Skip to contentSkip to conversation

rebalance locks by converting write_lock_bh to write_lock+local_bh_disable

PATCH In review Nicholas Mc Guire 3 participants 5 messages Nov 22, 2013 – Nov 25, 2013

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 rebalance locks by converting write_lock_bh to write_lock+local_bh_disable

Quoted 1 line

From 2c8e669b691b825c0ed2a02bd7a698d8ed5c6d29 Mon Sep 17 00:00:00 2001

From: Nicholas Mc Guire <[email protected]>
Date: Thu, 21 Nov 2013 18:22:55 -0500
Subject: [PATCH] rebalance locks by converting write_lock_bh to write_lock+local_bh_disable

in __neigh_event_send write_lock_bh(&neigh->lock) is implicitly balanced by write_unlock(&neigh->lock)+local_bh_disable() - while this is equivalent with respect to the effective low level locking primitives it breaks balancing in the locking api. This makes automatic lock-checking trigger false positives, creates an implicit dependency between *_lock_bh and *_lock functions as well as making the extremly simply locking of net core even easier to understand.

The api inbalance was introduced in:
commit cd28ca0a3dd17c68d24b839602a0e6268ad28b5d
Author: Eric Dumazet <[email protected]>
This patch just rebalances the lock api

No change of functionality

---
 net/core/neighbour.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)
net/core/neighbour.c +2 -1
index ca15f32..d681c75 100644
@@ -966,7 +966,8 @@ int __neigh_event_send(struct neighbour *neigh, struct sk_buff *skb)
 	int rc;
 	bool immediate_probe = false;
 
-	write_lock_bh(&neigh->lock);
+	local_bh_disable();
+	write_lock(&neigh->lock);
 
 	rc = 0;
 	if (neigh->nud_state & (NUD_CONNECTED | NUD_DELAY | NUD_PROBE))
Signature · 1 line
1.7.2.5
maintainer
#1 lore raw

From: Nicholas Mc Guire <[email protected]>
Date: Fri, 22 Nov 2013 00:54:02 +0100

Quoted 14 lines

From 2c8e669b691b825c0ed2a02bd7a698d8ed5c6d29 Mon Sep 17 00:00:00 2001
From: Nicholas Mc Guire <[email protected]>
Date: Thu, 21 Nov 2013 18:22:55 -0500
Subject: [PATCH] rebalance locks by converting write_lock_bh to write_lock+local_bh_disable

in __neigh_event_send write_lock_bh(&neigh->lock) is implicitly balanced by write_unlock(&neigh->lock)+local_bh_disable() - while this is equivalent with respect to the effective low level locking primitives it breaks balancing in the locking api. This makes automatic lock-checking trigger false positives, creates an implicit dependency between *_lock_bh and *_lock functions as well as making the extremly simply locking of net core even easier to understand.

Quoted 8 lines

The api inbalance was introduced in:
commit cd28ca0a3dd17c68d24b839602a0e6268ad28b5d
Author: Eric Dumazet <[email protected]>
This patch just rebalances the lock api

No change of functionality

This is a valid locking idiom, fix the lock checking.

#2 lore raw
Quoted 22 lines · David Miller

From: Nicholas Mc Guire <[email protected]>
Date: Fri, 22 Nov 2013 00:54:02 +0100

Quoted 19 lines

From 2c8e669b691b825c0ed2a02bd7a698d8ed5c6d29 Mon Sep 17 00:00:00 2001
From: Nicholas Mc Guire <[email protected]>
Date: Thu, 21 Nov 2013 18:22:55 -0500
Subject: [PATCH] rebalance locks by converting write_lock_bh to write_lock+local_bh_disable

in __neigh_event_send write_lock_bh(&neigh->lock) is implicitly balanced by write_unlock(&neigh->lock)+local_bh_disable() - while this is equivalent with respect to the effective low level locking primitives it breaks balancing in the locking api. This makes automatic lock-checking trigger false positives, creates an implicit dependency between *_lock_bh and *_lock functions as well as making the extremly simply locking of net core even easier to understand.

The api inbalance was introduced in:
commit cd28ca0a3dd17c68d24b839602a0e6268ad28b5d
Author: Eric Dumazet <[email protected]>
This patch just rebalances the lock api

Quoted 5 lines
Quoted 3 lines

No change of functionality

This is a valid locking idiom, fix the lock checking.

for lock checking that is doable but what is with the api coupling and readability ?

any change you do to the spin_lock_bh/spin_unlock_bh side would need to also
take care of the spin_lock/spin_unlock variance and keep them functionally
equivalent - currently there is a very small number of such inbalances in
place it seems (scan of 3.12.1 found 1 write_lock/write_lock_bh,
2 spin_lock/spin_lock_bh, 0 in read_lock/read_lock_bh) so is this idiomatic extension sensible given that it introduces implicit api-coupling ?

in one of the cases I do not understand the intent behind the split:
in net/core/sock.c:lock_sock_fast

	spin_lock_bh(&sk->sk_lock.slock);
	...
        spin_unlock(&sk->sk_lock.slock);
        /*
         * The sk_lock has mutex_lock() semantics here:
         */
        mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
        local_bh_enable();

I think that

	spin_lock_bh(&sk->sk_lock.slock);
	...
        /*
         * The sk_lock has mutex_lock() semantics here:
         */
        mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
        spin_unlock_bh(&sk->sk_lock.slock);

should be equivalent ?

thx!
hofrat

#3 lore raw

RE: [PATCH] rebalance locks by converting write_lock_bh towrite_lock+local_bh_disable

Quoted 22 lines

in one of the cases I do not understand the intent behind the split:
in net/core/sock.c:lock_sock_fast

	spin_lock_bh(&sk->sk_lock.slock);
	...
        spin_unlock(&sk->sk_lock.slock);
        /*
         * The sk_lock has mutex_lock() semantics here:
         */
        mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
        local_bh_enable();

I think that

	spin_lock_bh(&sk->sk_lock.slock);
	...
        /*
         * The sk_lock has mutex_lock() semantics here:
         */
        mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
        spin_unlock_bh(&sk->sk_lock.slock);
Quoted 1 line

should be equivalent ?

You've added a lock ordering that wasn't there before. Also I suspect that mutex_acquire() might be allowed to sleep, whereas you shouldn't sleep with a spin lock held.

David

#4 lore raw

Re: [PATCH] rebalance locks by converting write_lock_bh towrite_lock+local_bh_disable

Quoted 22 lines · David Laight
Quoted 22 lines

in one of the cases I do not understand the intent behind the split:
in net/core/sock.c:lock_sock_fast

	spin_lock_bh(&sk->sk_lock.slock);
	...
        spin_unlock(&sk->sk_lock.slock);
        /*
         * The sk_lock has mutex_lock() semantics here:
         */
        mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
        local_bh_enable();

I think that

	spin_lock_bh(&sk->sk_lock.slock);
	...
        /*
         * The sk_lock has mutex_lock() semantics here:
         */
        mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
        spin_unlock_bh(&sk->sk_lock.slock);
Quoted 6 lines
Quoted 1 line

should be equivalent ?

You've added a lock ordering that wasn't there before. Also I suspect that mutex_acquire() might be allowed to sleep, whereas you shouldn't sleep with a spin lock held.

mutex_acquire is not a lock but a lockdep entry.
As far as I understand it it nither can sleep nor is there a change of order here. Am I missing something ?

thx!
hofrat