| Message ID | 1454297472-24261-1-git-send-email-zyjzyj2000@gmail.com |
|---|---|
| State | Changes Requested, archived |
| Delegated to: | David Miller |
| Headers | show |
On 02/01/2016 04:31 AM, zyjzyj2000@gmail.com wrote: > From: Zhu Yanjun <zyjzyj2000@gmail.com> > > rcu_read_lock and rcu_read_unlock are to protect the function > bond_miimon_inspect. As such, moving rcu_read_lock and rcu_read_unlock > to the function bond_miimon_inspect to make the source code compact. > > CC: Jay Vosburgh <j.vosburgh@gmail.com> > CC: Veaceslav Falico <vfalico@gmail.com> > CC: Andy Gospodarek <gospo@cumulusnetworks.com> > Signed-off-by: Zhu Yanjun <zyjzyj2000@gmail.com> > --- > drivers/net/bonding/bond_main.c | 9 ++++----- > 1 file changed, 4 insertions(+), 5 deletions(-) > Not true, RCU also protects the slave dereference in bond_should_notify_peers(). Even though there's already a rcu_read_lock() while doing the dereference itself, it is there only to please RCU, IIRC. The only "simplification" you can do is remove the rcu_read_lock/unlock() around the slave deref in bond_should_notify_peers and use rcu_dereference_rtnl() as it can be used either in RCU protected region or with RTNL held. Also I think net-next is still closed (and that's where this should be targeted at). Cheers, Nik
On 02/01/2016 12:01 PM, Nikolay Aleksandrov wrote: > On 02/01/2016 04:31 AM, zyjzyj2000@gmail.com wrote: >> From: Zhu Yanjun <zyjzyj2000@gmail.com> >> >> rcu_read_lock and rcu_read_unlock are to protect the function >> bond_miimon_inspect. As such, moving rcu_read_lock and rcu_read_unlock >> to the function bond_miimon_inspect to make the source code compact. >> >> CC: Jay Vosburgh <j.vosburgh@gmail.com> >> CC: Veaceslav Falico <vfalico@gmail.com> >> CC: Andy Gospodarek <gospo@cumulusnetworks.com> >> Signed-off-by: Zhu Yanjun <zyjzyj2000@gmail.com> >> --- >> drivers/net/bonding/bond_main.c | 9 ++++----- >> 1 file changed, 4 insertions(+), 5 deletions(-) >> > > Not true, RCU also protects the slave dereference in bond_should_notify_peers(). > Even though there's already a rcu_read_lock() while doing the dereference > itself, it is there only to please RCU, IIRC. The only "simplification" you Err, I meant to please lockdep when being called with RTNL. > can do is remove the rcu_read_lock/unlock() around the slave deref in > bond_should_notify_peers and use rcu_dereference_rtnl() as it can be used either > in RCU protected region or with RTNL held. Also I think net-next is still > closed (and that's where this should be targeted at). > > Cheers, > Nik > >
diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c index 56b5605..4043181 100644 --- a/drivers/net/bonding/bond_main.c +++ b/drivers/net/bonding/bond_main.c @@ -2022,6 +2022,7 @@ static int bond_miimon_inspect(struct bonding *bond) struct slave *slave; bool ignore_updelay; + rcu_read_lock(); ignore_updelay = !rcu_dereference(bond->curr_active_slave); bond_for_each_slave_rcu(bond, slave, iter) { @@ -2113,6 +2114,8 @@ static int bond_miimon_inspect(struct bonding *bond) } } + rcu_read_unlock(); + return commit; } @@ -2231,12 +2234,9 @@ static void bond_mii_monitor(struct work_struct *work) if (!bond_has_slaves(bond)) goto re_arm; - rcu_read_lock(); - should_notify_peers = bond_should_notify_peers(bond); if (bond_miimon_inspect(bond)) { - rcu_read_unlock(); /* Race avoidance with bond_close cancel of workqueue */ if (!rtnl_trylock()) { @@ -2248,8 +2248,7 @@ static void bond_mii_monitor(struct work_struct *work) bond_miimon_commit(bond); rtnl_unlock(); /* might sleep, hold no other locks */ - } else - rcu_read_unlock(); + } re_arm: if (bond->params.miimon)