sock_map: clean up and refactor code for BPF_SK_SKB_VERDICT
Subsystems (11)
v3 cover sock_map: clean up and refactor code for BPF_SK_SKB_VERDICT
From: Cong Wang <[email protected]>
This patchset is the first series of patches separated out from the original large patchset, to make reviews easier. This patchset does not add any new feature or change any functionality but merely cleans up the existing sockmap and skmsg code and refactors it, to prepare for the patches followed up. This passed all BPF selftests.
The original whole patchset is available on github:
https://github.com/congwang/linux/tree/sockmap
and this patchset is also available on github:
https://github.com/congwang/linux/tree/sockmap1
---
v3: fix a few Kconfig compile errors
remove an unused variable
add a comment for bpf_convert_data_end_access()
v2: split the original patchset
compute data_end with bpf_convert_data_end_access() get rid of psock->bpf_running
reduce the scope of CONFIG_BPF_STREAM_PARSER do not add CONFIG_BPF_SOCK_MAP
Cong Wang (5):
bpf: clean up sockmap related Kconfigs
skmsg: get rid of struct sk_psock_parser
bpf: compute data_end dynamically with JIT code
skmsg: use skb ext instead of TCP_SKB_CB
sock_map: rename skb_parser and skb_verdict
include/linux/bpf.h | 20 +- include/linux/bpf_types.h | 2 - include/linux/skbuff.h | 3 + include/linux/skmsg.h | 85 ++++++-- include/net/tcp.h | 38 +--- include/net/udp.h | 4 +- init/Kconfig | 1 + net/Kconfig | 7 +- net/core/Makefile | 2 +- net/core/filter.c | 48 +++-- net/core/skbuff.c | 7 + net/core/skmsg.c | 194 +++++++++--------- net/core/sock_map.c | 74 +++---- net/ipv4/Makefile | 2 +- net/ipv4/tcp_bpf.c | 2 - .../selftests/bpf/prog_tests/sockmap_listen.c | 8 +- .../selftests/bpf/progs/test_sockmap_listen.c | 4 +- 17 files changed, 269 insertions(+), 232 deletions(-)
Signature · 1 line
2.25.1
PATCH v3 1/5 bpf: clean up sockmap related Kconfigs
From: Cong Wang <[email protected]>
As suggested by John, clean up sockmap related Kconfigs:
Reduce the scope of CONFIG_BPF_STREAM_PARSER down to TCP stream parser, to reflect its name.
Make the rest sockmap code simply depend on CONFIG_BPF_SYSCALL. And leave CONFIG_NET_SOCK_MSG untouched, as it is used by non-sockmap cases.
--- include/linux/bpf.h | 20 +++--- include/linux/bpf_types.h | 2 - include/linux/skmsg.h | 22 ++++++ include/net/tcp.h | 13 ++-- include/net/udp.h | 4 +- init/Kconfig | 1 + net/Kconfig | 6 +- net/core/Makefile | 2 +- net/core/skmsg.c | 139 ++++++++++++++++++++------------------ net/core/sock_map.c | 2 + net/ipv4/Makefile | 2 +- net/ipv4/tcp_bpf.c | 2 - 12 files changed, 117 insertions(+), 98 deletions(-)
include/linux/bpf.h +9 -11
@@ -1741,6 +1741,14 @@ static inline bool bpf_map_is_dev_bound(struct bpf_map *map)struct bpf_map *bpf_map_offload_map_alloc (union bpf_attr *attr);void bpf_map_offload_map_free (struct bpf_map *map); + +int sock_map_prog_update (struct bpf_map *map,struct bpf_prog *prog, +struct bpf_prog *old,u32 which); +int sock_map_get_from_fd (const union bpf_attr *attr,struct bpf_prog *prog); +int sock_map_prog_detach (const union bpf_attr *attr,enum bpf_prog_type ptype); +int sock_map_update_elem_sys (struct bpf_map *map,void *key,void *value,u64 flags); +void sock_map_unhash (struct sock *sk); +void sock_map_close (struct sock *sk,long timeout);#else static inline int bpf_prog_offload_init (struct bpf_prog *prog,union bpf_attr *attr)
@@ -1766,17 +1774,7 @@ static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr)static inline void bpf_map_offload_map_free (struct bpf_map *map) { } -#endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */ -#if defined(CONFIG_BPF_STREAM_PARSER ) -int sock_map_prog_update (struct bpf_map *map,struct bpf_prog *prog, -struct bpf_prog *old,u32 which); -int sock_map_get_from_fd (const union bpf_attr *attr,struct bpf_prog *prog); -int sock_map_prog_detach (const union bpf_attr *attr,enum bpf_prog_type ptype); -int sock_map_update_elem_sys (struct bpf_map *map,void *key,void *value,u64 flags); -void sock_map_unhash (struct sock *sk); -void sock_map_close (struct sock *sk,long timeout); -#else static inline int sock_map_prog_update (struct bpf_map *map,struct bpf_prog *prog,struct bpf_prog *old,u32 which)
@@ -1801,7 +1799,7 @@ static inline int sock_map_update_elem_sys(struct bpf_map *map, void *key, void {return -EOPNOTSUPP ; } -#endif /* CONFIG_BPF_STREAM_PARSER */ +#endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */ #if defined(CONFIG_INET ) && defined(CONFIG_BPF_SYSCALL )void bpf_sk_reuseport_detach (struct sock *sk);
include/linux/bpf_types.h +0 -2
@@ -103,10 +103,8 @@ BPF_MAP_TYPE(BPF_MAP_TYPE_HASH_OF_MAPS, htab_of_maps_map_ops) BPF_MAP_TYPE(BPF_MAP_TYPE_DEVMAP, dev_map_ops) BPF_MAP_TYPE(BPF_MAP_TYPE_DEVMAP_HASH, dev_map_hash_ops) BPF_MAP_TYPE(BPF_MAP_TYPE_SK_STORAGE, sk_storage_map_ops) -#if defined(CONFIG_BPF_STREAM_PARSER ) BPF_MAP_TYPE(BPF_MAP_TYPE_SOCKMAP, sock_map_ops) BPF_MAP_TYPE(BPF_MAP_TYPE_SOCKHASH, sock_hash_ops) -#endif #ifdef CONFIG_BPF_LSM BPF_MAP_TYPE(BPF_MAP_TYPE_INODE_STORAGE, inode_storage_map_ops)BPF_MAP_TYPE (BPF_MAP_TYPE_TASK_STORAGE , task_storage_map_ops)
include/linux/skmsg.h +22 -0
@@ -71,7 +71,9 @@ struct sk_psock_link { };struct sk_psock_parser { +#if IS_ENABLED (CONFIG_BPF_STREAM_PARSER )struct strparser strp; +#endif bool enabled;void (*saved_data_ready)(struct sock *sk); };
@@ -305,9 +307,29 @@ static inline void sk_psock_report_error(struct sk_psock *psock, int err)struct sk_psock *sk_psock_init (struct sock *sk,int node); +#if IS_ENABLED (CONFIG_BPF_STREAM_PARSER )int sk_psock_init_strp (struct sock *sk,struct sk_psock *psock);void sk_psock_start_strp (struct sock *sk,struct sk_psock *psock);void sk_psock_stop_strp (struct sock *sk,struct sk_psock *psock); +void sk_psock_done_strp (struct sk_psock *psock); +#else +static inline int sk_psock_init_strp (struct sock *sk,struct sk_psock *psock) +{ +return -EOPNOTSUPP ; +} + +static inline void sk_psock_start_strp (struct sock *sk,struct sk_psock *psock) +{ +} + +static inline void sk_psock_stop_strp (struct sock *sk,struct sk_psock *psock) +{ +} + +static inline void sk_psock_done_strp (struct sk_psock *psock) +{ +} +#endif void sk_psock_start_verdict (struct sock *sk,struct sk_psock *psock);void sk_psock_stop_verdict (struct sock *sk,struct sk_psock *psock);
include/net/tcp.h +5 -8
@@ -2204,23 +2204,20 @@ void tcp_update_ulp(struct sock *sk, struct proto *p,__MODULE_INFO (alias, alias_userspace, name); \ __MODULE_INFO(alias, alias_tcp_ulp, "tcp-ulp-" name) +#ifdef CONFIG_NET_SOCK_MSG struct sk_msg ;struct sk_psock ; - -#ifdef CONFIG_BPF_STREAM_PARSER struct proto *tcp_bpf_get_proto (struct sock *sk,struct sk_psock *psock);void tcp_bpf_clone (const struct sock *sk,struct sock *newsk); -#else -static inline void tcp_bpf_clone (const struct sock *sk,struct sock *newsk) -{ -} -#endif /* CONFIG_BPF_STREAM_PARSER */ -#ifdef CONFIG_NET_SOCK_MSG int tcp_bpf_sendmsg_redir (struct sock *sk,struct sk_msg *msg,u32 bytes,int flags);int __tcp_bpf_recvmsg (struct sock *sk,struct sk_psock *psock,struct msghdr *msg,int len,int flags); +#else +static inline void tcp_bpf_clone (const struct sock *sk,struct sock *newsk) +{ +}#endif /* CONFIG_NET_SOCK_MSG */ #ifdef CONFIG_CGROUP_BPF
include/net/udp.h +2 -2
@@ -511,9 +511,9 @@ static inline struct sk_buff *udp_rcv_segment(struct sock *sk, return segs; } -#ifdef CONFIG_BPF_STREAM_PARSER +#ifdef CONFIG_NET_SOCK_MSG struct sk_psock ;struct proto *udp_bpf_get_proto (struct sock *sk,struct sk_psock *psock); -#endif /* BPF_STREAM_PARSER */ +#endif /* CONFIG_NET_SOCK_MSG */ #endif /* _UDP_H */
init/Kconfig +1 -0
@@ -1703,6 +1703,7 @@ config BPF_SYSCALL select BPF select IRQ_WORK select TASKS_TRACE_RCU + select NET_SOCK_MSG if NET default n help Enable the bpf() system call that allows to manipulate eBPF
net/Kconfig +1 -5
@@ -313,13 +313,9 @@ config BPF_STREAM_PARSER select STREAM_PARSER select NET_SOCK_MSG help - Enabling this allows a stream parser to be used with + Enabling this allows a TCP stream parser to be used with BPF_MAP_TYPE_SOCKMAP. - BPF_MAP_TYPE_SOCKMAP provides a map type to use with network sockets. - It can be used to enforce socket policy, implement socket redirects, - etc. - config NET_FLOW_LIMIT bool depends on RPS
net/core/Makefile +1 -1
@@ -28,10 +28,10 @@ obj-$(CONFIG_CGROUP_NET_PRIO) += netprio_cgroup.o obj-$(CONFIG_CGROUP_NET_CLASSID) += netclassid_cgroup.o obj-$(CONFIG_LWTUNNEL) += lwtunnel.o obj-$(CONFIG_LWTUNNEL_BPF) += lwt_bpf.o -obj-$(CONFIG_BPF_STREAM_PARSER) += sock_map.o obj-$(CONFIG_DST_CACHE) += dst_cache.o obj-$(CONFIG_HWBM) += hwbm.o obj-$(CONFIG_NET_DEVLINK) += devlink.o obj-$(CONFIG_GRO_CELLS) += gro_cells.o obj-$(CONFIG_FAILOVER) += failover.o +obj-$(CONFIG_BPF_SYSCALL) += sock_map.o obj-$(CONFIG_BPF_SYSCALL) += bpf_sk_storage.o
net/core/skmsg.c +73 -66
181 lines not on this page · raw message
net/core/sock_map.c +2 -0
@@ -1461,9 +1461,11 @@ int sock_map_prog_update(struct bpf_map *map, struct bpf_prog *prog,case BPF_SK_MSG_VERDICT : pprog = &progs->msg_parser;break ; +#if IS_ENABLED (CONFIG_BPF_STREAM_PARSER )case BPF_SK_SKB_STREAM_PARSER : pprog = &progs->skb_parser;break ; +#endif case BPF_SK_SKB_STREAM_VERDICT : pprog = &progs->skb_verdict;break ;
net/ipv4/Makefile +1 -1
@@ -62,7 +62,7 @@ obj-$(CONFIG_TCP_CONG_LP) += tcp_lp.o obj-$(CONFIG_TCP_CONG_YEAH) += tcp_yeah.o obj-$(CONFIG_TCP_CONG_ILLINOIS) += tcp_illinois.o obj-$(CONFIG_NET_SOCK_MSG) += tcp_bpf.o -obj-$(CONFIG_BPF_STREAM_PARSER) += udp_bpf.o +obj-$(CONFIG_NET_SOCK_MSG) += udp_bpf.o obj-$(CONFIG_NETLABEL) += cipso_ipv4.o obj-$(CONFIG_XFRM) += xfrm4_policy.o xfrm4_state.o xfrm4_input.o \
net/ipv4/tcp_bpf.c +0 -2
@@ -229,7 +229,6 @@ int tcp_bpf_sendmsg_redir(struct sock *sk, struct sk_msg *msg, }EXPORT_SYMBOL_GPL (tcp_bpf_sendmsg_redir); -#ifdef CONFIG_BPF_STREAM_PARSER static bool tcp_bpf_stream_read (const struct sock *sk) {struct sk_psock *psock;
@@ -629,4 +628,3 @@ void tcp_bpf_clone(const struct sock *sk, struct sock *newsk)if (prot == &tcp_bpf_prots[family][TCP_BPF_BASE ]) newsk->sk_prot = sk->sk_prot_creator; } -#endif /* CONFIG_BPF_STREAM_PARSER */
Signature · 1 line
2.25.1
Re: [Patch bpf-next v3 1/5] bpf: clean up sockmap related Kconfigs
Quoted 11 lines · Cong Wang
From: Cong Wang <[email protected]>
As suggested by John, clean up sockmap related Kconfigs:
Reduce the scope of
CONFIG_BPF_STREAM_PARSERdown to TCP stream parser, to reflect its name.Make the rest sockmap code simply depend on
CONFIG_BPF_SYSCALL. And leaveCONFIG_NET_SOCK_MSGuntouched, as it is used by non-sockmap cases.
For the series:
Jakub, John: can you please take another look at the assembly in patch 3?
Signature · 4 lines
Lorenz Bauer | Systems Engineer 6th Floor, County Hall/The Riverside Building, SE1 7PB, UK www.cloudflare.com
RE: [Patch bpf-next v3 1/5] bpf: clean up sockmap related Kconfigs
In reply to Cong WangQuoted 11 lines · Cong Wang
From: Cong Wang <[email protected]>
As suggested by John, clean up sockmap related Kconfigs:
Reduce the scope of
CONFIG_BPF_STREAM_PARSERdown to TCP stream parser, to reflect its name.Make the rest sockmap code simply depend on
CONFIG_BPF_SYSCALL. And leaveCONFIG_NET_SOCK_MSGuntouched, as it is used by non-sockmap cases.
Quoted 6 lines
Cc John Fastabend <[email protected]>Cc Daniel Borkmann <[email protected]>Cc Jakub Sitnicki <[email protected]>Cc Lorenz Bauer <[email protected]>Signed-off-by Cong Wang <[email protected]>---
Thanks for doing this.
PATCH v3 2/5 skmsg: get rid of struct sk_psock_parser
In reply to Cong WangFrom: Cong Wang <[email protected]>
struct sk_psock_parser is embedded in sk_psock, it is unnecessary as skb verdict also uses ->saved_data_ready. We can simply fold these fields into sk_psock, and get rid of ->enabled.
--- include/linux/skmsg.h | 19 ++++++---------- net/core/skmsg.c | 53 +++++++++++++------------------------------ net/core/sock_map.c | 8 +++---- 3 files changed, 27 insertions(+), 53 deletions(-)
include/linux/skmsg.h +7 -12
@@ -70,14 +70,6 @@ struct sk_psock_link {void *link_raw; }; -struct sk_psock_parser { -#if IS_ENABLED (CONFIG_BPF_STREAM_PARSER ) -struct strparser strp; -#endif -bool enabled; -void (*saved_data_ready)(struct sock *sk); -}; -struct sk_psock_work_state {struct sk_buff *skb;u32 len;
@@ -92,7 +84,9 @@ struct sk_psock {u32 eval;struct sk_msg *cork;struct sk_psock_progs progs; -struct sk_psock_parser parser; +#if IS_ENABLED (CONFIG_BPF_STREAM_PARSER ) +struct strparser strp; +#endif struct sk_buff_head ingress_skb;struct list_head ingress_msg;unsigned long state;
@@ -102,6 +96,7 @@ struct sk_psock {void (*saved_unhash)(struct sock *sk);void (*saved_close)(struct sock *sk,long timeout);void (*saved_write_space)(struct sock *sk); +void (*saved_data_ready)(struct sock *sk);struct proto *sk_proto;struct sk_psock_work_state work_state;struct work_struct work;
@@ -422,8 +417,8 @@ static inline void sk_psock_put(struct sock *sk, struct sk_psock *psock)static inline void sk_psock_data_ready (struct sock *sk,struct sk_psock *psock) { -if (psock->parser.enabled) - psock->parser.saved_data_ready (sk); +if (psock->saved_data_ready) + psock->saved_data_ready (sk);else sk->sk_data_ready (sk); }
@@ -462,6 +457,6 @@ static inline bool sk_psock_strp_enabled(struct sk_psock *psock) {if (!psock)return false; -return psock->parser.enabled; +return !!psock->saved_data_ready; }#endif /* _LINUX_SKMSG_H */
net/core/skmsg.c +16 -37
109 lines not on this page · raw message
net/core/sock_map.c +4 -4
@@ -148,9 +148,9 @@ static void sock_map_del_link(struct sock *sk,struct bpf_map *map = link->map; struct bpf_stab *stab = container_of(map, struct bpf_stab, map); -if (psock->parser.enabled && stab->progs.skb_parser) +if (psock->saved_data_ready && stab->progs.skb_parser) strp_stop = true; -if (psock->parser.enabled && stab->progs.skb_verdict) +if (psock->saved_data_ready && stab->progs.skb_verdict) verdict_stop = true;list_del (&link->list);sk_psock_free_link (link);
@@ -283,14 +283,14 @@ static int sock_map_link(struct bpf_map *map, struct sk_psock_progs *progs, goto out_drop;write_lock_bh (&sk->sk_callback_lock); -if (skb_parser && skb_verdict && !psock->parser.enabled) { +if (skb_parser && skb_verdict && !psock->saved_data_ready) { ret =sk_psock_init_strp (sk, psock);if (ret) goto out_unlock_drop;psock_set_prog (&psock->progs.skb_verdict, skb_verdict);psock_set_prog (&psock->progs.skb_parser, skb_parser);sk_psock_start_strp (sk, psock); - }else if (!skb_parser && skb_verdict && !psock->parser.enabled) { + }else if (!skb_parser && skb_verdict && !psock->saved_data_ready) {psock_set_prog (&psock->progs.skb_verdict, skb_verdict);sk_psock_start_verdict (sk,psock); }
Signature · 1 line
2.25.1
RE: [Patch bpf-next v3 2/5] skmsg: get rid of struct sk_psock_parser
Quoted 12 lines · Cong Wang
From: Cong Wang <[email protected]>
struct
sk_psock_parseris embedded insk_psock, it is unnecessary as skb verdict also uses ->saved_data_ready. We can simply fold these fields intosk_psock, and get rid of ->enabled.Cc John Fastabend <[email protected]>Cc Daniel Borkmann <[email protected]>Cc Jakub Sitnicki <[email protected]>Cc Lorenz Bauer <[email protected]>Signed-off-by Cong Wang <[email protected]>
Quoted 6 lines
--- include/linux/skmsg.h | 19 ++++++---------- net/core/skmsg.c | 53 +++++++++++++------------------------------ net/core/sock_map.c | 8 +++---- 3 files changed, 27 insertions(+), 53 deletions(-)
Re: [Patch bpf-next v3 2/5] skmsg: get rid of struct sk_psock_parser
In reply to Cong WangQuoted 7 lines · Cong Wang
From: Cong Wang <[email protected]>
struct
sk_psock_parseris embedded insk_psock, it is unnecessary as skb verdict also uses ->saved_data_ready. We can simply fold these fields intosk_psock, and get rid of ->enabled.
Quoted 6 lines
Cc John Fastabend <[email protected]>Cc Daniel Borkmann <[email protected]>Cc Jakub Sitnicki <[email protected]>Cc Lorenz Bauer <[email protected]>Signed-off-by Cong Wang <[email protected]>---
PATCH v3 3/5 bpf: compute data_end dynamically with JIT code
In reply to Cong WangFrom: Cong Wang <[email protected]>
Currently, we compute ->data_end with a compile-time constant offset of skb. But as Jakub pointed out, we can actually compute it in eBPF JIT code at run-time, so that we can competely get rid of ->data_end. This is similar to skb_shinfo(skb) computation in bpf_convert_shinfo_access().
--- include/net/tcp.h | 6 ------ net/core/filter.c | 48 +++++++++++++++++++++++++++-------------------- net/core/skmsg.c | 1 - 3 files changed, 28 insertions(+), 27 deletions(-)
include/net/tcp.h +0 -6
@@ -885,18 +885,12 @@ struct tcp_skb_cb {struct {__u32 flags;struct sock *sk_redir; -void *data_end; } bpf; }; };#define TCP_SKB_CB (__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0])) -static inline void bpf_compute_data_end_sk_skb (struct sk_buff *skb) -{ -TCP_SKB_CB (skb)->bpf.data_end = skb->data +skb_headlen (skb); -} -static inline bool tcp_skb_bpf_ingress (const struct sk_buff *skb) {return TCP_SKB_CB (skb)->bpf.flags &BPF_F_INGRESS ;
net/core/filter.c +28 -20
87 lines not on this page · raw message
net/core/skmsg.c +0 -1
@@ -744,7 +744,6 @@ EXPORT_SYMBOL_GPL(sk_psock_msg_verdict);static int sk_psock_bpf_run (struct sk_psock *psock,struct bpf_prog *prog,struct sk_buff *skb) { -bpf_compute_data_end_sk_skb (skb);return bpf_prog_run_pin_on_cpu (prog, skb); }
Signature · 1 line
2.25.1
RE: [Patch bpf-next v3 3/5] bpf: compute data_end dynamically with JIT code
Quoted 8 lines · Cong Wang
From: Cong Wang <[email protected]>
Currently, we compute ->data_end with a compile-time constant offset of skb. But as Jakub pointed out, we can actually compute it in eBPF JIT code at run-time, so that we can competely get rid of ->data_end. This is similar to
skb_shinfo(skb)computation inbpf_convert_shinfo_access().
Quoted 6 lines
Suggested-by Jakub Sitnicki <[email protected]>Cc John Fastabend <[email protected]>Cc Daniel Borkmann <[email protected]>Cc Lorenz Bauer <[email protected]>Signed-off-by Cong Wang <[email protected]>---
Re: [Patch bpf-next v3 3/5] bpf: compute data_end dynamically with JIT code
In reply to Cong WangQuoted 8 lines · Cong Wang
From: Cong Wang <[email protected]>
Currently, we compute ->data_end with a compile-time constant offset of skb. But as Jakub pointed out, we can actually compute it in eBPF JIT code at run-time, so that we can competely get rid of ->data_end. This is similar to
skb_shinfo(skb)computation inbpf_convert_shinfo_access().
Quoted 6 lines
Suggested-by Jakub Sitnicki <[email protected]>Cc John Fastabend <[email protected]>Cc Daniel Borkmann <[email protected]>Cc Lorenz Bauer <[email protected]>Signed-off-by Cong Wang <[email protected]>---
PATCH v3 4/5 skmsg: use skb ext instead of TCP_SKB_CB
In reply to Cong WangFrom: Cong Wang <[email protected]>
Currently TCP_SKB_CB() is hard-coded in skmsg code, it certainly does not work for any other non-TCP protocols. We can move them to skb ext instead of playing with skb cb, which is harder to make correct.
--- include/linux/skbuff.h | 3 +++ include/linux/skmsg.h | 40 ++++++++++++++++++++++++++++++++++++++++ include/net/tcp.h | 19 ------------------- net/Kconfig | 1 + net/core/skbuff.c | 7 +++++++ net/core/skmsg.c | 35 +++++++++++++++++++++++------------ net/core/sock_map.c | 12 ++++++------ 7 files changed, 80 insertions(+), 37 deletions(-)
include/linux/skbuff.h +3 -0
@@ -4166,6 +4166,9 @@ enum skb_ext_id {#endif #if IS_ENABLED (CONFIG_MPTCP ) SKB_EXT_MPTCP, +#endif +#if IS_ENABLED (CONFIG_NET_SOCK_MSG ) +SKB_EXT_BPF , #endif SKB_EXT_NUM, /* must be last */ };
include/linux/skmsg.h +40 -0
@@ -459,4 +459,44 @@ static inline bool sk_psock_strp_enabled(struct sk_psock *psock)return false;return !!psock->saved_data_ready; } + +struct skb_bpf_ext { +__u32 flags; +struct sock *sk_redir; +}; + +#if IS_ENABLED (CONFIG_NET_SOCK_MSG ) +static inline +bool skb_bpf_ext_ingress (const struct sk_buff *skb) +{ +struct skb_bpf_ext *ext =skb_ext_find (skb,SKB_EXT_BPF ); + +return ext->flags &BPF_F_INGRESS ; +} + +static inline +void skb_bpf_ext_set_ingress (const struct sk_buff *skb) +{ +struct skb_bpf_ext *ext =skb_ext_find (skb,SKB_EXT_BPF ); + + ext->flags |=BPF_F_INGRESS ; +} + +static inline +struct sock *skb_bpf_ext_redirect_fetch (struct sk_buff *skb) +{ +struct skb_bpf_ext *ext =skb_ext_find (skb,SKB_EXT_BPF ); + +return ext->sk_redir; +} + +static inline +void skb_bpf_ext_redirect_clear (struct sk_buff *skb) +{ +struct skb_bpf_ext *ext =skb_ext_find (skb,SKB_EXT_BPF ); + + ext->flags =0 ; + ext->sk_redir =NULL ; +} +#endif /* CONFIG_NET_SOCK_MSG */ #endif /* _LINUX_SKMSG_H */
include/net/tcp.h +0 -19
@@ -882,30 +882,11 @@ struct tcp_skb_cb {struct inet6_skb_parm h6;#endif } header;/* For incoming skbs */ -struct { -__u32 flags; -struct sock *sk_redir; - } bpf; }; };#define TCP_SKB_CB (__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0])) -static inline bool tcp_skb_bpf_ingress (const struct sk_buff *skb) -{ -return TCP_SKB_CB (skb)->bpf.flags &BPF_F_INGRESS ; -} - -static inline struct sock *tcp_skb_bpf_redirect_fetch (struct sk_buff *skb) -{ -return TCP_SKB_CB (skb)->bpf.sk_redir; -} - -static inline void tcp_skb_bpf_redirect_clear (struct sk_buff *skb) -{ -TCP_SKB_CB (skb)->bpf.sk_redir =NULL ; -} -extern const struct inet_connection_sock_af_ops ipv4_specific;#if IS_ENABLED (CONFIG_IPV6 )
net/Kconfig +1 -0
@@ -419,6 +419,7 @@ config SOCK_VALIDATE_XMIT config NET_SOCK_MSG bool + select SKB_EXTENSIONS default n help The NET_SOCK_MSG provides a framework for plain sockets (e.g. TCP) or
net/core/skbuff.c +7 -0
@@ -60,6 +60,7 @@#include <linux/prefetch.h> #include <linux/if_vlan.h> #include <linux/mpls.h> +#include <linux/skmsg.h> #include <net/protocol.h> #include <net/dst.h>
@@ -4259,6 +4260,9 @@ static const u8 skb_ext_type_len[] = {#if IS_ENABLED (CONFIG_MPTCP ) [SKB_EXT_MPTCP] = SKB_EXT_CHUNKSIZEOF(struct mptcp_ext), #endif +#if IS_ENABLED(CONFIG_NET_SOCK_MSG) + [SKB_EXT_BPF] = SKB_EXT_CHUNKSIZEOF(struct skb_bpf_ext), +#endif }; static __always_inline unsigned int skb_ext_total_length(void)
@@ -4275,6 +4279,9 @@ static __always_inline unsigned int skb_ext_total_length(void)#endif #if IS_ENABLED (CONFIG_MPTCP ) skb_ext_type_len[SKB_EXT_MPTCP] + +#endif +#if IS_ENABLED (CONFIG_NET_SOCK_MSG ) + skb_ext_type_len[SKB_EXT_BPF] +#endif 0 ; }
net/core/skmsg.c +23 -12
@@ -525,7 +525,8 @@ static void sk_psock_backlog(struct work_struct *work) len = skb->len; off =0 ; start: - ingress =tcp_skb_bpf_ingress (skb); + ingress =skb_bpf_ext_ingress (skb); +skb_ext_del (skb,SKB_EXT_BPF ); do { ret = -EIO; if (likely(psock->sk->sk_socket))
@@ -752,7 +753,7 @@ static void sk_psock_skb_redirect(struct sk_buff *skb)struct sk_psock *psock_other;struct sock *sk_other; - sk_other =tcp_skb_bpf_redirect_fetch (skb); + sk_other =skb_bpf_ext_redirect_fetch (skb);/* This error is a buggy BPF program, it returned a redirect * return code, but then didn't set a redirect interface. */
@@ -794,6 +795,9 @@ int sk_psock_tls_strp_read(struct sk_psock *psock, struct sk_buff *skb)struct bpf_prog *prog;int ret = __SK_PASS; +if (!skb_ext_add (skb,SKB_EXT_BPF )) +return __SK_DROP; +rcu_read_lock (); prog =READ_ONCE (psock->progs.skb_verdict);if (likely (prog)) {
@@ -802,9 +806,9 @@ int sk_psock_tls_strp_read(struct sk_psock *psock, struct sk_buff *skb)* TLS context. */ skb->sk = psock->sk; -tcp_skb_bpf_redirect_clear (skb); +skb_bpf_ext_redirect_clear (skb); ret =sk_psock_bpf_run (psock, prog, skb); - ret =sk_psock_map_verd (ret,tcp_skb_bpf_redirect_fetch (skb)); + ret =sk_psock_map_verd (ret,skb_bpf_ext_redirect_fetch (skb)); skb->sk =NULL ; }sk_psock_tls_verdict_apply (skb, psock->sk, ret);
@@ -816,7 +820,6 @@ EXPORT_SYMBOL_GPL(sk_psock_tls_strp_read);static void sk_psock_verdict_apply (struct sk_psock *psock,struct sk_buff *skb,int verdict) { -struct tcp_skb_cb *tcp;struct sock *sk_other;int err = -EIO ;
@@ -828,9 +831,7 @@ static void sk_psock_verdict_apply(struct sk_psock *psock, goto out_free; } - tcp =TCP_SKB_CB (skb); - tcp->bpf.flags |=BPF_F_INGRESS ; - +skb_bpf_ext_set_ingress (skb);/* If the queue is empty then we can submit directly * into the msg queue. If its not empty we have to * queue work otherwise we may get OOO data. Otherwise,
@@ -888,11 +889,15 @@ static void sk_psock_strp_read(struct strparser *strp, struct sk_buff *skb) goto out; }skb_set_owner_r (skb, sk); +if (!skb_ext_add (skb,SKB_EXT_BPF )) { +kfree_skb (skb); + goto out; + } prog =READ_ONCE (psock->progs.skb_verdict);if (likely (prog)) { -tcp_skb_bpf_redirect_clear (skb); +skb_bpf_ext_redirect_clear (skb); ret =sk_psock_bpf_run (psock, prog, skb); - ret =sk_psock_map_verd (ret,tcp_skb_bpf_redirect_fetch (skb)); + ret =sk_psock_map_verd (ret,skb_bpf_ext_redirect_fetch (skb)); }sk_psock_verdict_apply (psock, skb, ret); out:
@@ -1003,11 +1008,17 @@ static int sk_psock_verdict_recv(read_descriptor_t *desc, struct sk_buff *skb, goto out; }skb_set_owner_r (skb, sk); +if (!skb_ext_add (skb,SKB_EXT_BPF )) { + len =0 ; +kfree_skb (skb); + goto out; + } + prog =READ_ONCE (psock->progs.skb_verdict);if (likely (prog)) { -tcp_skb_bpf_redirect_clear (skb); +skb_bpf_ext_redirect_clear (skb); ret =sk_psock_bpf_run (psock, prog, skb); - ret =sk_psock_map_verd (ret,tcp_skb_bpf_redirect_fetch (skb)); + ret =sk_psock_map_verd (ret,skb_bpf_ext_redirect_fetch (skb)); }sk_psock_verdict_apply (psock, skb, ret); out:
net/core/sock_map.c +6 -6
@@ -657,7 +657,7 @@ const struct bpf_func_proto bpf_sock_map_update_proto = { BPF_CALL_4(bpf_sk_redirect_map, struct sk_buff *, skb, struct bpf_map *, map, u32, key, u64, flags) { -struct tcp_skb_cb *tcb =TCP_SKB_CB (skb); +struct skb_bpf_ext *ext =skb_ext_find (skb,SKB_EXT_BPF );struct sock *sk;if (unlikely (flags & ~(BPF_F_INGRESS )))
@@ -667,8 +667,8 @@ BPF_CALL_4(bpf_sk_redirect_map, struct sk_buff *, skb,if (unlikely (!sk || !sock_map_redirect_allowed (sk)))return SK_DROP ; - tcb->bpf.flags = flags; - tcb->bpf.sk_redir = sk; + ext->flags = flags; + ext->sk_redir = sk;return SK_PASS ; }
@@ -1250,7 +1250,7 @@ const struct bpf_func_proto bpf_sock_hash_update_proto = { BPF_CALL_4(bpf_sk_redirect_hash, struct sk_buff *, skb, struct bpf_map *, map, void *, key, u64, flags) { -struct tcp_skb_cb *tcb =TCP_SKB_CB (skb); +struct skb_bpf_ext *ext =skb_ext_find (skb,SKB_EXT_BPF );struct sock *sk;if (unlikely (flags & ~(BPF_F_INGRESS )))
@@ -1260,8 +1260,8 @@ BPF_CALL_4(bpf_sk_redirect_hash, struct sk_buff *, skb,if (unlikely (!sk || !sock_map_redirect_allowed (sk)))return SK_DROP ; - tcb->bpf.flags = flags; - tcb->bpf.sk_redir = sk; + ext->flags = flags; + ext->sk_redir = sk;return SK_PASS ; }
Signature · 1 line
2.25.1
John Fastabend
maintainerRE: [Patch bpf-next v3 4/5] skmsg: use skb ext instead of TCP_SKB_CB
Quoted 7 lines · Cong Wang
From: Cong Wang <[email protected]>
Currently
TCP_SKB_CB()is hard-coded in skmsg code, it certainly does not work for any other non-TCP protocols. We can move them to skb ext instead of playing with skb cb, which is harder to make correct.
Quoted 6 lines
Cc John Fastabend <[email protected]>Cc Daniel Borkmann <[email protected]>Cc Jakub Sitnicki <[email protected]>Reviewed-by Lorenz Bauer <[email protected]>Signed-off-by Cong Wang <[email protected]>---
I'm not seeing the advantage of doing this at the moment. We can continue to use cb[] here, which is simpler IMO and use the ext if needed for the other use cases. This is adding a per packet alloc cost that we don't have at the moment as I understand it.
[...]
Quoted 38 lines
include/linux/skmsg.h +30
-0@@ -459,4 +459,44 @@ static inline bool sk_psock_strp_enabled(struct sk_psock *psock)return false;return !!psock->saved_data_ready; } + +struct skb_bpf_ext { +__u32 flags; +struct sock *sk_redir; +}; + +#if IS_ENABLED (CONFIG_NET_SOCK_MSG ) +static inline +bool skb_bpf_ext_ingress (const struct sk_buff *skb) +{ +struct skb_bpf_ext *ext =skb_ext_find (skb,SKB_EXT_BPF ); + +return ext->flags &BPF_F_INGRESS ; +} + +static inline +void skb_bpf_ext_set_ingress (const struct sk_buff *skb) +{ +struct skb_bpf_ext *ext =skb_ext_find (skb,SKB_EXT_BPF ); + + ext->flags |=BPF_F_INGRESS ; +} + +static inline +struct sock *skb_bpf_ext_redirect_fetch (struct sk_buff *skb) +{ +struct skb_bpf_ext *ext =skb_ext_find (skb,SKB_EXT_BPF ); + +return ext->sk_redir; +}
include/linux/skmsg.h static inline bool sk_psock_strp_enabled(struct sk_psock *psock) · line 501
Quoted 10 lines
include/linux/skmsg.h +10
-0+ +static inline +void skb_bpf_ext_redirect_clear (struct sk_buff *skb) +{ +struct skb_bpf_ext *ext =skb_ext_find (skb,SKB_EXT_BPF ); + + ext->flags =0 ; + ext->sk_redir =NULL ; +} +#endif /* CONFIG_NET_SOCK_MSG */
So we will have some slight duplication for cb[] variant and ext variant above. I'm OK with that to avoid an allocation.
[...]
include/linux/skmsg.h static int sk_psock_verdict_recv(read_descriptor_t *desc, struct sk_buff *skb, · line 1016
Quoted 10 lines
include/linux/skmsg.h +6
-0@@ -1003,11 +1008,17 @@ static int sk_psock_verdict_recv(read_descriptor_t *desc, struct sk_buff *skb, goto out; }skb_set_owner_r (skb, sk); +if (!skb_ext_add (skb,SKB_EXT_BPF )) { + len =0 ; +kfree_skb (skb); + goto out; + } +
per packet cost here. Perhaps you can argue small alloc will usually not be noticable in such a large stack, but once we convert over it will be very hard to go back. And I'm looking at optimizing this path now.
include/linux/skmsg.h static int sk_psock_verdict_recv(read_descriptor_t *desc, struct sk_buff *skb, · line 1023
Quoted 9 lines
include/linux/skmsg.h +2
-2prog =READ_ONCE (psock->progs.skb_verdict);if (likely (prog)) { -tcp_skb_bpf_redirect_clear (skb); +skb_bpf_ext_redirect_clear (skb); ret =sk_psock_bpf_run (psock, prog, skb); - ret =sk_psock_map_verd (ret,tcp_skb_bpf_redirect_fetch (skb)); + ret =sk_psock_map_verd (ret,skb_bpf_ext_redirect_fetch (skb)); }sk_psock_verdict_apply (psock, skb, ret);
Thanks for the series Cong. Drop this patch and resubmit carry ACKs forward and then lets revisit this later.
Thanks,
John
Re: [Patch bpf-next v3 4/5] skmsg: use skb ext instead of TCP_SKB_CB
Quoted 16 lines · John Fastabend
Quoted 7 lines · Cong Wang
From: Cong Wang <[email protected]>
Currently
TCP_SKB_CB()is hard-coded in skmsg code, it certainly does not work for any other non-TCP protocols. We can move them to skb ext instead of playing with skb cb, which is harder to make correct.Quoted 6 lines
Cc John Fastabend <[email protected]>Cc Daniel Borkmann <[email protected]>Cc Jakub Sitnicki <[email protected]>Reviewed-by Lorenz Bauer <[email protected]>Signed-off-by Cong Wang <[email protected]>---
Quoted 4 lines
I'm not seeing the advantage of doing this at the moment. We can continue to use cb[] here, which is simpler IMO and use the ext if needed for the other use cases. This is adding a per packet alloc cost that we don't have at the moment as I understand it.
Hmm? How can we continue using TCP_SKB_CB() for UDP or AF_UNIX?
I am not sure I get your "at the moment" correctly, do you mean I should move this patch to a later patchset, maybe the UDP patchset? At least this patch is needed, no matter by which patchset, so it should not be dropped.
Quoted 46 lines
[...]
Quoted 43 lines
include/linux/skmsg.h +35
-0@@ -459,4 +459,44 @@ static inline bool sk_psock_strp_enabled(struct sk_psock *psock)return false;return !!psock->saved_data_ready; } + +struct skb_bpf_ext { +__u32 flags; +struct sock *sk_redir; +}; + +#if IS_ENABLED (CONFIG_NET_SOCK_MSG ) +static inline +bool skb_bpf_ext_ingress (const struct sk_buff *skb) +{ +struct skb_bpf_ext *ext =skb_ext_find (skb,SKB_EXT_BPF ); + +return ext->flags &BPF_F_INGRESS ; +} + +static inline +void skb_bpf_ext_set_ingress (const struct sk_buff *skb) +{ +struct skb_bpf_ext *ext =skb_ext_find (skb,SKB_EXT_BPF ); + + ext->flags |=BPF_F_INGRESS ; +} + +static inline +struct sock *skb_bpf_ext_redirect_fetch (struct sk_buff *skb) +{ +struct skb_bpf_ext *ext =skb_ext_find (skb,SKB_EXT_BPF ); + +return ext->sk_redir; +} + +static inline +void skb_bpf_ext_redirect_clear (struct sk_buff *skb) +{ +struct skb_bpf_ext *ext =skb_ext_find (skb,SKB_EXT_BPF );
Quoted 8 lines
include/linux/skmsg.h static inline bool sk_psock_strp_enabled(struct sk_psock *psock) · line 501
Quoted 5 lines
include/linux/skmsg.h +5
-0+ + ext->flags =0 ; + ext->sk_redir = NULL; +} +#endif /* CONFIG_NET_SOCK_MSG */ So we will have some slight duplication for cb[] variant and ext variant above. I'm OK with that to avoid an allocation.
Not sure what you mean by "duplication", these are removed from TCP_SKB_CB(), so there is clearly no duplication.
Quoted 17 lines
[...]
include/linux/skmsg.h static int sk_psock_verdict_recv(read_descriptor_t *desc, struct sk_buff *skb, · line 1016
Quoted 10 lines
include/linux/skmsg.h +6
-0@@ -1003,11 +1008,17 @@ static int sk_psock_verdict_recv(read_descriptor_t *desc, struct sk_buff *skb, goto out; }skb_set_owner_r (skb, sk); +if (!skb_ext_add (skb,SKB_EXT_BPF )) { + len =0 ; +kfree_skb (skb); + goto out; + } +per packet cost here. Perhaps you can argue small alloc will usually not be noticable in such a large stack, but once we convert over it will be very hard to go back. And I'm looking at optimizing this path now.
This is a price we need to pay to avoid CB, and skb_ext_add() has been used on other fast paths too, for example, tcf_classify_ingress() and mptcp_incoming_options(). So, it is definitely acceptable.
Quoted 11 lines
include/linux/skmsg.h static int sk_psock_verdict_recv(read_descriptor_t *desc, struct sk_buff *skb, · line 1023
Quoted 10 lines
include/linux/skmsg.h +2
-2prog =READ_ONCE (psock->progs.skb_verdict);if (likely (prog)) { -tcp_skb_bpf_redirect_clear (skb); +skb_bpf_ext_redirect_clear (skb); ret =sk_psock_bpf_run (psock, prog, skb); - ret =sk_psock_map_verd (ret,tcp_skb_bpf_redirect_fetch (skb)); + ret =sk_psock_map_verd (ret,skb_bpf_ext_redirect_fetch (skb)); }sk_psock_verdict_apply (psock, skb, ret);
Quoted 2 lines
Thanks for the series Cong. Drop this patch and resubmit carry ACKs forward and then lets revisit this later.
I still believe it is best to stay in this patchset, as it does not change any functionality and is a preparation too. And the next patchset will be UDP/AF_UNIX changes as you suggested, it is very awkward to put this patch into either UDP or AF_UNIX changes.
So, let's keep it in this patchset, and I am happy to address any concerns and open to other ideas than using skb ext.
Thanks.
John Fastabend
maintainerRe: [Patch bpf-next v3 4/5] skmsg: use skb ext instead of TCP_SKB_CB
Quoted 23 lines · Cong Wang
Quoted 16 lines · John Fastabend
Quoted 7 lines · Cong Wang
From: Cong Wang <[email protected]>
Currently
TCP_SKB_CB()is hard-coded in skmsg code, it certainly does not work for any other non-TCP protocols. We can move them to skb ext instead of playing with skb cb, which is harder to make correct.Quoted 6 lines
Cc John Fastabend <[email protected]>Cc Daniel Borkmann <[email protected]>Cc Jakub Sitnicki <[email protected]>Reviewed-by Lorenz Bauer <[email protected]>Signed-off-by Cong Wang <[email protected]>---
Quoted 4 lines
I'm not seeing the advantage of doing this at the moment. We can continue to use cb[] here, which is simpler IMO and use the ext if needed for the other use cases. This is adding a per packet alloc cost that we don't have at the moment as I understand it.
Quoted 7 lines
Hmm? How can we continue using
TCP_SKB_CB()for UDP orAF_UNIX?I am not sure I get your "at the moment" correctly, do you mean I should move this patch to a later patchset, maybe the UDP patchset? At least this patch is needed, no matter by which patchset, so it should not be dropped.
Agree, the skb_bpf_ext{} pieces are needed for UDP and AF_UNIX. Its not required for TCP side though. What I'm suggesting is leave the TCP side as-is, using the cb[] fields. Then use the skb_bpf_ext fields from UDP and AF_UNIX.
Quoted 57 lines
Quoted 50 lines
[...]
Quoted 34 lines
include/linux/skmsg.h +26
-0@@ -459,4 +459,44 @@ static inline bool sk_psock_strp_enabled(struct sk_psock *psock)return false;return !!psock->saved_data_ready; } + +struct skb_bpf_ext { +__u32 flags; +struct sock *sk_redir; +}; + +#if IS_ENABLED (CONFIG_NET_SOCK_MSG ) +static inline +bool skb_bpf_ext_ingress (const struct sk_buff *skb) +{ +struct skb_bpf_ext *ext =skb_ext_find (skb,SKB_EXT_BPF ); + +return ext->flags &BPF_F_INGRESS ; +} + +static inline +void skb_bpf_ext_set_ingress (const struct sk_buff *skb) +{ +struct skb_bpf_ext *ext =skb_ext_find (skb,SKB_EXT_BPF ); + + ext->flags |=BPF_F_INGRESS ; +} + +static inline +struct sock *skb_bpf_ext_redirect_fetch (struct sk_buff *skb) +{include/linux/skmsg.h static inline bool sk_psock_strp_enabled(struct sk_psock *psock) · line 497
Quoted 10 lines
include/linux/skmsg.h +10
-0+struct skb_bpf_ext *ext =skb_ext_find (skb,SKB_EXT_BPF ); + +return ext->sk_redir; +} + +static inline +void skb_bpf_ext_redirect_clear (struct sk_buff *skb) +{ +struct skb_bpf_ext *ext =skb_ext_find (skb,SKB_EXT_BPF ); ++
ext->flags= 0;Quoted 6 lines
include/linux/skmsg.h static inline bool sk_psock_strp_enabled(struct sk_psock *psock) · line 501
Quoted 3 lines
include/linux/skmsg.h +3
-0+ ext->sk_redir = NULL; +} +#endif /* CONFIG_NET_SOCK_MSG */ So we will have some slight duplication for cb[] variant and ext variant above. I'm OK with that to avoid an allocation.
Quoted 2 lines
Not sure what you mean by "duplication", these are removed from
TCP_SKB_CB(), so there is clearly no duplication.
In this patch yes, no duplication. But, I want to leave TCP alone and have it continue to use cb[] to avoid alloc per packet.
Quoted 19 lines
Quoted 4 lines
[...]
Quoted 14 lines
include/linux/skmsg.h static int sk_psock_verdict_recv(read_descriptor_t *desc, struct sk_buff *skb, · line 1016
Quoted 10 lines
include/linux/skmsg.h +6
-0@@ -1003,11 +1008,17 @@ static int sk_psock_verdict_recv(read_descriptor_t *desc, struct sk_buff *skb, goto out; }skb_set_owner_r (skb, sk); +if (!skb_ext_add (skb,SKB_EXT_BPF )) { + len =0 ; +kfree_skb (skb); + goto out; + } +per packet cost here. Perhaps you can argue small alloc will usually not be noticable in such a large stack, but once we convert over it will be very hard to go back. And I'm looking at optimizing this path now.
Quoted 3 lines
This is a price we need to pay to avoid CB, and
skb_ext_add()has been used on other fast paths too, for example,tcf_classify_ingress()andmptcp_incoming_options(). So, it is definitely acceptable.
For TCP case we can continue to use CB and not pay the price. For UDP and AF_UNIX we can do the extra alloc.
The use in tcf_classify_ingress is a miss case so not the common path. If it is/was in the common path I would suggest we rip it out.
Quoted 12 lines
Quoted 12 lines
Quoted 11 lines
include/linux/skmsg.h +2
-2prog =READ_ONCE (psock->progs.skb_verdict);if (likely (prog)) { -tcp_skb_bpf_redirect_clear (skb); +skb_bpf_ext_redirect_clear (skb); ret =sk_psock_bpf_run (psock, prog, skb); - ret =sk_psock_map_verd (ret,tcp_skb_bpf_redirect_fetch (skb)); + ret =sk_psock_map_verd (ret,skb_bpf_ext_redirect_fetch (skb)); }sk_psock_verdict_apply (psock, skb, ret);
Quoted 7 lines
Quoted 2 lines
Thanks for the series Cong. Drop this patch and resubmit carry ACKs forward and then lets revisit this later.
I still believe it is best to stay in this patchset, as it does not change any functionality and is a preparation too. And the next patchset will be UDP/AF_UNIX changes as you suggested, it is very awkward to put this patch into either UDP or
AF_UNIXchanges.
Disagree. It adds extra code to the TCP side that I think is not needed. Any reason the TCP implementation can't continue to use cb[]?
Quoted 3 lines
So, let's keep it in this patchset, and I am happy to address any concerns and open to other ideas than using skb ext.
The idea here is to just use cb[] in TCP case per above. I only scanned your other patches, but presumably this can be patch 1 with just the functions
skb_bpf_ext_ingress()
skb_bpf_ext_set_ingress()
skb_bpf_ext_redirect_fetch()
skb_bpf_ext_redirect_clear()
And none of the removals from TCP side.
.John
Re: [Patch bpf-next v3 4/5] skmsg: use skb ext instead of TCP_SKB_CB
Quoted 3 lines · John Fastabend
For TCP case we can continue to use CB and not pay the price. For UDP and
AF_UNIXwe can do the extra alloc.
I see your point, but specializing TCP case does not give much benefit here, the skmsg code would have to check skb->protocol etc. to decide whether to use TCP_SKB_CB() or skb_ext:
if (skb->protocol == ...)
TCP_SKB_CB(skb) = ...;
else
ext = skb_ext_find(skb);
which looks ugly to me. And I doubt skb->protocol alone is sufficient to distinguish TCP, so we may end up having more checks above.
So do you really want to trade code readability with an extra alloc?
Quoted 4 lines
The use in
tcf_classify_ingressis a miss case so not the common path. If it is/was in the common path I would suggest we rip it out.
Excellent point, what about nf_bridge_unshare()? It is a common path for bridge netfilter, which is also probably why skb ext was introduced (IIRC). secpath_set() seems on a common path for XFRM too.
Are you suggesting to remove them all? ;)
Thanks.
John Fastabend
maintainerRe: [Patch bpf-next v3 4/5] skmsg: use skb ext instead of TCP_SKB_CB
Quoted 14 lines · Cong Wang
Quoted 3 lines · John Fastabend
For TCP case we can continue to use CB and not pay the price. For UDP and
AF_UNIXwe can do the extra alloc.I see your point, but specializing TCP case does not give much benefit here, the skmsg code would have to check
skb->protocoletc. to decide whether to useTCP_SKB_CB()orskb_ext:if (
skb->protocol== ...)
TCP_SKB_CB(skb)= ...;
else
ext =skb_ext_find(skb);
Quoted 4 lines
which looks ugly to me. And I doubt
skb->protocolalone is sufficient to distinguish TCP, so we may end up having more checks above.So do you really want to trade code readability with an extra alloc?
Above is ugly. So I look at where the patch replaces things,
sk_psock_tls_strp_read(), this is TLS specific read hook so can't really work in generic case anyways.
sk_psock_strp_read(), will you have UDP, AF_UNIX stream parsers? Do these even work outside TCP cases.
For these ones: sk_psock_verdict_apply(), sk_psock_verdict_recv(), sk_psock_backlog(), can't we just do some refactoring around their hook points so we know the context. For example sk_psock_tls_verdict_apply is calling sk_psock_skb_redirect(). Why not have a sk_psock_unix_redirect() and a sk_psock_udp_redirect(). There are likely some optimizations we can deploy this way. We've already don this for tls and sk_msg types for example.
Then the helpers will know their types by program type, just use the right variants.
So not suggestiong if/else the checks so much as having per type hooks.
Quoted 9 lines
Quoted 5 lines
The use in
tcf_classify_ingressis a miss case so not the common path. If it is/was in the common path I would suggest we rip it out.Excellent point, what about
nf_bridge_unshare()? It is a common path for bridge netfilter, which is also probably why skb ext was introduced (IIRC).secpath_set()seems on a common path for XFRM too.
Yeah not nice, but we don't use nf_bridge so doesn't bother me.
Quoted 2 lines
Are you suggesting to remove them all? ;)
From the hotpath where I care about perfromance yes.
Quoted 2 lines
Thanks.
Re: [Patch bpf-next v3 4/5] skmsg: use skb ext instead of TCP_SKB_CB
Quoted 36 lines · John Fastabend
Quoted 14 lines · Cong Wang
Quoted 3 lines · John Fastabend
For TCP case we can continue to use CB and not pay the price. For UDP and
AF_UNIXwe can do the extra alloc.I see your point, but specializing TCP case does not give much benefit here, the skmsg code would have to check
skb->protocoletc. to decide whether to useTCP_SKB_CB()orskb_ext:if (
skb->protocol== ...)
TCP_SKB_CB(skb)= ...;
else
ext =skb_ext_find(skb);Quoted 4 lines
which looks ugly to me. And I doubt
skb->protocolalone is sufficient to distinguish TCP, so we may end up having more checks above.So do you really want to trade code readability with an extra alloc?
Above is ugly. So I look at where the patch replaces things,
sk_psock_tls_strp_read(), this is TLS specific read hook so can't really work in generic case anyways.
sk_psock_strp_read(), will you have UDP,AF_UNIXstream parsers? Do these even work outside TCP cases.For these ones:
sk_psock_verdict_apply(),sk_psock_verdict_recv(),sk_psock_backlog(), can't we just do some refactoring around their hook points so we know the context. For examplesk_psock_tls_verdict_applyis callingsk_psock_skb_redirect(). Why not have ask_psock_unix_redirect()and ask_psock_udp_redirect(). There are likely some optimizations we can deploy this way. We've already don this for tls andsk_msgtypes for example.
Quoted 5 lines
Then the helpers will know their types by program type, just use the right variants.
So not suggestiong if/else the checks so much as having per type hooks.
Hmm, but sk_psock_backlog() is still the only one that handles all three above cases, right? It uses TCP_SKB_CB() too and more importantly it is also why we can't use a per-cpu struct here (see bpf_redirect_info).
Thanks.
John Fastabend
maintainerRe: [Patch bpf-next v3 4/5] skmsg: use skb ext instead of TCP_SKB_CB
Quoted 40 lines · Cong Wang
Quoted 39 lines · John Fastabend
Quoted 14 lines · Cong Wang
Quoted 3 lines · John Fastabend
For TCP case we can continue to use CB and not pay the price. For UDP and
AF_UNIXwe can do the extra alloc.I see your point, but specializing TCP case does not give much benefit here, the skmsg code would have to check
skb->protocoletc. to decide whether to useTCP_SKB_CB()orskb_ext:if (
skb->protocol== ...)
TCP_SKB_CB(skb)= ...;
else
ext =skb_ext_find(skb);Quoted 4 lines
which looks ugly to me. And I doubt
skb->protocolalone is sufficient to distinguish TCP, so we may end up having more checks above.So do you really want to trade code readability with an extra alloc?
Above is ugly. So I look at where the patch replaces things,
sk_psock_tls_strp_read(), this is TLS specific read hook so can't really work in generic case anyways.
sk_psock_strp_read(), will you have UDP,AF_UNIXstream parsers? Do these even work outside TCP cases.For these ones:
sk_psock_verdict_apply(),sk_psock_verdict_recv(),sk_psock_backlog(), can't we just do some refactoring around their hook points so we know the context. For examplesk_psock_tls_verdict_applyis callingsk_psock_skb_redirect(). Why not have ask_psock_unix_redirect()and ask_psock_udp_redirect(). There are likely some optimizations we can deploy this way. We've already don this for tls andsk_msgtypes for example.Then the helpers will know their types by program type, just use the right variants.
Quoted 6 lines
Quoted 2 lines
So not suggestiong if/else the checks so much as having per type hooks.
Hmm, but
sk_psock_backlog()is still the only one that handles all three above cases, right? It usesTCP_SKB_CB()too and more importantly it is also why we can't use a per-cpu struct here (seebpf_redirect_info).
Right, but the workqueue is created at init time where we will know the socket type based on the program/map types so can build the redirect backlog queue there based on the type needed. I also have a patch in mind that would do more specific TCP things in that code anyways. I can flush it out this week if anyone cares. The idea is we are wasting lots of cycles using skb_send_sock_locked when we can just inject the packet directlyy into the tcp stack.
Also on the original patch here, we can't just kfree_skb() on alloc errors because that will look like a data drop. Errors need to be handled gracefully without dropping data. At least in the TCP case, but probably also in UDP and AF_UNIX cases as well. Original code was pretty loose in this regard, but it caused users to write bug reports and then I've been fixing most of them. If you see more cases let me know.
Quoted 2 lines
Thanks.
Re: [Patch bpf-next v3 4/5] skmsg: use skb ext instead of TCP_SKB_CB
Quoted 45 lines · John Fastabend
Quoted 43 lines · Cong Wang
Quoted 41 lines · John Fastabend
Quoted 14 lines · Cong Wang
Quoted 3 lines · John Fastabend
For TCP case we can continue to use CB and not pay the price. For UDP and
AF_UNIXwe can do the extra alloc.I see your point, but specializing TCP case does not give much benefit here, the skmsg code would have to check
skb->protocoletc. to decide whether to useTCP_SKB_CB()orskb_ext:if (
skb->protocol== ...)
TCP_SKB_CB(skb)= ...;
else
ext =skb_ext_find(skb);Quoted 4 lines
which looks ugly to me. And I doubt
skb->protocolalone is sufficient to distinguish TCP, so we may end up having more checks above.So do you really want to trade code readability with an extra alloc?
Above is ugly. So I look at where the patch replaces things,
sk_psock_tls_strp_read(), this is TLS specific read hook so can't really work in generic case anyways.
sk_psock_strp_read(), will you have UDP,AF_UNIXstream parsers? Do these even work outside TCP cases.For these ones:
sk_psock_verdict_apply(),sk_psock_verdict_recv(),sk_psock_backlog(), can't we just do some refactoring around their hook points so we know the context. For examplesk_psock_tls_verdict_applyis callingsk_psock_skb_redirect(). Why not have ask_psock_unix_redirect()and ask_psock_udp_redirect(). There are likely some optimizations we can deploy this way. We've already don this for tls andsk_msgtypes for example.Then the helpers will know their types by program type, just use the right variants.
So not suggestiong if/else the checks so much as having per type hooks.
Quoted 7 lines
Quoted 3 lines
Hmm, but
sk_psock_backlog()is still the only one that handles all three above cases, right? It usesTCP_SKB_CB()too and more importantly it is also why we can't use a per-cpu struct here (seebpf_redirect_info).Right, but the workqueue is created at init time where we will know the socket type based on the program/map types so can build the redirect backlog queue there based on the type needed. I also have a patch in
Hmm? How could a socket type match the skb type when we redirect across-protocol?
In my use case, I want to redirect an AF_UNIX skb to a UDP socket, clearly checking the UDP socket workqueue can't find out it is an AF_UNIX skb. It has to be a per-skb check.
Quoted 4 lines
mind that would do more specific TCP things in that code anyways. I can flush it out this week if anyone cares. The idea is we are wasting lots of cycles using
skb_send_sock_lockedwhen we can just inject the packet directlyy into the tcp stack.
Actually I did try the same, it clearly doesn't work for cross-protocol.
Anyway, please let me know what your suggestion for skb ext here?
It looks like we either have some ugly packet type checks, or just use the skb ext. I don't see any other way yet, I also explored the struct sk_buff again and still can not find anything we can reuse. (_skb_refdst can only be reused very briefly with tcp_skb_tsorted_save().)
Therefore, I believe using skb ext is still the best solution here.
Quoted 8 lines
Also on the original patch here, we can't just
kfree_skb()on alloc errors because that will look like a data drop. Errors need to be handled gracefully without dropping data. At least in the TCP case, but probably also in UDP andAF_UNIXcases as well. Original code was pretty loose in this regard, but it caused users to write bug reports and then I've been fixing most of them. If you see more cases let me know.
What's your suggestion here? Return -EAGAIN? But it requires the caller put it in a loop to be graceful, but we can't do it in, for example, sk_psock_tls_strp_read().
Thanks.
Re: [Patch bpf-next v3 4/5] skmsg: use skb ext instead of TCP_SKB_CB
In reply to John FastabendQuoted 16 lines · John Fastabend
Quoted 7 lines · Cong Wang
From: Cong Wang <[email protected]>
Currently
TCP_SKB_CB()is hard-coded in skmsg code, it certainly does not work for any other non-TCP protocols. We can move them to skb ext instead of playing with skb cb, which is harder to make correct.Quoted 6 lines
Cc John Fastabend <[email protected]>Cc Daniel Borkmann <[email protected]>Cc Jakub Sitnicki <[email protected]>Reviewed-by Lorenz Bauer <[email protected]>Signed-off-by Cong Wang <[email protected]>---
Quoted 4 lines
I'm not seeing the advantage of doing this at the moment. We can continue to use cb[] here, which is simpler IMO and use the ext if needed for the other use cases. This is adding a per packet alloc cost that we don't have at the moment as I understand it.
John, do you have a benchmark we can look at? Right now we're arguing in the abstract.
John Fastabend
maintainerRe: [Patch bpf-next v3 4/5] skmsg: use skb ext instead of TCP_SKB_CB
Quoted 17 lines · Lorenz Bauer
Quoted 16 lines · John Fastabend
Quoted 13 lines · Cong Wang
From: Cong Wang <[email protected]>
Currently
TCP_SKB_CB()is hard-coded in skmsg code, it certainly does not work for any other non-TCP protocols. We can move them to skb ext instead of playing with skb cb, which is harder to make correct.Cc John Fastabend <[email protected]>Cc Daniel Borkmann <[email protected]>Cc Jakub Sitnicki <[email protected]>Reviewed-by Lorenz Bauer <[email protected]>Signed-off-by Cong Wang <[email protected]>---
Quoted 7 lines
Quoted 4 lines
I'm not seeing the advantage of doing this at the moment. We can continue to use cb[] here, which is simpler IMO and use the ext if needed for the other use cases. This is adding a per packet alloc cost that we don't have at the moment as I understand it.
John, do you have a benchmark we can look at? Right now we're arguing in the abstract.
Sure, but looks like Cong found some spare fields in sk_buff so that looks much nicer.
I'll mess aound a bit with our benchmarks and see where we can publish them. It would be good to have some repeatable tests here folks can use.
Thanks,
John
PATCH v3 5/5 sock_map: rename skb_parser and skb_verdict
In reply to Cong WangFrom: Cong Wang <[email protected]>
These two eBPF programs are tied to BPF_SK_SKB_STREAM_PARSER and BPF_SK_SKB_STREAM_VERDICT, rename them to reflect the fact they are only used for TCP. And save the name 'skb_verdict' for general use later.
--- include/linux/skmsg.h | 8 +-- net/core/skmsg.c | 14 ++--- net/core/sock_map.c | 60 +++++++++---------- .../selftests/bpf/prog_tests/sockmap_listen.c | 8 +-- .../selftests/bpf/progs/test_sockmap_listen.c | 4 +- 5 files changed, 47 insertions(+), 47 deletions(-)
include/linux/skmsg.h +4 -4
@@ -56,8 +56,8 @@ struct sk_msg {struct sk_psock_progs {struct bpf_prog *msg_parser; -struct bpf_prog *skb_parser; -struct bpf_prog *skb_verdict; +struct bpf_prog *stream_parser; +struct bpf_prog *stream_verdict; };enum sk_psock_state_bits {
@@ -447,8 +447,8 @@ static inline int psock_replace_prog(struct bpf_prog **pprog,static inline void psock_progs_drop (struct sk_psock_progs *progs) {psock_set_prog (&progs->msg_parser,NULL ); -psock_set_prog (&progs->skb_parser,NULL ); -psock_set_prog (&progs->skb_verdict,NULL ); +psock_set_prog (&progs->stream_parser,NULL ); +psock_set_prog (&progs->stream_verdict,NULL ); }int sk_psock_tls_strp_read (struct sk_psock *psock,struct sk_buff *skb);
net/core/skmsg.c +7 -7
@@ -684,9 +684,9 @@ void sk_psock_drop(struct sock *sk, struct sk_psock *psock)write_lock_bh (&sk->sk_callback_lock);sk_psock_restore_proto (sk, psock);rcu_assign_sk_user_data (sk,NULL ); -if (psock->progs.skb_parser) +if (psock->progs.stream_parser)sk_psock_stop_strp (sk, psock); -else if (psock->progs.skb_verdict) +else if (psock->progs.stream_verdict)sk_psock_stop_verdict (sk, psock);write_unlock_bh (&sk->sk_callback_lock);sk_psock_clear_state (psock,SK_PSOCK_TX_ENABLED );
@@ -799,7 +799,7 @@ int sk_psock_tls_strp_read(struct sk_psock *psock, struct sk_buff *skb)return __SK_DROP;rcu_read_lock (); - prog =READ_ONCE (psock->progs.skb_verdict); + prog =READ_ONCE (psock->progs.stream_verdict);if (likely (prog)) {/* We skip full set_owner_r here because if we do a SK_PASS * or SK_DROP we can skip skb memory accounting and use the
@@ -893,7 +893,7 @@ static void sk_psock_strp_read(struct strparser *strp, struct sk_buff *skb)kfree_skb (skb); goto out; } - prog =READ_ONCE (psock->progs.skb_verdict); + prog =READ_ONCE (psock->progs.stream_verdict);if (likely (prog)) {skb_bpf_ext_redirect_clear (skb); ret =sk_psock_bpf_run (psock, prog, skb);
@@ -916,7 +916,7 @@ static int sk_psock_strp_parse(struct strparser *strp, struct sk_buff *skb)int ret = skb->len;rcu_read_lock (); - prog =READ_ONCE (psock->progs.skb_parser); + prog =READ_ONCE (psock->progs.stream_parser);if (likely (prog)) { skb->sk = psock->sk; ret =sk_psock_bpf_run (psock, prog, skb);
@@ -979,7 +979,7 @@ void sk_psock_stop_strp(struct sock *sk, struct sk_psock *psock)void sk_psock_done_strp (struct sk_psock *psock) {/* Parser has been stopped */ -if (psock->progs.skb_parser) +if (psock->progs.stream_parser)strp_done (&psock->strp); }#endif
@@ -1014,7 +1014,7 @@ static int sk_psock_verdict_recv(read_descriptor_t *desc, struct sk_buff *skb, goto out; } - prog =READ_ONCE (psock->progs.skb_verdict); + prog =READ_ONCE (psock->progs.stream_verdict);if (likely (prog)) {skb_bpf_ext_redirect_clear (skb); ret =sk_psock_bpf_run (psock, prog, skb);
net/core/sock_map.c +30 -30
115 lines not on this page · raw message
tools/testing/selftests/bpf/prog_tests/sockmap_listen.c +4 -4
@@ -1014,8 +1014,8 @@ static void test_skb_redir_to_connected(struct test_sockmap_listen *skel, struct bpf_map *inner_map, int family, int sotype) { - int verdict = bpf_program__fd(skel->progs.prog_skb_verdict); -int parser =bpf_program__fd (skel->progs.prog_skb_parser); + int verdict = bpf_program__fd(skel->progs.prog_stream_verdict); +int parser =bpf_program__fd (skel->progs.prog_stream_parser);int verdict_map =bpf_map__fd (skel->maps.verdict_map);int sock_map =bpf_map__fd (inner_map);int err;
@@ -1125,8 +1125,8 @@ static void test_skb_redir_to_listening(struct test_sockmap_listen *skel, struct bpf_map *inner_map, int family, int sotype) { - int verdict = bpf_program__fd(skel->progs.prog_skb_verdict); -int parser =bpf_program__fd (skel->progs.prog_skb_parser); + int verdict = bpf_program__fd(skel->progs.prog_stream_verdict); +int parser =bpf_program__fd (skel->progs.prog_stream_parser);int verdict_map =bpf_map__fd (skel->maps.verdict_map);int sock_map =bpf_map__fd (inner_map);int err;
tools/testing/selftests/bpf/progs/test_sockmap_listen.c +2 -2
@@ -31,13 +31,13 @@ struct {static volatile bool test_sockmap;/* toggled by user-space */ SEC("sk_skb/stream_parser") -int prog_skb_parser (struct __sk_buff *skb) +int prog_stream_parser (struct __sk_buff *skb) {return skb->len; } SEC("sk_skb/stream_verdict") -int prog_skb_verdict (struct __sk_buff *skb) +int prog_stream_verdict (struct __sk_buff *skb) {unsigned int *count;__u32 zero =0 ;
Signature · 1 line
2.25.1
RE: [Patch bpf-next v3 5/5] sock_map: rename skb_parser and skb_verdict
Quoted 7 lines · Cong Wang
From: Cong Wang <[email protected]>
These two eBPF programs are tied to
BPF_SK_SKB_STREAM_PARSERandBPF_SK_SKB_STREAM_VERDICT, rename them to reflect the fact they are only used for TCP. And save the name 'skb_verdict' for general use later.
Quoted 6 lines
Cc John Fastabend <[email protected]>Cc Daniel Borkmann <[email protected]>Cc Jakub Sitnicki <[email protected]>Reviewed-by Lorenz Bauer <[email protected]>Signed-off-by Cong Wang <[email protected]>---