| Message ID | 20180627221642.247448-1-shakeelb@google.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 28.06.2018 01:16, Shakeel Butt wrote: > Currently the kernel accounts the memory for network traffic through > mem_cgroup_[un]charge_skmem() interface. However the memory accounted > only includes the truesize of sk_buff which does not include the size of > sock objects. In our production environment, with opt-out kmem > accounting, the sock kmem caches (TCP[v6], UDP[v6], RAW[v6], UNIX) are > among the top most charged kmem caches and consume a significant amount > of memory which can not be left as system overhead. So, this patch > converts the kmem caches of all sock objects to SLAB_ACCOUNT. > > Signed-off-by: Shakeel Butt <shakeelb@google.com> > Suggested-by: Eric Dumazet <edumazet@google.com> Looks good for me. Reviewed-by: Kirill Tkhai <ktkhai@virtuozzo.com> > --- > Changelog since v1: > - Instead of specific sock kmem_caches, convert all sock kmem_caches to > use SLAB_ACCOUNT. > > net/core/sock.c | 7 +++++-- > 1 file changed, 5 insertions(+), 2 deletions(-) > > diff --git a/net/core/sock.c b/net/core/sock.c > index bcc41829a16d..9e8f65585b81 100644 > --- a/net/core/sock.c > +++ b/net/core/sock.c > @@ -3243,7 +3243,8 @@ static int req_prot_init(const struct proto *prot) > > rsk_prot->slab = kmem_cache_create(rsk_prot->slab_name, > rsk_prot->obj_size, 0, > - prot->slab_flags, NULL); > + SLAB_ACCOUNT | prot->slab_flags, > + NULL); > > if (!rsk_prot->slab) { > pr_crit("%s: Can't create request sock SLAB cache!\n", > @@ -3258,7 +3259,8 @@ int proto_register(struct proto *prot, int alloc_slab) > if (alloc_slab) { > prot->slab = kmem_cache_create_usercopy(prot->name, > prot->obj_size, 0, > - SLAB_HWCACHE_ALIGN | prot->slab_flags, > + SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT | > + prot->slab_flags, > prot->useroffset, prot->usersize, > NULL); > > @@ -3281,6 +3283,7 @@ int proto_register(struct proto *prot, int alloc_slab) > kmem_cache_create(prot->twsk_prot->twsk_slab_name, > prot->twsk_prot->twsk_obj_size, > 0, > + SLAB_ACCOUNT | > prot->slab_flags, > NULL); > if (prot->twsk_prot->twsk_slab == NULL) >
On 06/27/2018 03:16 PM, Shakeel Butt wrote: > Currently the kernel accounts the memory for network traffic through > mem_cgroup_[un]charge_skmem() interface. However the memory accounted > only includes the truesize of sk_buff which does not include the size of > sock objects. In our production environment, with opt-out kmem > accounting, the sock kmem caches (TCP[v6], UDP[v6], RAW[v6], UNIX) are > among the top most charged kmem caches and consume a significant amount > of memory which can not be left as system overhead. So, this patch > converts the kmem caches of all sock objects to SLAB_ACCOUNT. > > Signed-off-by: Shakeel Butt <shakeelb@google.com> > Suggested-by: Eric Dumazet <edumazet@google.com> > --- Reviewed-by: Eric Dumazet <edumazet@google.com> Thanks !
diff --git a/net/core/sock.c b/net/core/sock.c index bcc41829a16d..9e8f65585b81 100644 --- a/net/core/sock.c +++ b/net/core/sock.c @@ -3243,7 +3243,8 @@ static int req_prot_init(const struct proto *prot) rsk_prot->slab = kmem_cache_create(rsk_prot->slab_name, rsk_prot->obj_size, 0, - prot->slab_flags, NULL); + SLAB_ACCOUNT | prot->slab_flags, + NULL); if (!rsk_prot->slab) { pr_crit("%s: Can't create request sock SLAB cache!\n", @@ -3258,7 +3259,8 @@ int proto_register(struct proto *prot, int alloc_slab) if (alloc_slab) { prot->slab = kmem_cache_create_usercopy(prot->name, prot->obj_size, 0, - SLAB_HWCACHE_ALIGN | prot->slab_flags, + SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT | + prot->slab_flags, prot->useroffset, prot->usersize, NULL); @@ -3281,6 +3283,7 @@ int proto_register(struct proto *prot, int alloc_slab) kmem_cache_create(prot->twsk_prot->twsk_slab_name, prot->twsk_prot->twsk_obj_size, 0, + SLAB_ACCOUNT | prot->slab_flags, NULL); if (prot->twsk_prot->twsk_slab == NULL)
Currently the kernel accounts the memory for network traffic through mem_cgroup_[un]charge_skmem() interface. However the memory accounted only includes the truesize of sk_buff which does not include the size of sock objects. In our production environment, with opt-out kmem accounting, the sock kmem caches (TCP[v6], UDP[v6], RAW[v6], UNIX) are among the top most charged kmem caches and consume a significant amount of memory which can not be left as system overhead. So, this patch converts the kmem caches of all sock objects to SLAB_ACCOUNT. Signed-off-by: Shakeel Butt <shakeelb@google.com> Suggested-by: Eric Dumazet <edumazet@google.com> --- Changelog since v1: - Instead of specific sock kmem_caches, convert all sock kmem_caches to use SLAB_ACCOUNT. net/core/sock.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)