Subject: Re: i386 1.4Q hangs nonrandomly?
To: None <email@example.com>
From: maximum entropy <firstname.lastname@example.org>
Date: 01/28/2000 03:06:51
>From: Jason Thorpe <email@example.com>
>Date: Thu, 27 Jan 2000 08:58:24 -0800
>I'm thinking what happened there is that you ran out of kernel virtual
>address space, and went boom.
Aha! While your explanation is still a possibility, I think I found a
more likely cause.
I still had softdeps enabled on my root partition! This would explain
why I got the lockup, and why it took a lot more abuse to get it to
lock up (/tmp would see some activity from the "make build"s, but
that's about all).
Here's why I screwed up, and thought I had disabled softdeps. Start
with a root filesystem with softdeps enabled, and do this:
boot netbsd -s
# tunefs -N /dev/rwd0a # reports softdeps enabled
# tunefs -n disable /dev/rwd0a
# tunefs -N /dev/rwd0a # reports softdeps disabled
# mount -u / # or you could reboot to multiuser
# tunefs -N /dev/rwd0a # reports softdeps enabled!!!
Is this behavior correct? Should I send-pr, or was this just a user
error on my part?
If tunefs can't really turn off softdeps on the partition while it's
mounted readonly, shouldn't it report an error, instead of seeming to
succeed, even to the point that verifying it was disabled with "tunefs
-N" says that it was successful until the partition is mounted rw?
The same strange behavior applies for enabling softdeps.