Messages in this thread | | | Subject | Re: [RFC][PATCH 7/7] lockdep: spin_lock_nest_lock() | From | Peter Zijlstra <> | Date | Mon, 04 Aug 2008 18:38:55 +0200 |
| |
On Mon, 2008-08-04 at 18:26 +0200, Andrea Arcangeli wrote: > On Mon, Aug 04, 2008 at 04:56:03PM +0200, Peter Zijlstra wrote: > > My guess is that the kvm thing attaches before there are any vma, and > > leaves after all the vma are gone. So it would never actually trigger. > > Yes, lockdep seems to be fine with kvm in kernel mainline in my > current and past testing. Andrew asked me to check this long ago. > > vmx ~ # zgrep LOCKDEP /proc/config.gz > CONFIG_LOCKDEP_SUPPORT=y > CONFIG_LOCKDEP=y > CONFIG_DEBUG_LOCKDEP=y
You also need CONFIG_PROVE_LOCKING, but I'll assume that's set too.
> CPA self-test: > 4k 26112 large 1997 gb 0 x 2652[ffff880000000000-ffff8800bffff000] > miss 262144 > 4k 186880 large 1683 gb 0 x 43021[ffff880000000000-ffff8800bffff000] > miss 262144 > 4k 186880 large 1683 gb 0 x 43021[ffff880000000000-ffff8800bffff000] > miss 262144 > ok. > loaded kvm module () > apic write: bad size=1 fee00030 > Ignoring de-assert INIT to vcpu 0 > Ignoring de-assert INIT to vcpu 0 > kvm: emulating exchange as write > apic write: bad size=1 fee00030 > Ignoring de-assert INIT to vcpu 0 > Ignoring de-assert INIT to vcpu 0 > apic write: bad size=1 fee00030 > Ignoring de-assert INIT to vcpu 0 > Ignoring de-assert INIT to vcpu 0 > apic write: bad size=1 fee00030 > Ignoring de-assert INIT to vcpu 0 > apic write: bad size=1 fee00030 > Ignoring de-assert INIT to vcpu 0 > apic write: bad size=1 fee00030 > Ignoring de-assert INIT to vcpu 0 > Ignoring de-assert INIT to vcpu 0 > Ignoring de-assert INIT to vcpu 0 > > I can't see lockdep errors in dmesg starting one more multiple VM in a > loop (all run on a quadcore).
Dave Jones just handed me:
https://bugzilla.redhat.com/show_bug.cgi?id=457779
> GRU is likely the same. > > The only real place where lockdep is unusable in my experience is > preempt-RT, it grinds it to an halt during boot on 8-way before > reaching the shell.
David Miller just did a patch that might fix that.
| |