Re: [2/4] [IPSEC] Kill spurious hard expire messages

[Herbert Xu <herbert@xxxxxxxxxxxxxxxxxxx>]

On Mon, Apr 11, 2005 at 07:57:29AM -0400, jamal wrote:
> Herbert,
> 
> Ok, what you are saying is sensible, but: does this mean there was never
> an issue then there was never a possibility of delete and expire
> intefering with each other then?

The problem before was that the timer didn't check the return value
of xfrm_policy_delete (in fact the latter didn't return anything).
xfrm_policy_delete tself always checked whether it succeeded in
deleting the entry from xfrm_policy_list.  This is used to determine
whether it should call xfrm_policy_kill.

Not checking the return value in the timer allows the following
situation to occur:

CPU 0                           CPU 1
xfrm_policy_timer
        goto expired
                                xfrm_get_policy(delete == 1)
                                        xfrm_policy_byid() succeeds
                                        notification is sent
        xfrm_policy_delete
        km_policy_expired <--- BUG

So as long as xfrm_policy_delete returns the correct value and it's
checked by the xfrm_policy_timer then we're OK.

Cheers,
-- 
Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <herbert@xxxxxxxxxxxxxxxxxxx>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt