Subject: kern/23718: umount delay causes repeated keystrokes
To: None <gnats-bugs@gnats.NetBSD.org>
From: None <grant@NetBSD.org>
List: netbsd-bugs
Date: 12/12/2003 00:45:42
>Number:         23718
>Category:       kern
>Synopsis:       umount delay causes repeated keystrokes
>Confidential:   no
>Severity:       non-critical
>Priority:       medium
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Thu Dec 11 13:46:00 UTC 2003
>Closed-Date:
>Last-Modified:
>Originator:     grant beattie
>Release:        NetBSD 1.6ZG
>Organization:
>Environment:
System: NetBSD mofo 1.6ZG NetBSD 1.6ZG (MOFO) #2: Thu Dec 11 00:30:00 EST 2003 grant@mofo:/src/sys/arch/i386/compile/MOFO i386
Architecture: i386
Machine: i386
>Description:
on my workstation, with a low delay-to-repeat, running "umount
/mountpoint", regardless of whether or not anything is mounted there,
causes repeated keystrokes.

I'm guessing this happens with more than just umount...

>How-To-Repeat:

mofo: {1} xset r rate 200 40
mofo: {2} time umount /cdrom

umount: /cdrom: not currently mounted
0.0u 0.2s 0:00.28 96.4% 0+0k 0+11io 0pf+0w
mofo: {3} mofo: {3} 

two extra <enter> keypresses somehow find their way in.

with a higher time to repeat, this doesn't happen:

mofo: {3} xset r rate 400 40
mofo: {4} time umount /cdrom
umount: /cdrom: not currently mounted
0.0u 0.2s 0:00.27 92.5% 0+0k 0+8io 0pf+0w
mofo: {5}

it seems the time taken making system calls contributes to this, and
200ms is a fairly long time to determine there's nothing mounted
there (this is a p3-866, it clearly shouldn't take that long or screw
up keyboard input :)

this is easily reproducable:

- when the machine is busy
- with a low time-to-repeat

>Fix:
>Release-Note:
>Audit-Trail:
>Unformatted: