Cong Wang <xiyou.wangc...@gmail.com> wrote:
> On Fri, Jan 4, 2019 at 5:19 AM Florian Westphal <f...@strlen.de> wrote:
> >
> > An xfrm hash rebuild has to reset the inexact policy list before the
> > policies get re-inserted: A change of hash thresholds will result in
> > policies to get moved from inexact tree to the policy hash table.
> >
> > If the thresholds are increased again later, they get moved from hash
> > table to inexact tree.
> >
> > We must unlink all policies from the inexact tree before re-insertion.
> >
> > Otherwise 'migrate' may find policies that are in main hash table a
> > second time, when it searches the inexact lists.
> >
> > Furthermore, re-insertion without deletion can cause elements ->next to
> > point back to itself, causing soft lockups or double-frees.
> >
> > Reported-by: syzbot+9d971dd21eb265670...@syzkaller.appspotmail.com
> > Fixes: 9cf545ebd591da ("xfrm: policy: store inexact policies in a tree
> > ordered by destination address")
> > Signed-off-by: Florian Westphal <f...@strlen.de>
> > ---
> > net/xfrm/xfrm_policy.c | 23 ++++++++++-------------
> > 1 file changed, 10 insertions(+), 13 deletions(-)
> >
> > diff --git a/net/xfrm/xfrm_policy.c b/net/xfrm/xfrm_policy.c
> > index d8fba27a4bfb..24dfd1e47cf0 100644
> > --- a/net/xfrm/xfrm_policy.c
> > +++ b/net/xfrm/xfrm_policy.c
> > @@ -680,16 +680,6 @@ static void xfrm_hash_resize(struct work_struct *work)
> > mutex_unlock(&hash_resize_mutex);
> > }
> >
> > -static void xfrm_hash_reset_inexact_table(struct net *net)
> > -{
> > - struct xfrm_pol_inexact_bin *b;
> > -
> > - lockdep_assert_held(&net->xfrm.xfrm_policy_lock);
> > -
> > - list_for_each_entry(b, &net->xfrm.inexact_bins, inexact_bins)
> > - INIT_HLIST_HEAD(&b->hhead);
> > -}
> > -
> > /* Make sure *pol can be inserted into fastbin.
> > * Useful to check that later insert requests will be sucessful
> > * (provided xfrm_policy_lock is held throughout).
> > @@ -1279,10 +1269,14 @@ static void xfrm_hash_rebuild(struct work_struct
> > *work)
> > }
> >
> > /* reset the bydst and inexact table in all directions */
>
>
> Does this comment need to update?
I don't think so, it predates my changes and we still reset the
inexact trees.
What would you suggest the comment should read?
> > + hlist_del_rcu(&policy->bydst);
>
>
> Need to test hlist_unhashed()?
No, all policies are in a hlist, either via the exact hash table or
an inexact tree list.
Also, the _rcu version leaves entry in undefined state so hlist_unhashed
doesn't return true.
