Re: [PATCH net] netfilter: conntrack: avoid possible false sharing

2019-10-09 Thread Eric Dumazet
On 10/9/19 9:24 PM, Jakub Kicinski wrote: > On Wed, 9 Oct 2019 09:19:13 -0700, Eric Dumazet wrote: >> As hinted by KCSAN, we need at least one READ_ONCE() >> to prevent a compiler optimization. >> >> More details on : >> https://github.com/google/ktsan/wiki/READ_ONCE-and-WRITE_ONCE#it-may-impro

Re: [PATCH net] netfilter: conntrack: avoid possible false sharing

2019-10-09 Thread Jakub Kicinski
On Wed, 9 Oct 2019 09:19:13 -0700, Eric Dumazet wrote: > As hinted by KCSAN, we need at least one READ_ONCE() > to prevent a compiler optimization. > > More details on : > https://github.com/google/ktsan/wiki/READ_ONCE-and-WRITE_ONCE#it-may-improve-performance > > [...] > > Fixes: cc16921351d8

Re: [PATCH net] netfilter: conntrack: avoid possible false sharing

2019-10-09 Thread Pablo Neira Ayuso
On Wed, Oct 09, 2019 at 09:19:13AM -0700, Eric Dumazet wrote: > As hinted by KCSAN, we need at least one READ_ONCE() > to prevent a compiler optimization. > > More details on : > https://github.com/google/ktsan/wiki/READ_ONCE-and-WRITE_ONCE#it-may-improve-performance > > sysbot report : > BUG: KC

[PATCH net] netfilter: conntrack: avoid possible false sharing

2019-10-09 Thread Eric Dumazet
As hinted by KCSAN, we need at least one READ_ONCE() to prevent a compiler optimization. More details on : https://github.com/google/ktsan/wiki/READ_ONCE-and-WRITE_ONCE#it-may-improve-performance sysbot report : BUG: KCSAN: data-race in __nf_ct_refresh_acct / __nf_ct_refresh_acct read to 0x8