6008f42e>] sys_read+0x45/0x6c
    [ 1080.320000] 9200bee8:  [<6001cd9c>] handle_syscall+0x58/0x70
    [ 1080.320000] 9200bf08:  [<6002c00f>] userspace+0x2d4/0x381
    [ 1080.320000] 9200bfc8:  [<6001a6f3>] fork_handler+0x62/0x69
    [ 1080.320000]

The error repeats every 120 seconds, and the machine never shuts down.

We're using kernel64-2.6.38.4 from http://uml.devloop.co.uk/ and the
system image is an up-to-date Debian 6 (squeeze).

Any suggestions on how to debug or work around this problem?

Thanks,
Riccardo