Can't kill a umount
Cameron Simpson
cs at zip.com.au
Sun Sep 6 20:35:28 UTC 2015
On 06Sep2015 13:23, Tom Horsley <horsley1953 at gmail.com> wrote:
>On Sun, 6 Sep 2015 12:11:18 -0500 (CDT)
>Michael Hennebry wrote:
>
>> Sometimes I get a "drive busy" message from umount,
>> which then fails.
>> In that case, a file has usually been open.
>> Does umount regard this as a different situation?
>
>The device busy means something is actively using it.
>The taking a long time to umount means no one is
>actively using it, but the in memory buffers haven't
>been completely flushed.
Though in this case the umount command itself will not be using any CPU - it
will just be waiting for the OS umount call to come back.
I'm actually astounded that unout will spin out unless in-kernel work is being
counted to the CPU time (not normally the case). Yes, the drive appears to be
buggy, but that should be concealed inside the kernel.
Cheers,
Cameron Simpson <cs at zip.com.au>
More information about the users
mailing list