Message ID | 1434734269-4545-4-git-send-email-pablo@netfilter.org |
---|---|
State | Accepted, archived |
Delegated to: | David Miller |
Headers | show |
From: Pablo Neira Ayuso > Sent: 19 June 2015 18:18 > From: Eric Dumazet <edumazet@google.com> > > Let's force a 16 bytes alignment on xt_counter percpu allocations, > so that bytes and packets sit in same cache line. > > xt_counter being exported to user space, we cannot add __align(16) on > the structure itself. You could allocate a wrapper structure that contains xt_counter and is itself marked __align(16). David -- To unsubscribe from this list: send the line "unsubscribe netdev" in
On Mon, 2015-06-22 at 11:49 +0000, David Laight wrote: > From: Pablo Neira Ayuso > > Sent: 19 June 2015 18:18 > > From: Eric Dumazet <edumazet@google.com> > > > > Let's force a 16 bytes alignment on xt_counter percpu allocations, > > so that bytes and packets sit in same cache line. > > > > xt_counter being exported to user space, we cannot add __align(16) on > > the structure itself. > > You could allocate a wrapper structure that contains xt_counter > and is itself marked __align(16). Sure. I took one solution, out of multiple ones. -- To unsubscribe from this list: send the line "unsubscribe netdev" in
diff --git a/include/linux/netfilter/x_tables.h b/include/linux/netfilter/x_tables.h index 95693c4..1c97a22 100644 --- a/include/linux/netfilter/x_tables.h +++ b/include/linux/netfilter/x_tables.h @@ -356,7 +356,8 @@ static inline unsigned long ifname_compare_aligned(const char *_a, * so nothing needs to be done there. * * xt_percpu_counter_alloc returns the address of the percpu - * counter, or 0 on !SMP. + * counter, or 0 on !SMP. We force an alignment of 16 bytes + * so that bytes/packets share a common cache line. * * Hence caller must use IS_ERR_VALUE to check for error, this * allows us to return 0 for single core systems without forcing @@ -365,7 +366,8 @@ static inline unsigned long ifname_compare_aligned(const char *_a, static inline u64 xt_percpu_counter_alloc(void) { if (nr_cpu_ids > 1) { - void __percpu *res = alloc_percpu(struct xt_counters); + void __percpu *res = __alloc_percpu(sizeof(struct xt_counters), + sizeof(struct xt_counters)); if (res == NULL) return (u64) -ENOMEM;