| Message ID | 1323941672-14324-3-git-send-email-glommer@parallels.com |
|---|---|
| State | Changes Requested, archived |
| Delegated to: | David Miller |
| Headers | show |
On Thu, 15 Dec 2011 13:34:32 +0400 Glauber Costa <glommer@parallels.com> wrote: > Walking the proto_list holds a read_lock, which prevents us from doing > allocations. Splitting the tcp create function into create + init is > good, but it is not enough since create_files will do allocations as well > (dentry ones, mostly). > > Since this does not involve any protocol state, I propose we call the tcp > functions explicitly from memcontrol.c > > With this, we lose by now the ability of doing cgroup memcontrol for > protocols that are loaded as modules. But at least the ones I have in mind > won't really need it (tcp_ipv6 being the only one, but it uses the same data > structures as tcp_ipv4). So I believe this to be the simpler solution to this > problem. > > Signed-off-by: Glauber Costa <glommer@parallels.com> > CC: Hiroyouki Kamezawa <kamezawa.hiroyu@jp.fujitsu.com> > CC: David S. Miller <davem@davemloft.net> > CC: Eric Dumazet <eric.dumazet@gmail.com> > CC: Stephen Rothwell <sfr@canb.auug.org.au> Could you remake the patch onto the 'latest' linux-next ? As Dave mentioned, some bandaids are already applied and this patch hunks. Thanks, -Kame -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 12/15/2011 08:13 PM, KAMEZAWA Hiroyuki wrote: > On Thu, 15 Dec 2011 13:34:32 +0400 > Glauber Costa<glommer@parallels.com> wrote: > >> Walking the proto_list holds a read_lock, which prevents us from doing >> allocations. Splitting the tcp create function into create + init is >> good, but it is not enough since create_files will do allocations as well >> (dentry ones, mostly). >> >> Since this does not involve any protocol state, I propose we call the tcp >> functions explicitly from memcontrol.c >> >> With this, we lose by now the ability of doing cgroup memcontrol for >> protocols that are loaded as modules. But at least the ones I have in mind >> won't really need it (tcp_ipv6 being the only one, but it uses the same data >> structures as tcp_ipv4). So I believe this to be the simpler solution to this >> problem. >> >> Signed-off-by: Glauber Costa<glommer@parallels.com> >> CC: Hiroyouki Kamezawa<kamezawa.hiroyu@jp.fujitsu.com> >> CC: David S. Miller<davem@davemloft.net> >> CC: Eric Dumazet<eric.dumazet@gmail.com> >> CC: Stephen Rothwell<sfr@canb.auug.org.au> > > Could you remake the patch onto the 'latest' linux-next ? > As Dave mentioned, some bandaids are already applied and this patch hunks. Sure thing. -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
From: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Date: Fri, 16 Dec 2011 01:13:16 +0900 > Could you remake the patch onto the 'latest' linux-next ? > As Dave mentioned, some bandaids are already applied and this patch hunks. He was correct to submit this against net-next. The bandaid in linux-next is temporary and Stephen will remove it once we fix things properly in net-next. -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
From: Glauber Costa <glommer@parallels.com> Date: Thu, 15 Dec 2011 13:34:32 +0400 > Walking the proto_list holds a read_lock, which prevents us from doing > allocations. Splitting the tcp create function into create + init is > good, but it is not enough since create_files will do allocations as well > (dentry ones, mostly). > > Since this does not involve any protocol state, I propose we call the tcp > functions explicitly from memcontrol.c > > With this, we lose by now the ability of doing cgroup memcontrol for > protocols that are loaded as modules. But at least the ones I have in mind > won't really need it (tcp_ipv6 being the only one, but it uses the same data > structures as tcp_ipv4). So I believe this to be the simpler solution to this > problem. > > Signed-off-by: Glauber Costa <glommer@parallels.com> This is an unnecessary limitation, please fix this properly otherwise DCCP, SCTP, etc. won't be supportable with this stuff. -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, 16 Dec 2011 01:13:16 +0900 KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> wrote: > On Thu, 15 Dec 2011 13:34:32 +0400 > Glauber Costa <glommer@parallels.com> wrote: > > > Walking the proto_list holds a read_lock, which prevents us from doing > > allocations. Splitting the tcp create function into create + init is > > good, but it is not enough since create_files will do allocations as well > > (dentry ones, mostly). > > > > Since this does not involve any protocol state, I propose we call the tcp > > functions explicitly from memcontrol.c > > > > With this, we lose by now the ability of doing cgroup memcontrol for > > protocols that are loaded as modules. But at least the ones I have in mind > > won't really need it (tcp_ipv6 being the only one, but it uses the same data > > structures as tcp_ipv4). So I believe this to be the simpler solution to this > > problem. > > > > Signed-off-by: Glauber Costa <glommer@parallels.com> > > CC: Hiroyouki Kamezawa <kamezawa.hiroyu@jp.fujitsu.com> > > CC: David S. Miller <davem@davemloft.net> > > CC: Eric Dumazet <eric.dumazet@gmail.com> > > CC: Stephen Rothwell <sfr@canb.auug.org.au> > > Could you remake the patch onto the 'latest' linux-next ? > As Dave mentioned, some bandaids are already applied and this patch hunks. Applied patches by hand and did small test for hours. seems good. Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 12/16/2011 02:20 AM, KAMEZAWA Hiroyuki wrote: > On Fri, 16 Dec 2011 01:13:16 +0900 > KAMEZAWA Hiroyuki<kamezawa.hiroyu@jp.fujitsu.com> wrote: > >> On Thu, 15 Dec 2011 13:34:32 +0400 >> Glauber Costa<glommer@parallels.com> wrote: >> >>> Walking the proto_list holds a read_lock, which prevents us from doing >>> allocations. Splitting the tcp create function into create + init is >>> good, but it is not enough since create_files will do allocations as well >>> (dentry ones, mostly). >>> >>> Since this does not involve any protocol state, I propose we call the tcp >>> functions explicitly from memcontrol.c >>> >>> With this, we lose by now the ability of doing cgroup memcontrol for >>> protocols that are loaded as modules. But at least the ones I have in mind >>> won't really need it (tcp_ipv6 being the only one, but it uses the same data >>> structures as tcp_ipv4). So I believe this to be the simpler solution to this >>> problem. >>> >>> Signed-off-by: Glauber Costa<glommer@parallels.com> >>> CC: Hiroyouki Kamezawa<kamezawa.hiroyu@jp.fujitsu.com> >>> CC: David S. Miller<davem@davemloft.net> >>> CC: Eric Dumazet<eric.dumazet@gmail.com> >>> CC: Stephen Rothwell<sfr@canb.auug.org.au> >> >> Could you remake the patch onto the 'latest' linux-next ? >> As Dave mentioned, some bandaids are already applied and this patch hunks. > > Applied patches by hand and did small test for hours. > seems good. > > Reviewed-by: KAMEZAWA Hiroyuki<kamezawa.hiroyu@jp.fujitsu.com> > Kame, Thanks. But see Dave's answer to this: He'd like me to follow a slightly different approach (I've attached a patch earlier) -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/include/net/sock.h b/include/net/sock.h index 1df44e2..6cbee80 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -64,8 +64,6 @@ #include <net/dst.h> #include <net/checksum.h> -int mem_cgroup_sockets_init(struct cgroup *cgrp, struct cgroup_subsys *ss); -void mem_cgroup_sockets_destroy(struct cgroup *cgrp, struct cgroup_subsys *ss); /* * This structure really needs to be cleaned up. * Most of it is for TCP, and not used by any of diff --git a/include/net/tcp_memcontrol.h b/include/net/tcp_memcontrol.h index 3512082..f1ff94a 100644 --- a/include/net/tcp_memcontrol.h +++ b/include/net/tcp_memcontrol.h @@ -11,9 +11,26 @@ struct tcp_memcontrol { int tcp_memory_pressure; }; -struct cg_proto *tcp_proto_cgroup(struct mem_cgroup *memcg); +struct cgroup; +struct cgroup_subsys; +#if defined(CONFIG_INET) && defined(CONFIG_CGROUP_MEM_RES_CTLR_KMEM) int tcp_init_cgroup(struct cgroup *cgrp, struct cgroup_subsys *ss); void tcp_destroy_cgroup(struct cgroup *cgrp, struct cgroup_subsys *ss); +void tcp_create_cgroup(struct mem_cgroup *memcg, struct mem_cgroup *parent); +#else +static inline int tcp_init_cgroup(struct cgroup *cgrp, struct cgroup_subsys *ss) +{ + return 0; +} +static inline void tcp_destroy_cgroup(struct cgroup *cgrp, struct cgroup_subsys *ss) +{ +} +static inline void +tcp_create_cgroup(struct mem_cgroup *memcg, struct mem_cgroup *parent) +{ +} +#endif +struct cg_proto *tcp_proto_cgroup(struct mem_cgroup *memcg); unsigned long long tcp_max_memory(const struct mem_cgroup *memcg); void tcp_prot_mem(struct mem_cgroup *memcg, long val, int idx); #endif /* _TCP_MEMCG_H */ diff --git a/mm/memcontrol.c b/mm/memcontrol.c index 7266202..e3d8886 100644 --- a/mm/memcontrol.c +++ b/mm/memcontrol.c @@ -4749,22 +4749,15 @@ static int register_kmem_files(struct cgroup *cont, struct cgroup_subsys *ss) ret = cgroup_add_files(cont, ss, kmem_cgroup_files, ARRAY_SIZE(kmem_cgroup_files)); - /* - * Part of this would be better living in a separate allocation - * function, leaving us with just the cgroup tree population work. - * We, however, depend on state such as network's proto_list that - * is only initialized after cgroup creation. I found the less - * cumbersome way to deal with it to defer it all to populate time - */ if (!ret) - ret = mem_cgroup_sockets_init(cont, ss); + tcp_init_cgroup(cont, ss); return ret; }; static void kmem_cgroup_destroy(struct cgroup_subsys *ss, struct cgroup *cont) { - mem_cgroup_sockets_destroy(cont, ss); + tcp_destroy_cgroup(cont, ss); } #else static int register_kmem_files(struct cgroup *cont, struct cgroup_subsys *ss) @@ -5093,6 +5086,7 @@ mem_cgroup_create(struct cgroup_subsys *ss, struct cgroup *cont) res_counter_init(&memcg->res, &parent->res); res_counter_init(&memcg->memsw, &parent->memsw); res_counter_init(&memcg->kmem, &parent->kmem); + tcp_create_cgroup(memcg, parent); /* * We increment refcnt of the parent to ensure that we can * safely access it on res_counter_charge/uncharge. @@ -5104,6 +5098,7 @@ mem_cgroup_create(struct cgroup_subsys *ss, struct cgroup *cont) res_counter_init(&memcg->res, NULL); res_counter_init(&memcg->memsw, NULL); res_counter_init(&memcg->kmem, NULL); + tcp_create_cgroup(memcg, NULL); } memcg->last_scanned_child = 0; memcg->last_scanned_node = MAX_NUMNODES; diff --git a/net/core/sock.c b/net/core/sock.c index 5de62d3..103f246 100644 --- a/net/core/sock.c +++ b/net/core/sock.c @@ -138,43 +138,6 @@ static DEFINE_RWLOCK(proto_list_lock); static LIST_HEAD(proto_list); -#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM -int mem_cgroup_sockets_init(struct cgroup *cgrp, struct cgroup_subsys *ss) -{ - struct proto *proto; - int ret = 0; - - read_lock(&proto_list_lock); - list_for_each_entry(proto, &proto_list, node) { - if (proto->init_cgroup) { - ret = proto->init_cgroup(cgrp, ss); - if (ret) - goto out; - } - } - - read_unlock(&proto_list_lock); - return ret; -out: - list_for_each_entry_continue_reverse(proto, &proto_list, node) - if (proto->destroy_cgroup) - proto->destroy_cgroup(cgrp, ss); - read_unlock(&proto_list_lock); - return ret; -} - -void mem_cgroup_sockets_destroy(struct cgroup *cgrp, struct cgroup_subsys *ss) -{ - struct proto *proto; - - read_lock(&proto_list_lock); - list_for_each_entry_reverse(proto, &proto_list, node) - if (proto->destroy_cgroup) - proto->destroy_cgroup(cgrp, ss); - read_unlock(&proto_list_lock); -} -#endif - /* * Each address family might have different locking rules, so we have * one slock key per address family: diff --git a/net/ipv4/tcp_memcontrol.c b/net/ipv4/tcp_memcontrol.c index 171d7b6..1433505 100644 --- a/net/ipv4/tcp_memcontrol.c +++ b/net/ipv4/tcp_memcontrol.c @@ -49,7 +49,7 @@ static void memcg_tcp_enter_memory_pressure(struct sock *sk) } EXPORT_SYMBOL(memcg_tcp_enter_memory_pressure); -int tcp_init_cgroup(struct cgroup *cgrp, struct cgroup_subsys *ss) +void tcp_create_cgroup(struct mem_cgroup *memcg, struct mem_cgroup *parent) { /* * The root cgroup does not use res_counters, but rather, @@ -59,13 +59,11 @@ int tcp_init_cgroup(struct cgroup *cgrp, struct cgroup_subsys *ss) struct res_counter *res_parent = NULL; struct cg_proto *cg_proto, *parent_cg; struct tcp_memcontrol *tcp; - struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp); - struct mem_cgroup *parent = parent_mem_cgroup(memcg); struct net *net = current->nsproxy->net_ns; cg_proto = tcp_prot.proto_cgroup(memcg); if (!cg_proto) - goto create_files; + return; tcp = tcp_from_cgproto(cg_proto); @@ -87,8 +85,11 @@ int tcp_init_cgroup(struct cgroup *cgrp, struct cgroup_subsys *ss) cg_proto->memory_allocated = &tcp->tcp_memory_allocated; cg_proto->sockets_allocated = &tcp->tcp_sockets_allocated; cg_proto->memcg = memcg; +} +EXPORT_SYMBOL(tcp_create_cgroup); -create_files: +int tcp_init_cgroup(struct cgroup *cgrp, struct cgroup_subsys *ss) +{ return cgroup_add_files(cgrp, ss, tcp_files, ARRAY_SIZE(tcp_files)); }
Walking the proto_list holds a read_lock, which prevents us from doing allocations. Splitting the tcp create function into create + init is good, but it is not enough since create_files will do allocations as well (dentry ones, mostly). Since this does not involve any protocol state, I propose we call the tcp functions explicitly from memcontrol.c With this, we lose by now the ability of doing cgroup memcontrol for protocols that are loaded as modules. But at least the ones I have in mind won't really need it (tcp_ipv6 being the only one, but it uses the same data structures as tcp_ipv4). So I believe this to be the simpler solution to this problem. Signed-off-by: Glauber Costa <glommer@parallels.com> CC: Hiroyouki Kamezawa <kamezawa.hiroyu@jp.fujitsu.com> CC: David S. Miller <davem@davemloft.net> CC: Eric Dumazet <eric.dumazet@gmail.com> CC: Stephen Rothwell <sfr@canb.auug.org.au> --- include/net/sock.h | 2 -- include/net/tcp_memcontrol.h | 19 ++++++++++++++++++- mm/memcontrol.c | 13 ++++--------- net/core/sock.c | 37 ------------------------------------- net/ipv4/tcp_memcontrol.c | 11 ++++++----- 5 files changed, 28 insertions(+), 54 deletions(-)