Re: kern/54434: ifconfig lo0 destroy leaves lo0 in unusable state

[Ryota Ozaki <ozaki-r%netbsd.org@localhost>]

On Sun, Aug 4, 2019 at 4:08 AM <logix%foobar.franken.de@localhost> wrote:
>
> >Number:         54434
> >Category:       kern
> >Synopsis:       ifconfig lo0 destroy leaves lo0 in unusable state
> >Confidential:   no
> >Severity:       serious
> >Priority:       medium
> >Responsible:    kern-bug-people
> >State:          open
> >Class:          sw-bug
> >Submitter-Id:   net
> >Arrival-Date:   Sat Aug 03 10:35:00 +0000 2019
> >Originator:     Harold Gutch
> >Release:        HEAD
> >Organization:
> >Environment:
> NetBSD localhost 9.99.2 NetBSD 9.99.2 (GENERIC) #0: Thu Aug  1 22:23:16 UTC 2019  mkrepro%mkrepro.NetBSD.org@localhost:/usr/src/sys/arch/amd64/compile/GENERIC amd64
> >Description:
> When calling "ifconfig <device> destroy", the generic clone destroy function sets the interface ioctl to if_nullioctl before running the device specific clone destroy function.  If the device specific destroy function fails, the device still exists and works but there is no reference anymore to its ioctl function and the device subsequently cannot be configured anymore until the next reboot.
>
> This can happen for lo0, ppp*, sl*, srt*, strip*.
>
> This was introduced in rev. 1.336 of src/sys/net/if.c.
> >How-To-Repeat:
> localhost# ifconfig -a
> pcn0: flags=0x8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500
>         address: 08:00:27:a8:b0:5a
>         inet 192.168.56.101/24 broadcast 192.168.56.255 flags 0x0
>         inet6 fe80::a00:27ff:fea8:b05a%pcn0/64 flags 0x0 scopeid 0x1
> lo0: flags=0x8049<UP,LOOPBACK,RUNNING,MULTICAST> mtu 33624
>         inet 127.0.0.1/8 flags 0x0
>         inet6 ::1/128 flags 0x20<NODAD>
>         inet6 fe80::1%lo0/64 flags 0x0 scopeid 0x2
> localhost# ifconfig lo0 destroy
> ifconfig: clone_command: Operation not permitted
> ifconfig: exec_matches: Operation not permitted
> localhost# ifconfig -a
> pcn0: flags=0x8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500
>         address: 08:00:27:a8:b0:5a
>         inet 192.168.56.101/24 broadcast 192.168.56.255 flags 0x0
>         inet6 fe80::a00:27ff:fea8:b05a%pcn0/64 flags 0x0 scopeid 0x1
> ifconfig: status: getifinfo: Device not configured
> localhost# ifconfig lo0
> ifconfig: SIOCGIFFLAGS lo0: Device not configured
> localhost# ifconfig -l
> pcn0 lo0
> >Fix:
> This patch seems to fix the issue, but I'm not sure if there are cases where it has side effects.
>
> --- src/sys/net/if.c.orig       2019-07-25 07:45:57.000000000 +0000
> +++ src/sys/net/if.c    2019-08-03 12:15:43.318986910 +0000
> @@ -1611,6 +1611,8 @@
>         struct if_clone *ifc;
>         struct ifnet *ifp;
>         struct psref psref;
> +       int errno;
> +       int (*if_ioctl)(struct ifnet *, u_long, void *);
>
>         KASSERT(mutex_owned(&if_clone_mtx));
>
> @@ -1627,6 +1629,7 @@
>
>         /* We have to disable ioctls here */
>         IFNET_LOCK(ifp);
> +       if_ioctl = ifp->if_ioctl;
>         ifp->if_ioctl = if_nullioctl;
>         IFNET_UNLOCK(ifp);
>
> @@ -1636,7 +1639,13 @@
>          */
>         if_put(ifp, &psref);
>
> -       return (*ifc->ifc_destroy)(ifp);
> +       errno = (*ifc->ifc_destroy)(ifp);
> +       if (errno) {
> +               IFNET_LOCK(ifp);
> +               ifp->if_ioctl = if_ioctl;
> +               IFNET_UNLOCK(ifp);
> +       }
> +       return errno;
>  }
>
>  static bool
>

Thank you for the fix!  I've committed it (with minor tweaks)
and will pull it up to NetBSD 8 and 9.

  ozaki-r