From: Vlad Buslov <vladbu@mellanox.com>
To: netdev@vger.kernel.org
Cc: jhs@mojatatu.com, xiyou.wangcong@gmail.com, jiri@resnulli.us,
davem@davemloft.net, jakub.kicinski@netronome.com,
pablo@netfilter.org, Vlad Buslov <vladbu@mellanox.com>
Subject: [PATCH net-next v2 06/10] net: sched: conditionally obtain rtnl lock in cls hw offloads API
Date: Fri, 23 Aug 2019 21:50:52 +0300 [thread overview]
Message-ID: <20190823185056.12536-7-vladbu@mellanox.com> (raw)
In-Reply-To: <20190823185056.12536-1-vladbu@mellanox.com>
In order to remove dependency on rtnl lock from offloads code of
classifiers, take rtnl lock conditionally before executing driver
callbacks. Only obtain rtnl lock if block is bound to devices that require
it.
Block bind/unbind code is rtnl-locked and obtains block->cb_lock while
holding rtnl lock. Obtain locks in same order in tc_setup_cb_*() functions
to prevent deadlock.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
---
Changes from V1 to V2:
- Speculatively read block->lockeddevcnt in tc_setup_cb_*() to obtain
rtnl mutex without retry when block is bound to locked device.
net/sched/cls_api.c | 65 +++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 65 insertions(+)
diff --git a/net/sched/cls_api.c b/net/sched/cls_api.c
index 87954f5370a4..f2dcecf34c6f 100644
--- a/net/sched/cls_api.c
+++ b/net/sched/cls_api.c
@@ -3076,11 +3076,28 @@ __tc_setup_cb_call(struct tcf_block *block, enum tc_setup_type type,
int tc_setup_cb_call(struct tcf_block *block, enum tc_setup_type type,
void *type_data, bool err_stop, bool rtnl_held)
{
+ bool take_rtnl = READ_ONCE(block->lockeddevcnt);
int ok_count;
+retry:
+ if (take_rtnl)
+ rtnl_lock();
down_read(&block->cb_lock);
+ /* Need to obtain rtnl lock if block is bound to devs that require it.
+ * In block bind code cb_lock is obtained while holding rtnl, so we must
+ * obtain the locks in same order here.
+ */
+ if (!rtnl_held && !take_rtnl && block->lockeddevcnt) {
+ up_read(&block->cb_lock);
+ take_rtnl = true;
+ goto retry;
+ }
+
ok_count = __tc_setup_cb_call(block, type, type_data, err_stop);
+
up_read(&block->cb_lock);
+ if (take_rtnl)
+ rtnl_unlock();
return ok_count;
}
EXPORT_SYMBOL(tc_setup_cb_call);
@@ -3095,9 +3112,23 @@ int tc_setup_cb_add(struct tcf_block *block, struct tcf_proto *tp,
enum tc_setup_type type, void *type_data, bool err_stop,
u32 *flags, unsigned int *in_hw_count, bool rtnl_held)
{
+ bool take_rtnl = READ_ONCE(block->lockeddevcnt);
int ok_count;
+retry:
+ if (take_rtnl)
+ rtnl_lock();
down_read(&block->cb_lock);
+ /* Need to obtain rtnl lock if block is bound to devs that require it.
+ * In block bind code cb_lock is obtained while holding rtnl, so we must
+ * obtain the locks in same order here.
+ */
+ if (!rtnl_held && !take_rtnl && block->lockeddevcnt) {
+ up_read(&block->cb_lock);
+ take_rtnl = true;
+ goto retry;
+ }
+
/* Make sure all netdevs sharing this block are offload-capable. */
if (block->nooffloaddevcnt && err_stop) {
ok_count = -EOPNOTSUPP;
@@ -3115,6 +3146,8 @@ int tc_setup_cb_add(struct tcf_block *block, struct tcf_proto *tp,
ok_count, true);
err_unlock:
up_read(&block->cb_lock);
+ if (take_rtnl)
+ rtnl_unlock();
return ok_count < 0 ? ok_count : 0;
}
EXPORT_SYMBOL(tc_setup_cb_add);
@@ -3131,9 +3164,23 @@ int tc_setup_cb_replace(struct tcf_block *block, struct tcf_proto *tp,
u32 *new_flags, unsigned int *new_in_hw_count,
bool rtnl_held)
{
+ bool take_rtnl = READ_ONCE(block->lockeddevcnt);
int ok_count;
+retry:
+ if (take_rtnl)
+ rtnl_lock();
down_read(&block->cb_lock);
+ /* Need to obtain rtnl lock if block is bound to devs that require it.
+ * In block bind code cb_lock is obtained while holding rtnl, so we must
+ * obtain the locks in same order here.
+ */
+ if (!rtnl_held && !take_rtnl && block->lockeddevcnt) {
+ up_read(&block->cb_lock);
+ take_rtnl = true;
+ goto retry;
+ }
+
/* Make sure all netdevs sharing this block are offload-capable. */
if (block->nooffloaddevcnt && err_stop) {
ok_count = -EOPNOTSUPP;
@@ -3155,6 +3202,8 @@ int tc_setup_cb_replace(struct tcf_block *block, struct tcf_proto *tp,
new_flags, ok_count, true);
err_unlock:
up_read(&block->cb_lock);
+ if (take_rtnl)
+ rtnl_unlock();
return ok_count < 0 ? ok_count : 0;
}
EXPORT_SYMBOL(tc_setup_cb_replace);
@@ -3167,9 +3216,23 @@ int tc_setup_cb_destroy(struct tcf_block *block, struct tcf_proto *tp,
enum tc_setup_type type, void *type_data, bool err_stop,
u32 *flags, unsigned int *in_hw_count, bool rtnl_held)
{
+ bool take_rtnl = READ_ONCE(block->lockeddevcnt);
int ok_count;
+retry:
+ if (take_rtnl)
+ rtnl_lock();
down_read(&block->cb_lock);
+ /* Need to obtain rtnl lock if block is bound to devs that require it.
+ * In block bind code cb_lock is obtained while holding rtnl, so we must
+ * obtain the locks in same order here.
+ */
+ if (!rtnl_held && !take_rtnl && block->lockeddevcnt) {
+ up_read(&block->cb_lock);
+ take_rtnl = true;
+ goto retry;
+ }
+
ok_count = __tc_setup_cb_call(block, type, type_data, err_stop);
tc_cls_offload_cnt_reset(block, tp, in_hw_count, flags);
@@ -3177,6 +3240,8 @@ int tc_setup_cb_destroy(struct tcf_block *block, struct tcf_proto *tp,
tp->ops->hw_del(tp, type_data);
up_read(&block->cb_lock);
+ if (take_rtnl)
+ rtnl_unlock();
return ok_count < 0 ? ok_count : 0;
}
EXPORT_SYMBOL(tc_setup_cb_destroy);
--
2.21.0
next prev parent reply other threads:[~2019-08-23 18:51 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-08-23 18:50 [PATCH net-next v2 00/10] Refactor cls hardware offload API to support rtnl-independent drivers Vlad Buslov
2019-08-23 18:50 ` [PATCH net-next v2 01/10] net: sched: protect block offload-related fields with rw_semaphore Vlad Buslov
2019-08-23 18:50 ` [PATCH net-next v2 02/10] net: sched: change tcf block offload counter type to atomic_t Vlad Buslov
2019-08-23 18:50 ` [PATCH net-next v2 03/10] net: sched: refactor block offloads counter usage Vlad Buslov
2019-08-24 0:26 ` Jakub Kicinski
2019-08-24 14:44 ` Vlad Buslov
2019-08-23 18:50 ` [PATCH net-next v2 04/10] net: sched: notify classifier on successful offload add/delete Vlad Buslov
2019-08-23 18:50 ` [PATCH net-next v2 05/10] net: sched: add API for registering unlocked offload block callbacks Vlad Buslov
2019-08-23 18:50 ` Vlad Buslov [this message]
2019-08-23 18:50 ` [PATCH net-next v2 07/10] net: sched: take rtnl lock in tc_setup_flow_action() Vlad Buslov
2019-08-23 18:50 ` [PATCH net-next v2 08/10] net: sched: take reference to action dev before calling offloads Vlad Buslov
2019-08-23 18:50 ` [PATCH net-next v2 09/10] net: sched: copy tunnel info when setting flow_action entry->tunnel Vlad Buslov
2019-08-23 18:50 ` [PATCH net-next v2 10/10] net: sched: flower: don't take rtnl lock for cls hw offloads API Vlad Buslov
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=20190823185056.12536-7-vladbu@mellanox.com \
--to=vladbu@mellanox.com \
--cc=davem@davemloft.net \
--cc=jakub.kicinski@netronome.com \
--cc=jhs@mojatatu.com \
--cc=jiri@resnulli.us \
--cc=netdev@vger.kernel.org \
--cc=pablo@netfilter.org \
--cc=xiyou.wangcong@gmail.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).