Re: ipfilter cannot be build within because warning's are present

Max Laier wrote:
> On Thursday 18 October 2007, Victor M. Blood wrote:
>   
>> On 18.10.2007, Max Laier wrote:
>>     
>>> On Wednesday 17 October 2007, Victor M. Blood wrote:
>>>       
>>>> Hi, All.
>>>>
>>>> I try to use options in kernel instead of a module build of the
>>>> ipfilter and got error then kernel builds.
>>>>
>>>> I'm edit files: fil.c, ip_auth.h, ip_auth.h, ip_log.c ip_compat.h
>>>> and correct #ifdef statament :) no more warnings...
>>>>         
>>> ipf is likely broken anyway.  See thread: "7.0 CURRENT, need help
>>> with panic: Trying sleep, but thread marked as sleeping prohibited"
>>> on this ML a few days back.  That this warning went unnoticed tells
>>> you something, too.
>>>       
>> New version, new problems, no more to say) few days back I have panic
>> with ip filter, now it seems as worked, after upgrade.
>>
>> #ipf -V
>> ipf: IP Filter: v4.1.27 (404)
>> Kernel: IP Filter: v4.1.27
>> Running: yes
>> Log Flags: 0 = none set
>> Default: block all, Logging: available
>> Active list: 0
>> Feature mask: 0xe
>>     
>
> Is this with a WITNESS/INVARIANTS enabled kernel?  From a quick glance at 
> the code you should see warnings on the ioctl path as there are copy 
> operations to/from userland with the lock held.
>
> Moving to rw_locks is the right direction, but the config path is still 
> broken.
>   

The real fault here is if the system needs to page in/out over NFS
in order to fulfill the copyin/out.

Does devfs prevent a kldunload from being executed while calls
(ioctl/read/write) are outstanding?

Darren