On Fri, Mar 09, 2018 at 06:52:34AM +0800, Boqun Feng wrote:
> On Fri, Mar 09, 2018 at 05:41:27AM +0800, kbuild test robot wrote:
> > tree:
> > https://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu.git
> > rcu/dev
> > head: b8909ec707bb5beba94e7c7d62cc6b3115ceae50
> > commit: b8909ec707bb5beba94e7c7d62cc6b3115ceae50 [39/39] rcu: Protect all
> > sync_rcu_preempt_exp_done() with rcu_node lock
> > reproduce:
> > # apt-get install sparse
> > git checkout b8909ec707bb5beba94e7c7d62cc6b3115ceae50
> > make ARCH=x86_64 allmodconfig
> > make C=1 CF=-D__CHECK_ENDIAN__
> >
> >
> > sparse warnings: (new ones prefixed by >>)
> >
> [...]
> > kernel/rcu/tree.c:345:6: sparse: symbol 'rcu_dynticks_curr_cpu_in_eqs'
> > was not declared. Should it be static?
> > kernel/rcu/tree.c:3953:21: sparse: incorrect type in argument 1
> > (different modifiers) @@ expected int ( *threadfn )( ... ) @@ got int
> > ( [noreint ( *threadfn )( ... ) @@
> > kernel/rcu/tree.c:3953:21: expected int ( *threadfn )( ... )
> > kernel/rcu/tree.c:3953:21: got int ( [noreturn] *<noident> )( ... )
> > >> kernel/rcu/tree_exp.h:163:9: sparse: incorrect type in argument 1
> > >> (different modifiers) @@ expected struct lockdep_map const *lock @@
> > >> got strustruct lockdep_map const *lock @@
> > kernel/rcu/tree_exp.h:163:9: expected struct lockdep_map const *lock
> > kernel/rcu/tree_exp.h:163:9: got struct lockdep_map [noderef]
> > *<noident>
> > kernel/rcu/tree.c:1752:9: sparse: context imbalance in
> > 'rcu_start_future_gp' - different lock contexts for basic block
> > kernel/rcu/tree.c:2786:9: sparse: context imbalance in 'force_qs_rnp' -
> > different lock contexts for basic block
> > kernel/rcu/tree.c:2849:25: sparse: context imbalance in
> > 'force_quiescent_state' - unexpected unlock
> > kernel/rcu/tree_exp.h:203:9: sparse: too many warnings
> >
> > vim +163 kernel/rcu/tree_exp.h
> >
> > 151
> > 152 /*
> > 153 * Return non-zero if there is no RCU expedited grace period in
> > progress
> > 154 * for the specified rcu_node structure, in other words, if all
> > CPUs and
> > 155 * tasks covered by the specified rcu_node structure have done
> > their bit
> > 156 * for the current expedited grace period. Works only for
> > preemptible
> > 157 * RCU -- other RCU implementation use other means.
> > 158 *
> > 159 * Caller must hold the specificed rcu_node structure's ->lock
> > 160 */
> > 161 static bool sync_rcu_preempt_exp_done(struct rcu_node *rnp)
> > 162 {
> > > 163 lockdep_assert_held(&rnp->lock);
>
> OK, so we need ACCESS_PRIVATE() to visit ->lock in rcu_node. I will
> introduce something like:
>
> #define rcu_node_lock_assert_held(rnp)
> lockdep_assert_held(&ACCESS_PRIVATE(rnp, lock))
>
> in v3.
Or use this, which is in kernel/rcu/rcu.h:
#define raw_lockdep_assert_held_rcu_node(p) \
lockdep_assert_held(&ACCESS_PRIVATE(p, lock))
Thanx, Paul
