| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Date: Sat, 16 May 2015 09:50:11 -0700
From: Eric Dumazet <eric.dumazet@...il.com>
To: David Miller <davem@...hat.com>
Cc: herbert@...dor.apana.org.au, tgraf@...g.ch, netdev@...r.kernel.org,
ying.xue@...driver.com
Subject: Re: netlink: Kill bogus lock_sock in netlink_insert
On Fri, 2015-05-15 at 11:01 -0700, Eric Dumazet wrote:
> On Fri, 2015-05-15 at 12:49 -0400, David Miller wrote:
> > From: Herbert Xu <herbert@...dor.apana.org.au>
> > Date: Thu, 14 May 2015 14:02:30 +0800
> >
> > > On Thu, May 14, 2015 at 01:58:24PM +0800, Herbert Xu wrote:
> > >>
> > >> This patch also removes a bogus socket lock introduced by that
> > >> very same patch.
> > >
> > > This bogus socket lock is still there in the current tree.
> > >
> > > ---8<---
> > > The commit c5adde9468b0714a051eac7f9666f23eb10b61f7 ("netlink:
> > > eliminate nl_sk_hash_lock") added a lock_sock to netlink_insert
> > > with no justifications whatsoever.
> > >
> > > This patch kills it.
> > >
> > > Signed-off-by: Herbert Xu <herbert@...dor.apana.org.au>
> >
> > Eric, does this improve your benchmark?
>
> I don't think so, but I can test this in about one hour.
Maybe the difference between 3.15 and 4.1 is not coming from
rhashtable/netlink but other core stuff, like mutex code.
The workload is largely dominated by mutex acquisition anyway.
88.42% [kernel] [k] osq_lock
1.91% [kernel] [k] mutex_spin_on_owner.isra.4
0.46% [kernel] [k] mutex_optimistic_spin
0.33% [kernel] [k] _raw_spin_lock
0.32% [kernel] [k] inet_dump_ifaddr
0.31% [kernel] [k] inet6_dump_addr
0.23% [kernel] [k] mutex_unlock
0.23% [kernel] [k] _raw_read_lock_bh
0.21% [kernel] [k] _raw_spin_lock_irqsave
0.20% perf [.] 0x0000000000046273
- 88.49% getaddrinfo [kernel.kallsyms] [k] osq_lock ◆
- osq_lock ▒
- 100.00% mutex_optimistic_spin ▒
__mutex_lock_slowpath ▒
- mutex_lock ▒
- 45.48% netlink_dump ▒
+ 87.27% netlink_recvmsg ▒
+ 12.73% __netlink_dump_start ▒
- 38.93% rtnl_lock ▒
+ 52.52% rtnetlink_rcv ▒
+ 47.48% rtnetlink_rcv_msg ▒
- 15.59% __netlink_dump_start ▒
rtnetlink_rcv_msg ▒
netlink_rcv_skb ▒
rtnetlink_rcv ▒
netlink_unicast ▒
netlink_sendmsg ▒
sock_sendmsg ▒
SYSC_sendto ▒
sys_sendto ▒
system_call_fastpath ▒
+ __sendto_nocancel ▒
+ 1.96% getaddrinfo [kernel.kallsyms] [k] mutex_spin_on_owner.isra.4
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists