Re: towards mpsafe rtfree(9)

2015-10-23 Thread Alexander Bluhm
On Fri, Oct 23, 2015 at 02:41:34PM +0200, Martin Pieuchot wrote: > On 22/10/15(Thu) 19:42, Martin Pieuchot wrote: > > Now that we have a single refcounting mechanism for route entries, I'd > > like to use atomic operations and grab the KERNEL_LOCK only if a CPU is > > dropping the last reference

Re: towards mpsafe rtfree(9)

2015-10-23 Thread Martin Pieuchot
On 22/10/15(Thu) 19:42, Martin Pieuchot wrote: > Now that we have a single refcounting mechanism for route entries, I'd > like to use atomic operations and grab the KERNEL_LOCK only if a CPU is > dropping the last reference on an entry. > > Currently this only matters for MPLS. I intentionally

Re: towards mpsafe rtfree(9)

2015-10-23 Thread Martin Pieuchot
On 22/10/15(Thu) 20:41, Bret Lambert wrote: > On Thu, Oct 22, 2015 at 07:42:24PM +0200, Martin Pieuchot wrote: > > Now that we have a single refcounting mechanism for route entries, I'd > > like to use atomic operations and grab the KERNEL_LOCK only if a CPU is > > dropping the last reference on

Re: towards mpsafe rtfree(9)

2015-10-22 Thread Bret Lambert
On Thu, Oct 22, 2015 at 07:42:24PM +0200, Martin Pieuchot wrote: > Now that we have a single refcounting mechanism for route entries, I'd > like to use atomic operations and grab the KERNEL_LOCK only if a CPU is > dropping the last reference on an entry. > > Currently this only matters for MPLS.

towards mpsafe rtfree(9)

2015-10-22 Thread Martin Pieuchot
Now that we have a single refcounting mechanism for route entries, I'd like to use atomic operations and grab the KERNEL_LOCK only if a CPU is dropping the last reference on an entry. Currently this only matters for MPLS. I intentionally use atomic_* ops because I'd like to see be able to see if