[Differential] [Commented On] D1944: PF and VIMAGE fixes

nvass-gmx.com (Nikos Vassiliadis) phabric-noreply at FreeBSD.org
Sat May 16 12:25:57 UTC 2015

nvass-gmx.com added a comment.

In https://reviews.freebsd.org/D1944#47231, @rodrigc wrote:

> I tested this patch.
>   # kldload pf
>   # kldunload pf
>   kldunload: can't unload file: Device busy
> The fact that the pf module cannot be unloaded was one of the
>  reasons that @glebius used to back out the entire changeset last time
>  I committed your pf changes.  Can you fix this?

This is intended behaviour, regadless of VIMAGE. You need to use kldunload -f
to unload it.

> I also saw this in dmesg:


>   CURVNET_SET() recursion in pfi_vnet_initialize() line 130, prev in vnet_register_sysinit()

>       0xfffff800056e4100 -> 0xfffff800056e4100

>   KDB: stack backtrace:

>   db_trace_self_wrapper() at db_trace_self_wrapper+0x2b/frame 0xfffffe046389a550

>   pfi_vnet_initialize() at pfi_vnet_initialize+0x21b/frame 0xfffffe046389a590

>   pf_vnet_init() at pf_vnet_init+0x35/frame 0xfffffe046389a5c0

>   vnet_register_sysinit() at vnet_register_sysinit+0x13c/frame 0xfffffe046389a600

>   linker_load_module() at linker_load_module+0xc87/frame 0xfffffe046389a920

>   kern_kldload() at kern_kldload+0x10e/frame 0xfffffe046389a970

>   sys_kldload() at sys_kldload+0x5b/frame 0xfffffe046389a9a0

>   amd64_syscall() at amd64_syscall+0x27f/frame 0xfffffe046389aab0

>   Xfast_syscall() at Xfast_syscall+0xfb/frame 0xfffffe046389aab0


This should be ok. It is just a warning that we are setting curvnet although
it is already set. Maybe it can be avoided.



To: nvass-gmx.com, bz, zec, trociny, glebius, rodrigc, kristof, gnn
Cc: julian, robak, freebsd-virtualization, freebsd-pf, freebsd-net

More information about the freebsd-virtualization mailing list