Re: unsafe locks seen with netperf on net-2.6.29 tree

[Ingo Molnar <mingo@elte.hu>]

* Herbert Xu <herbert@gondor.apana.org.au> wrote:

> On Mon, Dec 29, 2008 at 01:38:19PM +0100, Ingo Molnar wrote:
> >
> > hm, even with the revert i got the splat below. So some other commits are 
> > causing this too?
> 
> Indeed, there is more :)
> 
> > stack backtrace:
> > Pid: 1435, comm: kjournald Not tainted 2.6.28-tip-03883-gf855e6c-dirty #13150
> > Call Trace:
> >  [<c015a0d6>] print_usage_bug+0x176/0x1d0
> >  [<c015b800>] mark_lock+0xbd0/0xd80
> >  [<c015da13>] __lock_acquire+0x483/0xae0
> >  [<c015bcdb>] ? trace_hardirqs_on+0xb/0x10
> >  [<c015e0f9>] lock_acquire+0x89/0xc0
> >  [<c034fc75>] ? __percpu_counter_add+0x65/0xb0
> >  [<c07247a8>] _spin_lock+0x38/0x50
> >  [<c034fc75>] ? __percpu_counter_add+0x65/0xb0
> >  [<c034fc75>] __percpu_counter_add+0x65/0xb0
> >  [<c060dc49>] tcp_v4_destroy_sock+0x1d9/0x240
> 
> This came from 1748376b6626acf59c24e9592ac67b3fe2a0e026 which also
> has the same bug (although this particular trace is bogus and is
> fixed by Peter's first patch).
> 
> I think these are the only two percpu counter patches around that
> time frame.
> 
> Also watch out for 6976a1d6c222c50ac93d2273b9cf57e6fd047e59 when
> reverting.

okay, since i'm not really in the business of reverting various networking 
patches, would you mind to Cc: me to the real fixes, once they are 
available? Or do you think a revert will be the approach in -git? (in 
which case i can guinea-pig the above reverts i guess)

	Ingo