From: Matthew Wilcox <willy@infradead.org>
To: Vlad Buslov <vladbu@mellanox.com>
Cc: "netdev@vger.kernel.org" <netdev@vger.kernel.org>
Subject: Re: [PATCH 33/38] act_api: Convert action_idr to XArray
Date: Wed, 21 Aug 2019 13:35:14 -0700 [thread overview]
Message-ID: <20190821203514.GA21442@bombadil.infradead.org> (raw)
In-Reply-To: <vbfpnky4884.fsf@mellanox.com>
On Wed, Aug 21, 2019 at 07:41:19PM +0000, Vlad Buslov wrote:
> > @@ -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);
Ah, this one is buggy anyway because we call xa_erase() inside the
loop, so it'll deadlock.
We could just drop the xa_lock() / xa_unlock() around the loop. There's
no need to hold the lock unless we're trying to prevent other simultaneous
additions/deletions, which shouldn't be happening? ie this:
@@ -268,8 +268,10 @@ static int tcf_idr_release_unsafe(struct tc_action *p)
if (atomic_read(&p->tcfa_bindcnt) > 0)
return -EPERM;
- if (refcount_dec_and_test(&p->tcfa_refcnt)) {
- xa_erase(&p->idrinfo->actions, p->tcfa_index);
+ if (refcount_dec_and_lock(&p->tcfa_refcnt,
+ &p->idrinfo->actions.xa_lock)) {
+ __xa_erase(&p->idrinfo->actions, p->tcfa_index);
+ xa_unlock(&p->idrinfo->actions);
tcf_action_cleanup(p);
return ACT_P_DELETED;
}
@@ -292,18 +294,15 @@ 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;
- xa_lock(&idrinfo->actions);
xa_for_each(&idrinfo->actions, index, p) {
ret = tcf_idr_release_unsafe(p);
if (ret == ACT_P_DELETED) {
module_put(ops->owner);
n_i++;
} else if (ret < 0) {
- xa_unlock(&idrinfo->actions);
goto nla_put_failure;
}
}
- xa_unlock(&idrinfo->actions);
if (nla_put_u32(skb, TCA_FCNT, n_i))
goto nla_put_failure;
next prev parent reply other threads:[~2019-08-21 20:35 UTC|newest]
Thread overview: 56+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-08-20 22:32 [PATCH 00/38] Convert networking to use the XArray Matthew Wilcox
2019-08-20 22:32 ` [PATCH 01/38] mlx4: Convert cq_table->tree to XArray Matthew Wilcox
2019-08-20 22:32 ` [PATCH 02/38] mlx4: Convert srq_table->tree " Matthew Wilcox
2019-08-20 22:32 ` [PATCH 03/38] mlx4: Convert qp_table_tree " Matthew Wilcox
2019-08-27 19:18 ` Saeed Mahameed
2019-08-20 22:32 ` [PATCH 04/38] mlx5: Convert cq_table " Matthew Wilcox
2019-08-27 19:22 ` Saeed Mahameed
2019-08-20 22:32 ` [PATCH 05/38] mlx5: Convert mlx5_qp_table " Matthew Wilcox
2019-08-20 22:32 ` [PATCH 06/38] mlx5: Convert counters_idr " Matthew Wilcox
2019-08-20 22:32 ` [PATCH 07/38] mlx5: Convert fpga IDRs " Matthew Wilcox
2019-08-20 22:32 ` [PATCH 08/38] nfp: Convert " Matthew Wilcox
2019-08-21 3:59 ` Jakub Kicinski
2019-08-20 22:32 ` [PATCH 09/38] ath10k: Convert pending_tx " Matthew Wilcox
2019-08-20 22:32 ` [PATCH 10/38] ath10k: Convert mgmt_pending_tx IDR " Matthew Wilcox
2019-08-20 22:32 ` [PATCH 11/38] mt76: Convert token " Matthew Wilcox
2019-08-20 22:32 ` [PATCH 12/38] mwifiex: Convert ack_status_frames " Matthew Wilcox
2019-08-20 22:32 ` [PATCH 13/38] ppp: Convert units_idr " Matthew Wilcox
2019-08-20 22:32 ` [PATCH 14/38] tap: Convert minor_idr " Matthew Wilcox
2019-08-20 22:32 ` [PATCH 15/38] nfp: Convert internal ports " Matthew Wilcox
2019-08-20 22:32 ` [PATCH 16/38] qrtr: Convert qrtr_nodes " Matthew Wilcox
2019-08-20 22:32 ` [PATCH 17/38] qrtr: Convert qrtr_ports " Matthew Wilcox
2019-08-20 22:32 ` [PATCH 18/38] rxrpc: Convert " Matthew Wilcox
2019-08-20 22:32 ` [PATCH 19/38] 9p: Convert reqs IDR " Matthew Wilcox
2019-08-20 22:32 ` [PATCH 20/38] 9p: Convert fids " Matthew Wilcox
2019-08-20 22:32 ` [PATCH 21/38] 9p: Move lock from client to trans_fd Matthew Wilcox
2019-08-20 22:32 ` [PATCH 22/38] sctp: Convert sctp_assocs_id to XArray Matthew Wilcox
2019-08-20 22:32 ` [PATCH 23/38] cls_api: Convert tcf_net " Matthew Wilcox
2019-08-20 23:57 ` David Miller
2019-08-21 0:52 ` Matthew Wilcox
2019-08-20 22:32 ` [PATCH 24/38] cls_u32: Convert tc_u_common->handle_idr " Matthew Wilcox
2019-08-21 21:13 ` Jakub Kicinski
2019-08-21 21:25 ` Matthew Wilcox
2019-08-21 21:38 ` Jakub Kicinski
2019-08-20 22:32 ` [PATCH 25/38] cls_u32: Convert tc_u_hnode->handle_idr " Matthew Wilcox
2019-08-20 22:32 ` [PATCH 26/38] cls_bpf: Convert handle_idr " Matthew Wilcox
2019-08-20 22:32 ` [PATCH 27/38] cls_bpf: Remove list of programs Matthew Wilcox
2019-08-20 22:32 ` [PATCH 28/38] cls_bpf: Use XArray marks to accelerate re-offload Matthew Wilcox
2019-08-20 22:32 ` [PATCH 29/38] cls_flower: Convert handle_idr to XArray Matthew Wilcox
2019-08-20 23:58 ` David Miller
2019-08-21 0:50 ` Matthew Wilcox
2019-08-21 18:27 ` Vlad Buslov
2019-08-25 18:32 ` Cong Wang
2019-08-26 10:11 ` Vlad Buslov
2019-08-20 22:32 ` [PATCH 30/38] cls_flower: Use XArray list of filters in fl_walk Matthew Wilcox
2019-08-21 18:32 ` Vlad Buslov
2019-08-20 22:32 ` [PATCH 31/38] cls_flower: Use XArray marks instead of separate list Matthew Wilcox
2019-08-21 19:12 ` Vlad Buslov
2019-08-20 22:32 ` [PATCH 32/38] cls_basic: Convert handle_idr to XArray Matthew Wilcox
2019-08-20 22:32 ` [PATCH 33/38] act_api: Convert action_idr " Matthew Wilcox
2019-08-21 19:41 ` Vlad Buslov
2019-08-21 20:35 ` Matthew Wilcox [this message]
2019-08-20 22:32 ` [PATCH 34/38] net_namespace: Convert netns_ids " Matthew Wilcox
2019-08-20 22:32 ` [PATCH 35/38] tipc: Convert conn_idr " Matthew Wilcox
2019-08-20 22:32 ` [PATCH 36/38] netlink: Convert genl_fam_idr " Matthew Wilcox
2019-08-20 22:32 ` [PATCH 37/38] mac80211: Convert ack_status_frames " Matthew Wilcox
2019-08-20 22:32 ` [PATCH 38/38] mac80211: Convert function_inst_ids " Matthew Wilcox
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20190821203514.GA21442@bombadil.infradead.org \
--to=willy@infradead.org \
--cc=netdev@vger.kernel.org \
--cc=vladbu@mellanox.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).