Re: Deferred freeing of kauth_cred_ts

[Elad Efrat <elad%NetBSD.org@localhost>]

On Mon, Dec 28, 2009 at 2:41 AM, Adam Hamsik <haaaad%gmail.com@localhost> wrote:
>
> On Dec,Monday 28 2009, at 2:22 AM, Elad Efrat wrote:
>
>> rmind@ was right: ipintr -> tcp_input -> tcp_close -> in_pcbdetach -> sofree.
>>
>> Does anyone see a way to fix the path so it doesn't call sofree() in
>> interrupt context? If not, then (unless there's an option I'm missing)
>> I think we should do the deferred free thread for kauth(9) rather than
>> an isolated fix for the network stack(s)...
>
> Can we use workqueue for calling tcp_close ? I think that it might be a 
> solution.

Do we need to allocate a "work unit" every time we want to enqueue a
job for tcp_close?

-e.