Subject: Re: Cardbus
To: Ken Hornstein <email@example.com>
From: Warner Losh <firstname.lastname@example.org>
Date: 12/27/1999 23:15:37
In message <199912280552.AAA10025@ginger.cmf.nrl.navy.mil> Ken Hornstein writes:
: So, I take it from what you've been saying, FreeBSD really does get
: rid of the device completely upon eject now (that has changed since
: I last looked at the code).
Yes. This is new in the last few months.
: Hmmm, so I take it nothing is really done about the interface pointers
: left over in received mbufs in the FreeBSD code? Nothing really jumps
: out at me that this is dealt with.
Not directly. It just goes through the normal shutdown of the device
which seems to get rid of the dangling pointers, but I only have
negative evidence for that (eg the machine doesn't crash). I've not
had time to do a code review/grep to see if there are problems lurking
under the hood.
I'll be the first to admit that what I committed to FreeBSD was a hack
to get things stable. I had planned on defining a "this interface
went away" pr_ctlinput message so that the right stuff could happen
generically w/o the gross kludges in FreeBSD's if_detach currently.
This would most likely be the right way to deal with the mbuf problem.