sk Panic in 8.0-CURRENT
Cy Schubert
Cy.Schubert at komquats.com
Wed Feb 13 22:36:41 UTC 2008
In message <20080213003554.GA11251 at cdnetworks.co.kr>, Pyun YongHyeon writes:
>
> --PNTmBPCT7hxwcZjr
> Content-Type: text/plain; charset=us-ascii
> Content-Disposition: inline
>
> On Mon, Feb 11, 2008 at 07:45:31PM -0800, Cy Schubert wrote:
> > In message <20080212014312.GA6953 at cdnetworks.co.kr>, Pyun YongHyeon writes
> :
> > > On Mon, Feb 11, 2008 at 04:10:40PM -0800, Cy Schubert wrote:
> > > > Has anyone seen the following mutex panic in sk_jfree? The last time
> this
> > > > system booted was Jan 31.
> > > >
> > >
> > > [...]
> > >
> > > > panic: mtx_lock() of spin mutex (null) @ /dsk03/src/cvs-current/src/s
> ys/mo
> > > du
> > > > les/sk/../../dev/sk/if_sk.c:2439
> > > > cpuid = 0
> > > > KDB: enter: panic
> > > > [thread pid 12 tid 100038 ]
> > > > Stopped at kdb_enter+0x34: movl $0,kdb_why
> > > > db> bt
> > > > Tracing pid 12 tid 100038 td 0xc3363cc0
> > > > kdb_enter(c0a36183,c0a36183) at kdb_enter+0x34
> > > > panic(c0a34f9b,0,c0cefb36,987,e2583cc0,...) at panic+0x111
> > > > _mtx_lock_flags(e2586bbc,0,c0cefb36,987,c35d1000,...) at
> > > > _mtx_lock_flags+0x70
> > > > sk_jfree(c341f000,e2583cc0) at sk_jfree+0x3a
> > > > mb_free_ext(c35d1000) at mb_free_ext+0x18f
> > > > m_freem(c35d1000) at m_freem+0x1f
> > > > arpintr(c35d1000) at arpintr+0xc0b
> > > > netisr_dispatch(12,c35d1000) at netisr_dispatch+0x5d
> > > > ether_demux(c33d5400,c35d1000) at ether_demux+0x1c9
> > > > ether_input(c33d5400,c35d1000,c33f36e0,0,c0cefb36,...) at ether_input
> +0x2f
> > > 9
> > > > sk_jumbo_rxeof(c33f36e0,c341f000,c33d5400,0,c342d340,...) at
> > > > sk_jumbo_rxeof+0x215
> > > > sk_intr(c33f3680) at sk_intr+0xac
> > > > ithread_loop(c342ab40,e2589d38) at ithread_loop+0x175
> > > > fork_exit(c06eded0,c342ab40,e2589d38) at fork_exit+0xb0
> > > > fork_trampoline() at fork_trampoline+0x8
> > > > --- trap 0, eip = 0, esp = 0xe2589d70, ebp = 0 ---
> > > > db>
> > > >
> > >
> > > I'm not sure whether this panic is related with recent phk's change
> > > to MEXTADD(). If this is the case, you may have to use standard MTU
> > > instead of 9000.
> > > Since FreeBSD now have physically contiguous jumbos I have plan to
> > > take advantage of it instead of use of local allocator. That would
> > > also eliminate a jlist lock required to serialize accessing jumbo
> > > buffers allocated from driver. Give me a couple of days.
> >
> > Thanks. Reducing the MTU from 9000 to default (1500) circumvents the panic
> .
> >
>
> Would you try attached patch?
Great! The patch fixes jumbo frames. Thanks.
--
Cheers,
Cy Schubert <Cy.Schubert at komquats.com>
FreeBSD UNIX: <cy at FreeBSD.org> Web: http://www.FreeBSD.org
e**(i*pi)+1=0
More information about the freebsd-current
mailing list