CURRENT + amd64 + user-ppp = panic

[Victor Snezhko]

Mark Tinguely <tinguely at casselton.net> writes:

> I also was not thinking MBUF when I suggested trace locations. m_sanity()
> in kern/uipc_mbuf.c will also set "0xdeadc0de" if the "sanitize"
> option is requested.

Now I have a current of 2005.10.21.16.25.00, and there is no deadcode
there. Only a comment stating that deadc0de'ing would be not bad.

> But common sense says that your trash_dtor()
> code should catch this before it gets set, but if you wish to test
> there also it would put my mind at ease.

Sure. I have already submitted a PR(kern/88725) with all the results I
have so far, and when I find all the other places I'll make a followup
to it.

> In the mean time, I will look for callouts that are not stopped before
> they are freed. So far, I found the &sp->ifstart_callout in the file
> net/if_spppsubr.c started but not stopped before freed, but you are
> not using that routine. I will concentrate on the files that you listed
> were changed.

Yes, it's right. This commit needs a serious checking even if we'll
fix my issue with ppp.

-- 
WBR, Victor V. Snezhko
EMail: snezhko at indorsoft.ru