On Wed 21 Aug 2019 at 01:32, Matthew Wilcox <[email protected]> wrote:
> From: "Matthew Wilcox (Oracle)" <[email protected]>
>
> Replace the mutex protecting the IDR with the XArray spinlock.
>
> Signed-off-by: Matthew Wilcox (Oracle) <[email protected]>
> ---
> include/net/act_api.h | 6 +-
> net/sched/act_api.c | 127 +++++++++++++++++-------------------------
> 2 files changed, 53 insertions(+), 80 deletions(-)
>
> diff --git a/include/net/act_api.h b/include/net/act_api.h
[...]
> @@ -290,10 +283,8 @@ static int tcf_del_walker(struct tcf_idrinfo *idrinfo,
> struct sk_buff *skb,
> struct nlattr *nest;
> int n_i = 0;
> int ret = -EINVAL;
> - struct idr *idr = &idrinfo->action_idr;
> struct tc_action *p;
> - unsigned long id = 1;
> - unsigned long tmp;
> + unsigned long index;
>
> nest = nla_nest_start_noflag(skb, 0);
> if (nest == NULL)
> @@ -301,18 +292,18 @@ static int tcf_del_walker(struct tcf_idrinfo *idrinfo,
> struct sk_buff *skb,
> if (nla_put_string(skb, TCA_KIND, ops->kind))
> goto nla_put_failure;
>
> - mutex_lock(&idrinfo->lock);
> - idr_for_each_entry_ul(idr, p, tmp, id) {
> + xa_lock(&idrinfo->actions);
> + xa_for_each(&idrinfo->actions, index, p) {
> ret = tcf_idr_release_unsafe(p);
I like the simplification of reusing builtin xarray spinlock for this,
but the reason we are using mutex here is because the following call
chain: tcf_idr_release_unsafe() -> tcf_action_cleanup() -> free_tcf() ->
tcf_chain_put_by_act() -> __tcf_chain_put() -> mutex_lock(&block->lock);
