| Message ID | 20240919142149.282175-1-yyyynoom@gmail.com (mailing list archive) |
|---|---|
| State | Changes Requested |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [net] net: add inline annotation to fix the build warning | expand |
On Thu, Sep 19, 2024 at 11:21:49PM +0900, Moon Yeounsu wrote: > This patch fixes two sparse warnings (`make C=1`): > net/ipv6/icmp.c:103:20: warning: context imbalance in 'icmpv6_xmit_lock' - wrong count at exit > net/ipv6/icmp.c:119:13: warning: context imbalance in 'icmpv6_xmit_unlock' - unexpected unlock > > Since `icmp6_xmit_lock()` and `icmp6_xmit_unlock()` are designed as they > are named, entering/returning the function without lock/unlock doesn't > matter. > > Signed-off-by: Moon Yeounsu <yyyynoom@gmail.com> Hi Moon, Without this patch applied I see the warnings cited above. However, with this patch applied, I see the following. So I think this needs more work. net/ipv6/icmp.c: note: in included file (through include/linux/sched.h, include/linux/percpu.h, arch/x86/include/asm/msr.h, arch/x86/include/asm/tsc.h, arch/x86/include/asm/timex.h, include/linux/timex.h, ...): ./include/linux/spinlock.h:361:16: warning: context imbalance in 'icmpv6_xmit_lock' - different lock contexts for basic block net/ipv6/icmp.c: note: in included file (through include/linux/spinlock.h, include/linux/sched.h, include/linux/percpu.h, arch/x86/include/asm/msr.h, arch/x86/include/asm/tsc.h, arch/x86/include/asm/timex.h, ...): ./include/linux/bottom_half.h:33:30: warning: context imbalance in 'icmp6_send' - different lock contexts for basic block ./include/linux/bottom_half.h:33:30: warning: context imbalance in 'icmpv6_echo_reply' - different lock contexts for basic block Also, It is my feeling that addressing warnings of this nature is not a fix for net, but rather but rather an enhancement for net-next. net-next is currently closed for the v6.12 merge windows, so non-RFC, patches should not be posted for net-next until it re-opens once v6.12-rc1 has been released, most likely during the week of 30th September.
diff --git a/net/ipv6/icmp.c b/net/ipv6/icmp.c index 071b0bc1179d..d8cc3d63c942 100644 --- a/net/ipv6/icmp.c +++ b/net/ipv6/icmp.c @@ -101,6 +101,7 @@ static const struct inet6_protocol icmpv6_protocol = { /* Called with BH disabled */ static struct sock *icmpv6_xmit_lock(struct net *net) + __acquires(&sk->sk_lock.slock) { struct sock *sk; @@ -117,6 +118,7 @@ static struct sock *icmpv6_xmit_lock(struct net *net) } static void icmpv6_xmit_unlock(struct sock *sk) + __releases(&sk->sk_lock.slock) { sock_net_set(sk, &init_net); spin_unlock(&sk->sk_lock.slock);
This patch fixes two sparse warnings (`make C=1`): net/ipv6/icmp.c:103:20: warning: context imbalance in 'icmpv6_xmit_lock' - wrong count at exit net/ipv6/icmp.c:119:13: warning: context imbalance in 'icmpv6_xmit_unlock' - unexpected unlock Since `icmp6_xmit_lock()` and `icmp6_xmit_unlock()` are designed as they are named, entering/returning the function without lock/unlock doesn't matter. Signed-off-by: Moon Yeounsu <yyyynoom@gmail.com> --- net/ipv6/icmp.c | 2 ++ 1 file changed, 2 insertions(+)