Re: Pcmcia ne(4)/mii crash in -current

[Jason Thorpe <thorpej%me.com@localhost>]

> On Jun 29, 2021, at 4:35 PM, Björn Johannesson <rherdware%yahoo.com@localhost> wrote:
> 
> Hello and thanks again for looking into this. I really appreciate it.

No problem.  It’s all of my MII / ifmedia locking changes that probably broke it in the first place :-)

> We have some progress. The patch compiles after fixing two typos.
> Two instances of one struct too many
> struct struct dp8390_softc *sc = arg;

Fixed! :-)

> # ifconfig ne2
> ne2: flags=0x8802<BROADCAST,SIMPLEX,MULTICAST> mtu 1500
> ec_capabilities=0x1<VLAN_MTU>
> ec_enabled=0
> address: 00:0d:88:40:68:0e
> # dhcpcd ne2
> dhcpcd-9.4.0 starting
> DUID 00:01:00:01:1a:14:11:b8:00:0d:88:40:68:0e
> ne2: waiting for carrier
> ne2: carrier unknown, assuming up
> ne2: IAID 88:40:68:0e
> ne2: adding address fe80::53b7:6ccf:4e12:3fb7
> ne2: carrier lost - roaming
> ne2: carrier lost - roaming
> ne2: soliciting an IPv6 router
> ne2: carrier acquired
> ne2: IAID 88:40:68:0e
> ne2: soliciting an IPv6 router
> ne2: soliciting a DHCP lease
> ne2: offered 192.168.1.32 from 192.168.1.1
> ne2: leased 192.168.1.32 for 1800 seconds
> ne2: adding route to 192.168.1.0/24
> ne2: adding default route via 192.168.1.1
> forked to background, child pid 854
> # 
> 
> I don't remember at the top of my head if it acted like this in re: carrier previously but it gets there in the end.

Ok, that’s good.

> However there is still no detach event for when its been unused.

That I have no idea what’s going on.  Let’s get the interface more-or-less working first, then we can dive into the CardBus code.

> And when detaching a used card it still crashes.
> 
> # dhcpcd -k ne2
> sending signal ALRM to pid 854
> waiting for pid 854 to exit
> # Jun 29 00:05:35 abbie dhcpcd[851]: ps_root_log: Connection reset by peer
> route flush
> # ifconfig ne2
> ne2: flags=0x8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500
> ec_capabilities=0x1<VLAN_MTU>
> ec_enabled=0
> address: 00:0d:88:40:68:0e
> status: active
> # <--- detach here
> [ 279.3474089] nsphyter0: detached
> [ 279.3838727] ne2: detached
> [ 279.8574170] panic: kernel diagnostic assertion "mii_locked(mii)" failed: file "../../../../dev/mii/mii.c", line 342 

Oops.  Adding a "callout_stop(&sc->sc_tick_ch);” before the mii_detach() call should fix that.

New composite patch:

	https://www.netbsd.org/~thorpej/dp8390-assert-patch-v4.txt

-- thorpej