Date: Wed, 30 Apr 1997 19:05:12 +1000 From: Bruce Evans <bde@zeta.org.au> To: avalon@coombs.anu.edu.au, bde@zeta.org.au Cc: hackers@FreeBSD.org, msmith@atrad.adelaide.edu.au Subject: Re: Unloading LKMs (was Re: A Desparate Plea for Help...) Message-ID: <199704300905.TAA13932@godzilla.zeta.org.au>
next in thread | raw e-mail | index | archive | help
>> >> > Er, AFAIK unloading LKM drivers is broken in all cases. It certainly
>> ^^^^^^^
>>
>> >Hmmm, IP Filter unloads and reloads easily enough.
>>
>> It's not a driver.
>
>Actually, it is a character device module:
>MOD_DECL(if_ipl);
>
>static struct lkm_dev _module = {
> LM_DEV,
> LKM_VERSION,
> IPL_VERSION,
> CDEV_MAJOR,
> LM_DT_CHAR,
> (void *)&ipl_cdevsw
>};
Oops. It's probably broken then. I think there is no problem in 2.1.x.
I think the problem is caused by incomplete conversion to indirect devsw's.
>> Drivers are supposed to do nothing return nonzero in their unload function
>> if they are busy.
>
>I found that attempting to keep track of whether you are busy or not made
>things worse. Whether it be via signal 9 or whatever, it is somehow possible
>for things to not get cleaned up properly and you end up thinking you are
>busy when you aren't. Maybe this has changed or someone wants to look
>into it further ?
At least for pure device drivers, just keep track of opens and don't allow
unloading if anything is open.
Bruce
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199704300905.TAA13932>