- Kernel - mailweb.openeuler.org

[PATCH OLK-6.6 v3 0/4] Terrace Service Acceleration
by Zhengchao Shao 30 Dec '23

30 Dec '23

Terrace Service Acceleration. Liu Jian (1): bpf: Add new bpf helper to get SO_ORIGINAL_DST/REPLY_SRC Lu Wei (1): net: core: Add a GID field to struct sock. Wang Yufen (1): bpf, sockmap: Add sk_rmem_alloc check for sockmap Zhengchao Shao (1): bpf: Add bpf_get_sockops_uid_gid helper function include/net/netfilter/nf_conntrack.h | 6 +++ include/net/sock.h | 18 +++++++ include/uapi/linux/bpf.h | 2 + include/uapi/linux/netfilter_ipv4.h | 2 + net/Kconfig | 7 +++ net/core/filter.c | 64 +++++++++++++++++++++++++ net/core/skmsg.c | 10 +++- net/core/sock.c | 6 +++ net/ipv4/tcp_bpf.c | 10 ++++ net/netfilter/nf_conntrack_proto.c | 71 ++++++++++++++++++++++++++++ net/socket.c | 8 +++- 11 files changed, 201 insertions(+), 3 deletions(-) -- 2.34.1

2 5

[PATCH OLK-6.6 0/1] Support priority load balance for qos scheduler
by Song Zhang 29 Dec '23

29 Dec '23

Song Zhang (1): sched: Introduce priority load balance for qos scheduler arch/arm64/configs/openeuler_defconfig | 1 + arch/x86/configs/openeuler_defconfig | 1 + init/Kconfig | 9 +++ kernel/sched/core.c | 3 + kernel/sched/fair.c | 83 +++++++++++++++++++++++++- kernel/sched/sched.h | 3 + 6 files changed, 99 insertions(+), 1 deletion(-) -- 2.33.0

2 2

[PATCH openEuler-1.0-LTS] Bluetooth: af_bluetooth: Fix Use-After-Free in bt_sock_recvmsg
by Ziyang Xuan 29 Dec '23

29 Dec '23

From: Hyunwoo Kim <v4bel(a)theori.io> mainline inclusion from mainline-v6.7-rc7 commit 2e07e8348ea454615e268222ae3fc240421be768 category: bugfix bugzilla: https://gitee.com/src-openeuler/kernel/issues/I8RXOD CVE: CVE-2023-51779 Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?… -------------------------------- This can cause a race with bt_sock_ioctl() because bt_sock_recvmsg() gets the skb from sk->sk_receive_queue and then frees it without holding lock_sock. A use-after-free for a skb occurs with the following flow. ``` bt_sock_recvmsg() -> skb_recv_datagram() -> skb_free_datagram() bt_sock_ioctl() -> skb_peek() ``` Add lock_sock to bt_sock_recvmsg() to fix this issue. Cc: stable(a)vger.kernel.org Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Signed-off-by: Hyunwoo Kim <v4bel(a)theori.io> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz(a)intel.com> Conflicts: net/bluetooth/af_bluetooth.c Signed-off-by: Ziyang Xuan <william.xuanziyang(a)huawei.com> --- net/bluetooth/af_bluetooth.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/net/bluetooth/af_bluetooth.c b/net/bluetooth/af_bluetooth.c index ee60c30f3be2..363f85c617bb 100644 --- a/net/bluetooth/af_bluetooth.c +++ b/net/bluetooth/af_bluetooth.c @@ -263,11 +263,14 @@ int bt_sock_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, if (flags & MSG_OOB) return -EOPNOTSUPP; + lock_sock(sk); + skb = skb_recv_datagram(sk, flags, noblock, &err); if (!skb) { if (sk->sk_shutdown & RCV_SHUTDOWN) - return 0; + err = 0; + release_sock(sk); return err; } @@ -290,6 +293,8 @@ int bt_sock_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, skb_free_datagram(sk, skb); + release_sock(sk); + if (flags & MSG_TRUNC) copied = skblen; -- 2.25.1

2 1

[PATCH OLK-5.10] Bluetooth: af_bluetooth: Fix Use-After-Free in bt_sock_recvmsg
by Ziyang Xuan 29 Dec '23

29 Dec '23

From: Hyunwoo Kim <v4bel(a)theori.io> mainline inclusion from mainline-v6.7-rc7 commit 2e07e8348ea454615e268222ae3fc240421be768 category: bugfix bugzilla: https://gitee.com/src-openeuler/kernel/issues/I8RXOD CVE: CVE-2023-51779 Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?… -------------------------------- This can cause a race with bt_sock_ioctl() because bt_sock_recvmsg() gets the skb from sk->sk_receive_queue and then frees it without holding lock_sock. A use-after-free for a skb occurs with the following flow. ``` bt_sock_recvmsg() -> skb_recv_datagram() -> skb_free_datagram() bt_sock_ioctl() -> skb_peek() ``` Add lock_sock to bt_sock_recvmsg() to fix this issue. Cc: stable(a)vger.kernel.org Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Signed-off-by: Hyunwoo Kim <v4bel(a)theori.io> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz(a)intel.com> Conflicts: net/bluetooth/af_bluetooth.c Signed-off-by: Ziyang Xuan <william.xuanziyang(a)huawei.com> --- net/bluetooth/af_bluetooth.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/net/bluetooth/af_bluetooth.c b/net/bluetooth/af_bluetooth.c index 2f87f57e7a4f..14a917e70f3e 100644 --- a/net/bluetooth/af_bluetooth.c +++ b/net/bluetooth/af_bluetooth.c @@ -263,11 +263,14 @@ int bt_sock_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, if (flags & MSG_OOB) return -EOPNOTSUPP; + lock_sock(sk); + skb = skb_recv_datagram(sk, flags, noblock, &err); if (!skb) { if (sk->sk_shutdown & RCV_SHUTDOWN) - return 0; + err = 0; + release_sock(sk); return err; } @@ -293,6 +296,8 @@ int bt_sock_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, skb_free_datagram(sk, skb); + release_sock(sk); + if (flags & MSG_TRUNC) copied = skblen; -- 2.25.1

2 2

[PATCH OLK-6.6 v4] tcp_comp: implement tcp compression
by Lu Wei 29 Dec '23

29 Dec '23

hulk inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I4PNEK CVE: NA ------------------------------------------------- Signed-off-by: Wang Yufen <wangyufen(a)huawei.com> Signed-off-by: Wei Yongjun <weiyongjun1(a)huawei.com> Signed-off-by: Lu Wei <luwei32(a)huawei.com> --- arch/arm64/configs/openeuler_defconfig | 1 + arch/x86/configs/openeuler_defconfig | 1 + include/linux/tcp.h | 8 +- include/net/inet_sock.h | 3 + include/net/sock.h | 1 + include/net/tcp.h | 40 ++ net/ipv4/Kconfig | 10 + net/ipv4/Makefile | 1 + net/ipv4/syncookies.c | 4 + net/ipv4/sysctl_net_ipv4.c | 42 ++ net/ipv4/tcp.c | 5 + net/ipv4/tcp_comp.c | 912 +++++++++++++++++++++++++ net/ipv4/tcp_input.c | 26 + net/ipv4/tcp_ipv4.c | 2 + net/ipv4/tcp_minisocks.c | 3 + net/ipv4/tcp_output.c | 52 ++ net/ipv6/syncookies.c | 4 + 17 files changed, 1114 insertions(+), 1 deletion(-) create mode 100644 net/ipv4/tcp_comp.c diff --git a/arch/arm64/configs/openeuler_defconfig b/arch/arm64/configs/openeuler_defconfig index 33ba39711884..abc33d2c29a6 100644 --- a/arch/arm64/configs/openeuler_defconfig +++ b/arch/arm64/configs/openeuler_defconfig @@ -1222,6 +1222,7 @@ CONFIG_DEFAULT_CUBIC=y # CONFIG_DEFAULT_RENO is not set CONFIG_DEFAULT_TCP_CONG="cubic" CONFIG_TCP_MD5SIG=y +CONFIG_TCP_COMP=y CONFIG_IPV6=y CONFIG_IPV6_ROUTER_PREF=y CONFIG_IPV6_ROUTE_INFO=y diff --git a/arch/x86/configs/openeuler_defconfig b/arch/x86/configs/openeuler_defconfig index 44040b835333..177a6fdcce58 100644 --- a/arch/x86/configs/openeuler_defconfig +++ b/arch/x86/configs/openeuler_defconfig @@ -1244,6 +1244,7 @@ CONFIG_DEFAULT_CUBIC=y # CONFIG_DEFAULT_RENO is not set CONFIG_DEFAULT_TCP_CONG="cubic" CONFIG_TCP_MD5SIG=y +CONFIG_TCP_COMP=y CONFIG_IPV6=y CONFIG_IPV6_ROUTER_PREF=y CONFIG_IPV6_ROUTE_INFO=y diff --git a/include/linux/tcp.h b/include/linux/tcp.h index 3c5efeeb024f..d94a84c291f0 100644 --- a/include/linux/tcp.h +++ b/include/linux/tcp.h @@ -123,7 +123,10 @@ struct tcp_options_received { snd_wscale : 4, /* Window scaling received from sender */ rcv_wscale : 4; /* Window scaling to send to receiver */ u8 saw_unknown:1, /* Received unknown option */ - unused:7; +#if IS_ENABLED(CONFIG_TCP_COMP) + comp_ok:1, /* COMP seen on SYN packet */ +#endif + unused:6; u8 num_sacks; /* Number of SACK blocks */ u16 user_mss; /* mss requested by user in ioctl */ u16 mss_clamp; /* Maximal mss, negotiated at connection setup */ @@ -136,6 +139,9 @@ static inline void tcp_clear_options(struct tcp_options_received *rx_opt) #if IS_ENABLED(CONFIG_SMC) rx_opt->smc_ok = 0; #endif +#if IS_ENABLED(CONFIG_TCP_COMP) + rx_opt->comp_ok = 0; +#endif } /* This is the max number of SACKS that we'll generate and process. It's safe diff --git a/include/net/inet_sock.h b/include/net/inet_sock.h index 2de0e4d4a027..13c1f1e83de5 100644 --- a/include/net/inet_sock.h +++ b/include/net/inet_sock.h @@ -87,6 +87,9 @@ struct inet_request_sock { ecn_ok : 1, acked : 1, no_srccheck: 1, +#if IS_ENABLED(CONFIG_TCP_COMP) + comp_ok : 1, +#endif smc_ok : 1; u32 ir_mark; union { diff --git a/include/net/sock.h b/include/net/sock.h index 7753354d59c0..c86845136ec5 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -959,6 +959,7 @@ enum sock_flags { SOCK_XDP, /* XDP is attached */ SOCK_TSTAMP_NEW, /* Indicates 64 bit timestamps always */ SOCK_RCVMARK, /* Receive SO_MARK ancillary data with packet */ + SOCK_COMP, }; #define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE)) diff --git a/include/net/tcp.h b/include/net/tcp.h index e9d387fffe22..cb33a2c46b2f 100644 --- a/include/net/tcp.h +++ b/include/net/tcp.h @@ -195,6 +195,7 @@ void tcp_time_wait(struct sock *sk, int state, int timeo); */ #define TCPOPT_FASTOPEN_MAGIC 0xF989 #define TCPOPT_SMC_MAGIC 0xE2D4C3D9 +#define TCPOPT_COMP_MAGIC 0x7954 /* * TCP option lengths @@ -208,6 +209,7 @@ void tcp_time_wait(struct sock *sk, int state, int timeo); #define TCPOLEN_FASTOPEN_BASE 2 #define TCPOLEN_EXP_FASTOPEN_BASE 4 #define TCPOLEN_EXP_SMC_BASE 6 +#define TCPOLEN_EXP_COMP_BASE 4 /* But this is what stacks really send out. */ #define TCPOLEN_TSTAMP_ALIGNED 12 @@ -2531,4 +2533,42 @@ static inline u64 tcp_transmit_time(const struct sock *sk) return 0; } +#if IS_ENABLED(CONFIG_TCP_COMP) +extern struct static_key_false tcp_have_comp; + +extern unsigned long *sysctl_tcp_compression_ports; +extern int sysctl_tcp_compression_local; + +bool tcp_syn_comp_enabled(struct sock *sk); +bool tcp_synack_comp_enabled(struct sock *sk, + const struct inet_request_sock *ireq); +void tcp_init_compression(struct sock *sk); +void tcp_cleanup_compression(struct sock *sk); +int tcp_comp_init(void); +#else +static inline bool tcp_syn_comp_enabled(struct tcp_sock *tp) +{ + return false; +} + +static inline bool tcp_synack_comp_enabled(struct sock *sk, + const struct inet_request_sock *ireq) +{ + return false; +} + +static inline void tcp_init_compression(struct sock *sk) +{ +} + +static inline void tcp_cleanup_compression(struct sock *sk) +{ +} + +static inline int tcp_comp_init(void) +{ + return 0; +} +#endif + #endif /* _TCP_H */ diff --git a/net/ipv4/Kconfig b/net/ipv4/Kconfig index 2dfb12230f08..a4405ea38338 100644 --- a/net/ipv4/Kconfig +++ b/net/ipv4/Kconfig @@ -751,3 +751,13 @@ config TCP_MD5SIG on the Internet. If unsure, say N. + +config TCP_COMP + bool "TCP: Transport Layer Compression support" + depends on CRYPTO_ZSTD=y + select STREAM_PARSER + help + Enable kernel payload compression support for TCP protocol. This allows + payload compression handling of the TCP protocol to be done in-kernel. + + If unsure, say Y. diff --git a/net/ipv4/Makefile b/net/ipv4/Makefile index b18ba8ef93ad..aa458d9f534a 100644 --- a/net/ipv4/Makefile +++ b/net/ipv4/Makefile @@ -65,6 +65,7 @@ obj-$(CONFIG_TCP_CONG_ILLINOIS) += tcp_illinois.o obj-$(CONFIG_NET_SOCK_MSG) += tcp_bpf.o obj-$(CONFIG_BPF_SYSCALL) += udp_bpf.o obj-$(CONFIG_NETLABEL) += cipso_ipv4.o +obj-$(CONFIG_TCP_COMP) += tcp_comp.o obj-$(CONFIG_XFRM) += xfrm4_policy.o xfrm4_state.o xfrm4_input.o \ xfrm4_output.o xfrm4_protocol.o diff --git a/net/ipv4/syncookies.c b/net/ipv4/syncookies.c index 3b4dafefb4b0..584b29c154e8 100644 --- a/net/ipv4/syncookies.c +++ b/net/ipv4/syncookies.c @@ -391,6 +391,10 @@ struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb) if (IS_ENABLED(CONFIG_SMC)) ireq->smc_ok = 0; +#if IS_ENABLED(CONFIG_TCP_COMP) + ireq->comp_ok = 0; +#endif + ireq->ir_iif = inet_request_bound_dev_if(sk, skb); /* We throwed the options of the initial SYN away, so we hope diff --git a/net/ipv4/sysctl_net_ipv4.c b/net/ipv4/sysctl_net_ipv4.c index b17eb28a9690..f212133b05e9 100644 --- a/net/ipv4/sysctl_net_ipv4.c +++ b/net/ipv4/sysctl_net_ipv4.c @@ -469,6 +469,30 @@ static int proc_fib_multipath_hash_fields(struct ctl_table *table, int write, } #endif +#if IS_ENABLED(CONFIG_TCP_COMP) +static int proc_tcp_compression_ports(struct ctl_table *table, int write, + void __user *buffer, size_t *lenp, + loff_t *ppos) +{ + unsigned long *bitmap = *(unsigned long **)table->data; + unsigned long bitmap_len = table->maxlen; + int ret; + + ret = proc_do_large_bitmap(table, write, buffer, lenp, ppos); + if (write && ret == 0) { + if (bitmap_empty(bitmap, bitmap_len)) { + if (static_key_enabled(&tcp_have_comp)) + static_branch_disable(&tcp_have_comp); + } else { + if (!static_key_enabled(&tcp_have_comp)) + static_branch_enable(&tcp_have_comp); + } + } + + return ret; +} +#endif + static struct ctl_table ipv4_table[] = { { .procname = "tcp_max_orphans", @@ -587,6 +611,24 @@ static struct ctl_table ipv4_table[] = { .mode = 0644, .proc_handler = proc_dointvec, }, +#if IS_ENABLED(CONFIG_TCP_COMP) + { + .procname = "tcp_compression_ports", + .data = &sysctl_tcp_compression_ports, + .maxlen = 65536, + .mode = 0644, + .proc_handler = proc_tcp_compression_ports, + }, + { + .procname = "tcp_compression_local", + .data = &sysctl_tcp_compression_local, + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = proc_dointvec_minmax, + .extra1 = SYSCTL_ZERO, + .extra2 = SYSCTL_ONE, + }, +#endif { } }; diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c index 3d3a24f79573..2703be8a7316 100644 --- a/net/ipv4/tcp.c +++ b/net/ipv4/tcp.c @@ -302,6 +302,10 @@ DEFINE_STATIC_KEY_FALSE(tcp_have_smc); EXPORT_SYMBOL(tcp_have_smc); #endif +#if IS_ENABLED(CONFIG_TCP_COMP) +DEFINE_STATIC_KEY_FALSE(tcp_have_comp); +#endif + /* * Current number of TCP sockets. */ @@ -4707,5 +4711,6 @@ void __init tcp_init(void) tcp_metrics_init(); BUG_ON(tcp_register_congestion_control(&tcp_reno) != 0); tcp_tasklet_init(); + tcp_comp_init(); mptcp_init(); } diff --git a/net/ipv4/tcp_comp.c b/net/ipv4/tcp_comp.c new file mode 100644 index 000000000000..53157a413b58 --- /dev/null +++ b/net/ipv4/tcp_comp.c @@ -0,0 +1,912 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * TCP compression support + * + * Copyright(c) 2021 Huawei Technologies Co., Ltd + */ + +#include <linux/skmsg.h> +#include <linux/zstd.h> + +#define TCP_COMP_MAX_PADDING 64 +#define TCP_COMP_DATA_SIZE 65536 +#define TCP_COMP_SCRATCH_SIZE (TCP_COMP_DATA_SIZE - 1) +#define TCP_COMP_MAX_CSIZE (TCP_COMP_SCRATCH_SIZE + TCP_COMP_MAX_PADDING) +#define TCP_COMP_ALLOC_ORDER get_order(TCP_COMP_DATA_SIZE) +#define TCP_COMP_MAX_WINDOWLOG 17 +#define TCP_COMP_MAX_INPUT (1 << TCP_COMP_MAX_WINDOWLOG) + +#define TCP_COMP_SEND_PENDING 1 +#define ZSTD_COMP_DEFAULT_LEVEL 1 + +static unsigned long tcp_compression_ports[65536 / 8]; + +unsigned long *sysctl_tcp_compression_ports = tcp_compression_ports; +int sysctl_tcp_compression_local __read_mostly; + +static struct proto tcp_prot_override; + +struct tcp_comp_context_tx { + ZSTD_CStream *cstream; + void *cworkspace; + void *plaintext_data; + void *compressed_data; + struct sk_msg msg; + bool in_tcp_sendpages; +}; + +struct tcp_comp_context_rx { + ZSTD_DStream *dstream; + void *dworkspace; + void *plaintext_data; + + struct strparser strp; + void (*saved_data_ready)(struct sock *sk); + struct sk_buff *pkt; + struct sk_buff *dpkt; +}; + +struct tcp_comp_context { + struct rcu_head rcu; + + struct proto *sk_proto; + void (*sk_write_space)(struct sock *sk); + + struct tcp_comp_context_tx tx; + struct tcp_comp_context_rx rx; + + unsigned long flags; +}; + +static bool tcp_comp_is_write_pending(struct tcp_comp_context *ctx) +{ + return test_bit(TCP_COMP_SEND_PENDING, &ctx->flags); +} + +static void tcp_comp_err_abort(struct sock *sk, int err) +{ + sk->sk_err = err; + sk->sk_error_report(sk); +} + +static bool tcp_comp_enabled(__be32 saddr, __be32 daddr, int port) +{ + if (!sysctl_tcp_compression_local && + (saddr == daddr || ipv4_is_loopback(daddr))) + return false; + + return test_bit(port, sysctl_tcp_compression_ports); +} + +bool tcp_syn_comp_enabled(struct sock *sk) +{ + struct inet_sock *inet = inet_sk(sk); + + return tcp_comp_enabled(inet->inet_saddr, inet->inet_daddr, + ntohs(inet->inet_dport)); +} + +bool tcp_synack_comp_enabled(struct sock *sk, + const struct inet_request_sock *ireq) +{ + struct inet_sock *inet = inet_sk(sk); + + if (!ireq->comp_ok) + return false; + + return tcp_comp_enabled(ireq->ir_loc_addr, ireq->ir_rmt_addr, + ntohs(inet->inet_sport)); +} + +static struct tcp_comp_context *comp_get_ctx(const struct sock *sk) +{ + struct inet_connection_sock *icsk = inet_csk(sk); + + return (__force void *)icsk->icsk_ulp_data; +} + +static int tcp_comp_tx_context_init(struct tcp_comp_context *ctx) +{ + ZSTD_parameters params; + int csize; + + params = ZSTD_getParams(ZSTD_COMP_DEFAULT_LEVEL, PAGE_SIZE, 0); + csize = zstd_cstream_workspace_bound(&params.cParams); + if (csize <= 0) + return -EINVAL; + + ctx->tx.cworkspace = kmalloc(csize, GFP_KERNEL); + if (!ctx->tx.cworkspace) + return -ENOMEM; + + ctx->tx.cstream = zstd_init_cstream(&params, 0, ctx->tx.cworkspace, + csize); + if (!ctx->tx.cstream) + goto err_cstream; + + ctx->tx.plaintext_data = kvmalloc(TCP_COMP_SCRATCH_SIZE, GFP_KERNEL); + if (!ctx->tx.plaintext_data) + goto err_cstream; + + ctx->tx.compressed_data = kvmalloc(TCP_COMP_MAX_CSIZE, GFP_KERNEL); + if (!ctx->tx.compressed_data) + goto err_compressed; + + return 0; + +err_compressed: + kvfree(ctx->tx.plaintext_data); + ctx->tx.plaintext_data = NULL; +err_cstream: + kfree(ctx->tx.cworkspace); + ctx->tx.cworkspace = NULL; + + return -ENOMEM; +} + +static void *tcp_comp_get_tx_stream(struct sock *sk) +{ + struct tcp_comp_context *ctx = comp_get_ctx(sk); + + if (!ctx->tx.plaintext_data) + tcp_comp_tx_context_init(ctx); + + return ctx->tx.plaintext_data; +} + +static int alloc_compressed_msg(struct sock *sk, int len) +{ + struct tcp_comp_context *ctx = comp_get_ctx(sk); + struct sk_msg *msg = &ctx->tx.msg; + + sk_msg_init(msg); + + return sk_msg_alloc(sk, msg, len, 0); +} + +static int memcopy_from_iter(struct sock *sk, struct iov_iter *from, int copy) +{ + void *dest; + int rc; + + dest = tcp_comp_get_tx_stream(sk); + if (!dest) + return -ENOSPC; + + if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) + rc = copy_from_iter_nocache(dest, copy, from); + else + rc = copy_from_iter(dest, copy, from); + + if (rc != copy) + rc = -EFAULT; + + return rc; +} + +static int memcopy_to_msg(struct sock *sk, int bytes) +{ + struct tcp_comp_context *ctx = comp_get_ctx(sk); + struct sk_msg *msg = &ctx->tx.msg; + int i = msg->sg.curr; + struct scatterlist *sge; + u32 copy, buf_size; + void *from, *to; + + from = ctx->tx.compressed_data; + do { + sge = sk_msg_elem(msg, i); + /* This is possible if a trim operation shrunk the buffer */ + if (msg->sg.copybreak >= sge->length) { + msg->sg.copybreak = 0; + sk_msg_iter_var_next(i); + if (i == msg->sg.end) + break; + sge = sk_msg_elem(msg, i); + } + buf_size = sge->length - msg->sg.copybreak; + copy = (buf_size > bytes) ? bytes : buf_size; + to = sg_virt(sge) + msg->sg.copybreak; + msg->sg.copybreak += copy; + memcpy(to, from, copy); + bytes -= copy; + from += copy; + if (!bytes) + break; + msg->sg.copybreak = 0; + sk_msg_iter_var_next(i); + } while (i != msg->sg.end); + + msg->sg.curr = i; + return bytes; +} + +static int tcp_comp_compress_to_msg(struct sock *sk, int bytes) +{ + struct tcp_comp_context *ctx = comp_get_ctx(sk); + ZSTD_outBuffer outbuf; + ZSTD_inBuffer inbuf; + size_t ret; + + inbuf.src = ctx->tx.plaintext_data; + outbuf.dst = ctx->tx.compressed_data; + inbuf.size = bytes; + outbuf.size = TCP_COMP_MAX_CSIZE; + inbuf.pos = 0; + outbuf.pos = 0; + + ret = ZSTD_compressStream(ctx->tx.cstream, &outbuf, &inbuf); + if (ZSTD_isError(ret)) + return -EIO; + + ret = ZSTD_flushStream(ctx->tx.cstream, &outbuf); + if (ZSTD_isError(ret)) + return -EIO; + + if (inbuf.pos != inbuf.size) + return -EIO; + + if (memcopy_to_msg(sk, outbuf.pos)) + return -EIO; + + sk_msg_trim(sk, &ctx->tx.msg, outbuf.pos); + + return 0; +} + +static int tcp_comp_push_msg(struct sock *sk, struct sk_msg *msg, int flags) +{ + struct tcp_comp_context *ctx = comp_get_ctx(sk); + struct msghdr mh; + struct bio_vec bvec; + struct scatterlist *sg; + int ret, offset; + struct page *p; + size_t size; + + ctx->tx.in_tcp_sendpages = true; + while (1) { + sg = sk_msg_elem(msg, msg->sg.start); + offset = sg->offset; + size = sg->length; + p = sg_page(sg); +retry: + memset(&mh, 0, sizeof(struct msghdr)); + memset(&bvec, 0, sizeof(struct bio_vec)); + + mh.msg_flags = flags | MSG_SPLICE_PAGES; + bvec_set_page(&bvec, p, size, offset); + iov_iter_bvec(&mh.msg_iter, ITER_SOURCE, &bvec, 1, size); + + ret = tcp_sendmsg_locked(sk, &mh, size); + if (ret != size) { + if (ret > 0) { + sk_mem_uncharge(sk, ret); + sg->offset += ret; + sg->length -= ret; + size -= ret; + offset += ret; + goto retry; + } + ctx->tx.in_tcp_sendpages = false; + return ret; + } + + sk_mem_uncharge(sk, ret); + msg->sg.size -= size; + put_page(p); + sk_msg_iter_next(msg, start); + if (msg->sg.start == msg->sg.end) + break; + } + + clear_bit(TCP_COMP_SEND_PENDING, &ctx->flags); + ctx->tx.in_tcp_sendpages = false; + + return 0; +} + +static int tcp_comp_push(struct sock *sk, int bytes, int flags) +{ + struct tcp_comp_context *ctx = comp_get_ctx(sk); + int ret; + + ret = tcp_comp_compress_to_msg(sk, bytes); + if (ret < 0) { + pr_debug("%s: failed to compress sg\n", __func__); + return ret; + } + + set_bit(TCP_COMP_SEND_PENDING, &ctx->flags); + + ret = tcp_comp_push_msg(sk, &ctx->tx.msg, flags); + if (ret) { + pr_debug("%s: failed to tcp_comp_push_sg\n", __func__); + return ret; + } + + return 0; +} + +static int wait_on_pending_writer(struct sock *sk, long *timeo) +{ + DEFINE_WAIT_FUNC(wait, woken_wake_function); + int ret = 0; + + add_wait_queue(sk_sleep(sk), &wait); + while (1) { + if (!*timeo) { + ret = -EAGAIN; + break; + } + + if (signal_pending(current)) { + ret = sock_intr_errno(*timeo); + break; + } + + if (sk_wait_event(sk, timeo, !sk->sk_write_pending, &wait)) + break; + } + remove_wait_queue(sk_sleep(sk), &wait); + + return ret; +} + +static int tcp_comp_push_pending_msg(struct sock *sk, int flags) +{ + struct tcp_comp_context *ctx = comp_get_ctx(sk); + struct sk_msg *msg = &ctx->tx.msg; + + if (msg->sg.start == msg->sg.end) + return 0; + + return tcp_comp_push_msg(sk, msg, flags); +} + +static int tcp_comp_complete_pending_work(struct sock *sk, int flags, + long *timeo) +{ + struct tcp_comp_context *ctx = comp_get_ctx(sk); + int ret = 0; + + if (unlikely(sk->sk_write_pending)) + ret = wait_on_pending_writer(sk, timeo); + + if (!ret && tcp_comp_is_write_pending(ctx)) + ret = tcp_comp_push_pending_msg(sk, flags); + + return ret; +} + +static int tcp_comp_sendmsg(struct sock *sk, struct msghdr *msg, size_t size) +{ + struct tcp_comp_context *ctx = comp_get_ctx(sk); + int copied = 0, err = 0; + size_t try_to_copy; + int required_size; + long timeo; + + lock_sock(sk); + + timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT); + + err = tcp_comp_complete_pending_work(sk, msg->msg_flags, &timeo); + if (err) + goto out_err; + + while (msg_data_left(msg)) { + if (sk->sk_err) { + err = -sk->sk_err; + goto out_err; + } + + try_to_copy = msg_data_left(msg); + if (try_to_copy > TCP_COMP_SCRATCH_SIZE) + try_to_copy = TCP_COMP_SCRATCH_SIZE; + required_size = try_to_copy + TCP_COMP_MAX_PADDING; + + if (!sk_stream_memory_free(sk)) + goto wait_for_sndbuf; + +alloc_compressed: + err = alloc_compressed_msg(sk, required_size); + if (err) { + if (err != -ENOSPC) + goto wait_for_memory; + goto out_err; + } + + err = memcopy_from_iter(sk, &msg->msg_iter, try_to_copy); + if (err < 0) + goto out_err; + + copied += try_to_copy; + + err = tcp_comp_push(sk, try_to_copy, msg->msg_flags); + if (err < 0) { + if (err == -ENOMEM) + goto wait_for_memory; + if (err != -EAGAIN) + tcp_comp_err_abort(sk, EBADMSG); + goto out_err; + } + + continue; +wait_for_sndbuf: + set_bit(SOCK_NOSPACE, &sk->sk_socket->flags); +wait_for_memory: + err = sk_stream_wait_memory(sk, &timeo); + if (err) + goto out_err; + if (ctx->tx.msg.sg.size < required_size) + goto alloc_compressed; + } + +out_err: + err = sk_stream_error(sk, msg->msg_flags, err); + + release_sock(sk); + + return copied ? copied : err; +} + +static struct sk_buff *comp_wait_data(struct sock *sk, int flags, + long timeo, int *err) +{ + struct tcp_comp_context *ctx = comp_get_ctx(sk); + struct sk_buff *skb; + DEFINE_WAIT_FUNC(wait, woken_wake_function); + + while (!(skb = ctx->rx.pkt)) { + if (sk->sk_err) { + *err = sock_error(sk); + return NULL; + } + + if (!skb_queue_empty(&sk->sk_receive_queue)) { + __strp_unpause(&ctx->rx.strp); + if (ctx->rx.pkt) + return ctx->rx.pkt; + } + + if (sk->sk_shutdown & RCV_SHUTDOWN) + return NULL; + + if (sock_flag(sk, SOCK_DONE)) + return NULL; + + if ((flags & MSG_DONTWAIT) || !timeo) { + *err = -EAGAIN; + return NULL; + } + + add_wait_queue(sk_sleep(sk), &wait); + sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk); + sk_wait_event(sk, &timeo, ctx->rx.pkt != skb, &wait); + sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk); + remove_wait_queue(sk_sleep(sk), &wait); + + /* Handle signals */ + if (signal_pending(current)) { + *err = sock_intr_errno(timeo); + return NULL; + } + } + + return skb; +} + +static bool comp_advance_skb(struct sock *sk, struct sk_buff *skb, + unsigned int len) +{ + struct tcp_comp_context *ctx = comp_get_ctx(sk); + struct strp_msg *rxm = strp_msg(skb); + + if (len < rxm->full_len) { + rxm->offset += len; + rxm->full_len -= len; + return false; + } + + /* Finished with message */ + ctx->rx.pkt = NULL; + kfree_skb(skb); + __strp_unpause(&ctx->rx.strp); + + return true; +} + +static bool comp_advance_dskb(struct sock *sk, struct sk_buff *skb, + unsigned int len) +{ + struct tcp_comp_context *ctx = comp_get_ctx(sk); + struct strp_msg *rxm = strp_msg(skb); + + if (len < rxm->full_len) { + rxm->offset += len; + rxm->full_len -= len; + return false; + } + + /* Finished with message */ + ctx->rx.dpkt = NULL; + kfree_skb(skb); + return true; +} + +static int tcp_comp_rx_context_init(struct tcp_comp_context *ctx) +{ + int dsize; + + dsize = zstd_dstream_workspace_bound(TCP_COMP_MAX_INPUT); + if (dsize <= 0) + return -EINVAL; + + ctx->rx.dworkspace = kmalloc(dsize, GFP_KERNEL); + if (!ctx->rx.dworkspace) + return -ENOMEM; + + ctx->rx.dstream = zstd_init_dstream(TCP_COMP_MAX_INPUT, + ctx->rx.dworkspace, dsize); + if (!ctx->rx.dstream) + goto err_dstream; + + ctx->rx.plaintext_data = kvmalloc(TCP_COMP_MAX_CSIZE * 32, GFP_KERNEL); + if (!ctx->rx.plaintext_data) + goto err_dstream; + + return 0; + +err_dstream: + kfree(ctx->rx.dworkspace); + ctx->rx.dworkspace = NULL; + + return -ENOMEM; +} + +static void *tcp_comp_get_rx_stream(struct sock *sk) +{ + struct tcp_comp_context *ctx = comp_get_ctx(sk); + + if (!ctx->rx.plaintext_data) + tcp_comp_rx_context_init(ctx); + + return ctx->rx.plaintext_data; +} + +static int tcp_comp_decompress(struct sock *sk, struct sk_buff *skb, int flags) +{ + struct tcp_comp_context *ctx = comp_get_ctx(sk); + struct strp_msg *rxm = strp_msg(skb); + size_t ret, compressed_len = 0; + int nr_frags_over = 0; + ZSTD_outBuffer outbuf; + ZSTD_inBuffer inbuf; + struct sk_buff *nskb; + int len, plen; + void *to; + + to = tcp_comp_get_rx_stream(sk); + if (!to) + return -ENOSPC; + + if (skb_linearize_cow(skb)) + return -ENOMEM; + + nskb = skb_copy(skb, GFP_KERNEL); + if (!nskb) + return -ENOMEM; + + while (compressed_len < (skb->len - rxm->offset)) { + if (skb_shinfo(nskb)->nr_frags >= MAX_SKB_FRAGS) + break; + + len = 0; + plen = skb->len - rxm->offset - compressed_len; + if (plen > TCP_COMP_MAX_CSIZE) + plen = TCP_COMP_MAX_CSIZE; + + inbuf.src = (char *)skb->data + rxm->offset + compressed_len; + inbuf.pos = 0; + inbuf.size = plen; + + outbuf.dst = ctx->rx.plaintext_data; + outbuf.pos = 0; + outbuf.size = MAX_SKB_FRAGS * TCP_COMP_DATA_SIZE; + outbuf.size -= skb_shinfo(nskb)->nr_frags * TCP_COMP_DATA_SIZE; + + ret = ZSTD_decompressStream(ctx->rx.dstream, &outbuf, &inbuf); + if (ZSTD_isError(ret)) { + kfree_skb(nskb); + return -EIO; + } + + if (!compressed_len) { + len = outbuf.pos - skb->len; + if (len > skb_tailroom(nskb)) + len = skb_tailroom(nskb); + + __skb_put(nskb, len); + + len += skb->len; + skb_copy_to_linear_data(nskb, to, len); + } + + while ((to += len, outbuf.pos -= len) > 0) { + struct page *pages; + skb_frag_t *frag; + + if (skb_shinfo(nskb)->nr_frags >= MAX_SKB_FRAGS) { + nr_frags_over = 1; + break; + } + + frag = skb_shinfo(nskb)->frags + + skb_shinfo(nskb)->nr_frags; + pages = alloc_pages(__GFP_NOWARN | GFP_KERNEL | __GFP_COMP, + TCP_COMP_ALLOC_ORDER); + if (!pages) { + kfree_skb(nskb); + return -ENOMEM; + } + + frag->bv_page = pages; + len = PAGE_SIZE << TCP_COMP_ALLOC_ORDER; + if (outbuf.pos < len) + len = outbuf.pos; + + frag->bv_offset = 0; + skb_frag_size_set(frag, len); + memcpy(skb_frag_address(frag), to, len); + + nskb->truesize += len; + nskb->data_len += len; + nskb->len += len; + skb_shinfo(nskb)->nr_frags++; + } + + if (nr_frags_over) + break; + + compressed_len += inbuf.pos; + } + + ctx->rx.dpkt = nskb; + rxm = strp_msg(nskb); + rxm->full_len = nskb->len; + rxm->offset = 0; + comp_advance_skb(sk, skb, compressed_len); + + return 0; +} + +static int tcp_comp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, + int flags, int *addr_len) +{ + struct tcp_comp_context *ctx = comp_get_ctx(sk); + struct strp_msg *rxm; + struct sk_buff *skb; + ssize_t copied = 0; + int target, err = 0; + long timeo; + + if (unlikely(flags & MSG_ERRQUEUE)) + return sock_recv_errqueue(sk, msg, len, SOL_IP, IP_RECVERR); + + lock_sock(sk); + + target = sock_rcvlowat(sk, flags & MSG_WAITALL, len); + timeo = sock_rcvtimeo(sk, flags & MSG_WAITALL); + + do { + int chunk = 0; + + if (!ctx->rx.dpkt) { + skb = comp_wait_data(sk, flags, timeo, &err); + if (!skb) + goto recv_end; + + err = tcp_comp_decompress(sk, skb, flags); + if (err < 0) { + goto recv_end; + } + } + + skb = ctx->rx.dpkt; + rxm = strp_msg(skb); + chunk = min_t(unsigned int, rxm->full_len, len); + err = skb_copy_datagram_msg(skb, rxm->offset, msg, + chunk); + if (err < 0) + goto recv_end; + + copied += chunk; + len -= chunk; + if (likely(!(flags & MSG_PEEK))) + comp_advance_dskb(sk, skb, chunk); + else + break; + + if (copied >= target && !ctx->rx.dpkt) + break; + } while (len > 0); + +recv_end: + release_sock(sk); + return copied ? : err; +} + +bool comp_stream_read(struct sock *sk) +{ + struct tcp_comp_context *ctx = comp_get_ctx(sk); + + if (!ctx) + return false; + + if (ctx->rx.pkt || ctx->rx.dpkt) + return true; + + return false; +} + +static void comp_data_ready(struct sock *sk) +{ + struct tcp_comp_context *ctx = comp_get_ctx(sk); + + strp_data_ready(&ctx->rx.strp); +} + +static void comp_queue(struct strparser *strp, struct sk_buff *skb) +{ + struct tcp_comp_context *ctx = comp_get_ctx(strp->sk); + + ctx->rx.pkt = skb; + strp_pause(strp); + ctx->rx.saved_data_ready(strp->sk); +} + +static int comp_read_size(struct strparser *strp, struct sk_buff *skb) +{ + struct strp_msg *rxm = strp_msg(skb); + + if (rxm->offset > skb->len) + return 0; + + return skb->len - rxm->offset; +} + +void comp_setup_strp(struct sock *sk, struct tcp_comp_context *ctx) +{ + struct strp_callbacks cb; + + memset(&cb, 0, sizeof(cb)); + cb.rcv_msg = comp_queue; + cb.parse_msg = comp_read_size; + strp_init(&ctx->rx.strp, sk, &cb); + + write_lock_bh(&sk->sk_callback_lock); + ctx->rx.saved_data_ready = sk->sk_data_ready; + sk->sk_data_ready = comp_data_ready; + write_unlock_bh(&sk->sk_callback_lock); + + strp_check_rcv(&ctx->rx.strp); +} + +static void tcp_comp_write_space(struct sock *sk) +{ + struct tcp_comp_context *ctx = comp_get_ctx(sk); + + if (ctx->tx.in_tcp_sendpages) { + ctx->sk_write_space(sk); + return; + } + + if (!sk->sk_write_pending && tcp_comp_is_write_pending(ctx)) { + gfp_t sk_allocation = sk->sk_allocation; + int rc; + + sk->sk_allocation = GFP_ATOMIC; + rc = tcp_comp_push_pending_msg(sk, MSG_DONTWAIT | MSG_NOSIGNAL); + sk->sk_allocation = sk_allocation; + + if (rc < 0) + return; + } + + ctx->sk_write_space(sk); +} + +void tcp_init_compression(struct sock *sk) +{ + struct inet_connection_sock *icsk = inet_csk(sk); + struct tcp_comp_context *ctx = NULL; + struct sk_msg *msg = NULL; + struct tcp_sock *tp = tcp_sk(sk); + + if (!tp->rx_opt.comp_ok) + return; + + ctx = kzalloc(sizeof(*ctx), GFP_ATOMIC); + if (!ctx) + return; + + msg = &ctx->tx.msg; + sk_msg_init(msg); + + ctx->sk_write_space = sk->sk_write_space; + ctx->sk_proto = sk->sk_prot; + WRITE_ONCE(sk->sk_prot, &tcp_prot_override); + sk->sk_write_space = tcp_comp_write_space; + + rcu_assign_pointer(icsk->icsk_ulp_data, ctx); + + sock_set_flag(sk, SOCK_COMP); + comp_setup_strp(sk, ctx); +} + +static void tcp_comp_context_tx_free(struct tcp_comp_context *ctx) +{ + kfree(ctx->tx.cworkspace); + ctx->tx.cworkspace = NULL; + + kvfree(ctx->tx.plaintext_data); + ctx->tx.plaintext_data = NULL; + + kvfree(ctx->tx.compressed_data); + ctx->tx.compressed_data = NULL; +} + +static void tcp_comp_context_rx_free(struct tcp_comp_context *ctx) +{ + kfree(ctx->rx.dworkspace); + ctx->rx.dworkspace = NULL; + + kvfree(ctx->rx.plaintext_data); + ctx->rx.plaintext_data = NULL; +} + +static void tcp_comp_context_free(struct rcu_head *head) +{ + struct tcp_comp_context *ctx; + + ctx = container_of(head, struct tcp_comp_context, rcu); + + tcp_comp_context_tx_free(ctx); + tcp_comp_context_rx_free(ctx); + strp_done(&ctx->rx.strp); + kfree(ctx); +} + +void tcp_cleanup_compression(struct sock *sk) +{ + struct inet_connection_sock *icsk = inet_csk(sk); + struct tcp_comp_context *ctx = comp_get_ctx(sk); + + if (!ctx || !sock_flag(sk, SOCK_COMP)) + return; + + if (ctx->rx.pkt) { + kfree_skb(ctx->rx.pkt); + ctx->rx.pkt = NULL; + } + + if (ctx->rx.dpkt) { + kfree_skb(ctx->rx.dpkt); + ctx->rx.dpkt = NULL; + } + strp_stop(&ctx->rx.strp); + + rcu_assign_pointer(icsk->icsk_ulp_data, NULL); + call_rcu(&ctx->rcu, tcp_comp_context_free); +} + +int tcp_comp_init(void) +{ + tcp_prot_override = tcp_prot; + tcp_prot_override.sendmsg = tcp_comp_sendmsg; + tcp_prot_override.recvmsg = tcp_comp_recvmsg; + tcp_prot_override.sock_is_readable = comp_stream_read; + + return 0; +} diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index fd5c13c1fbc8..a5adf4822663 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -4012,6 +4012,24 @@ static bool smc_parse_options(const struct tcphdr *th, return false; } +static bool tcp_parse_comp_option(const struct tcphdr *th, + struct tcp_options_received *opt_rx, + const unsigned char *ptr, + int opsize) +{ +#if IS_ENABLED(CONFIG_TCP_COMP) + if (static_branch_unlikely(&tcp_have_comp)) { + if (th->syn && !(opsize & 1) && + opsize >= TCPOLEN_EXP_COMP_BASE && + get_unaligned_be16(ptr) == TCPOPT_COMP_MAGIC) { + opt_rx->comp_ok = 1; + return true; + } + } +#endif + return false; +} + /* Try to parse the MSS option from the TCP header. Return 0 on failure, clamped * value on success. */ @@ -4171,6 +4189,10 @@ void tcp_parse_options(const struct net *net, if (smc_parse_options(th, opt_rx, ptr, opsize)) break; + if (tcp_parse_comp_option(th, opt_rx, ptr, + opsize)) + break; + opt_rx->saw_unknown = 1; break; @@ -6097,6 +6119,7 @@ void tcp_init_transfer(struct sock *sk, int bpf_op, struct sk_buff *skb) /* Initialize congestion control unless BPF initialized it already: */ if (!icsk->icsk_ca_initialized) tcp_init_congestion_control(sk); + tcp_init_compression(sk); tcp_init_buffer_space(sk); } @@ -6835,6 +6858,9 @@ static void tcp_openreq_init(struct request_sock *req, ireq->smc_ok = rx_opt->smc_ok && !(tcp_sk(sk)->smc_hs_congested && tcp_sk(sk)->smc_hs_congested(sk)); #endif +#if IS_ENABLED(CONFIG_TCP_COMP) + ireq->comp_ok = rx_opt->comp_ok; +#endif } struct request_sock *inet_reqsk_alloc(const struct request_sock_ops *ops, diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c index 4167e8a48b60..9cbf23c1c039 100644 --- a/net/ipv4/tcp_ipv4.c +++ b/net/ipv4/tcp_ipv4.c @@ -2304,6 +2304,8 @@ void tcp_v4_destroy_sock(struct sock *sk) tcp_cleanup_congestion_control(sk); + tcp_cleanup_compression(sk); + tcp_cleanup_ulp(sk); /* Cleanup up the write buffer. */ diff --git a/net/ipv4/tcp_minisocks.c b/net/ipv4/tcp_minisocks.c index b98d476f1594..d67a84c114ce 100644 --- a/net/ipv4/tcp_minisocks.c +++ b/net/ipv4/tcp_minisocks.c @@ -543,6 +543,9 @@ struct sock *tcp_create_openreq_child(const struct sock *sk, newtp->rcv_ssthresh = req->rsk_rcv_wnd; newtp->rcv_wnd = req->rsk_rcv_wnd; newtp->rx_opt.wscale_ok = ireq->wscale_ok; +#if IS_ENABLED(CONFIG_TCP_COMP) + newtp->rx_opt.comp_ok = ireq->comp_ok; +#endif if (newtp->rx_opt.wscale_ok) { newtp->rx_opt.snd_wscale = ireq->snd_wscale; newtp->rx_opt.rcv_wscale = ireq->rcv_wscale; diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c index 1917c62ad3bf..eee34f6c7643 100644 --- a/net/ipv4/tcp_output.c +++ b/net/ipv4/tcp_output.c @@ -422,6 +422,7 @@ static inline bool tcp_urg_mode(const struct tcp_sock *tp) #define OPTION_FAST_OPEN_COOKIE BIT(8) #define OPTION_SMC BIT(9) #define OPTION_MPTCP BIT(10) +#define OPTION_COMP BIT(11) static void smc_options_write(__be32 *ptr, u16 *options) { @@ -438,6 +439,19 @@ static void smc_options_write(__be32 *ptr, u16 *options) #endif } +static void comp_options_write(__be32 *ptr, u16 *options) +{ +#if IS_ENABLED(CONFIG_TCP_COMP) + if (static_branch_unlikely(&tcp_have_comp)) { + if (unlikely(OPTION_COMP & *options)) { + *ptr++ = htonl((TCPOPT_EXP << 24) | + (TCPOLEN_EXP_COMP_BASE << 16) | + (TCPOPT_COMP_MAGIC)); + } + } +#endif +} + struct tcp_out_options { u16 options; /* bit field of OPTION_* */ u16 mss; /* 0 to disable */ @@ -711,6 +725,8 @@ static void tcp_options_write(struct tcphdr *th, struct tcp_sock *tp, smc_options_write(ptr, &options); mptcp_options_write(th, ptr, tp, opts); + + comp_options_write(ptr, &options); } static void smc_set_option(const struct tcp_sock *tp, @@ -729,6 +745,39 @@ static void smc_set_option(const struct tcp_sock *tp, #endif } +static void comp_set_option(struct sock *sk, + struct tcp_out_options *opts, + unsigned int *remaining) +{ +#if IS_ENABLED(CONFIG_TCP_COMP) + if (static_branch_unlikely(&tcp_have_comp)) { + if (tcp_syn_comp_enabled(sk)) { + if (*remaining >= TCPOLEN_EXP_COMP_BASE) { + opts->options |= OPTION_COMP; + *remaining -= TCPOLEN_EXP_COMP_BASE; + } + } + } +#endif +} + +static void comp_set_option_cond(struct sock *sk, + const struct inet_request_sock *ireq, + struct tcp_out_options *opts, + unsigned int *remaining) +{ +#if IS_ENABLED(CONFIG_TCP_COMP) + if (static_branch_unlikely(&tcp_have_comp)) { + if (tcp_synack_comp_enabled(sk, ireq)) { + if (*remaining >= TCPOLEN_EXP_COMP_BASE) { + opts->options |= OPTION_COMP; + *remaining -= TCPOLEN_EXP_COMP_BASE; + } + } + } +#endif +} + static void smc_set_option_cond(const struct tcp_sock *tp, const struct inet_request_sock *ireq, struct tcp_out_options *opts, @@ -830,6 +879,7 @@ static unsigned int tcp_syn_options(struct sock *sk, struct sk_buff *skb, } smc_set_option(tp, opts, &remaining); + comp_set_option(sk, opts, &remaining); if (sk_is_mptcp(sk)) { unsigned int size; @@ -910,6 +960,8 @@ static unsigned int tcp_synack_options(const struct sock *sk, smc_set_option_cond(tcp_sk(sk), ireq, opts, &remaining); + comp_set_option_cond((struct sock *)sk, ireq, opts, &remaining); + bpf_skops_hdr_opt_len((struct sock *)sk, skb, req, syn_skb, synack_type, opts, &remaining); diff --git a/net/ipv6/syncookies.c b/net/ipv6/syncookies.c index 8698b49dfc8d..75652826abbd 100644 --- a/net/ipv6/syncookies.c +++ b/net/ipv6/syncookies.c @@ -218,6 +218,10 @@ struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb) if (IS_ENABLED(CONFIG_SMC)) ireq->smc_ok = 0; +#if IS_ENABLED(CONFIG_TCP_COMP) + ireq->comp_ok = 0; +#endif + /* * We need to lookup the dst_entry to get the correct window size. * This is taken from tcp_v6_syn_recv_sock. Somebody please enlighten -- 2.34.1

2 1

[PATCH OLK-6.6 0/2] To offer an option for turn off spectre-v2
by Yi Yang 29 Dec '23

29 Dec '23

option for turn off spectre-v2 Cui GaoSheng (1): ARM: use ldr_l to replace ldr instruction for the symbol jump GONG, Ruiqi (1): ARM: spectre-v2: turn off the mitigation via boot cmdline param arch/arm/include/asm/system_misc.h | 1 + arch/arm/mm/proc-v7-bugs.c | 18 ++++++++++++++++++ arch/arm/mm/proc-v7.S | 9 +++++++++ 3 files changed, 28 insertions(+) -- 2.25.1

2 3

[PATCH OLK-6.6 0/1] arm64: Add non nmi ipi backtrace support
by Liao Chen 29 Dec '23

29 Dec '23

From: linruizhe <linruizhe(a)huawei.com> Dependencies: CONFIG_DETECT_HUNG_TASK=y CONFIG_WATCHDOG=y Li Zhengyu (1): arm64: Add non nmi ipi backtrace support arch/arm64/kernel/ipi_nmi.c | 18 +++++++++++++++--- 1 file changed, 15 insertions(+), 3 deletions(-) -- 2.25.1

2 2

[PATCH OLK-6.6 0/4] cgroup ns kabi
by Chen Ridong 29 Dec '23

29 Dec '23

Chen Ridong (4): kabi: reserve space for cgroup frame namespace:kabi:reserve for future namespace development blk-cgroup:kabi:reserve for future blk-cgroup develemenpt cgroup/misc: kabi reserve for future misc development block/blk-cgroup.h | 11 +++++++++++ include/linux/cgroup-defs.h | 32 ++++++++++++++++++++++++++++++++ include/linux/cgroup_rdma.h | 3 +++ include/linux/misc_cgroup.h | 11 +++++++++++ include/linux/nsproxy.h | 9 +++++++++ include/linux/pid_namespace.h | 2 ++ include/linux/user_namespace.h | 33 +++++++++++++++++++++++++++++++++ kernel/ucount.c | 18 ++++++++++++++++++ 8 files changed, 119 insertions(+) -- 2.34.1

2 5

[PATCH OLK-6.6] timer_list: avoid other cpu soft lockup when printing timer list
by Yang Yingliang 29 Dec '23

29 Dec '23

hulk inclusion category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I8SDKX --------------------------- If system has many cpus (e.g. 128), it will spend a lot of time to print message to the console when execute echo q > /proc/sysrq-trigger. When /proc/sys/kernel/numa_balancing is enabled, if the migration threads is woke up, the thread cannot continue until the print finish, it will trigger a soft lockup. PID: 619 TASK: ffffa02fdd8bec80 CPU: 121 COMMAND: "migration/121" #0 [ffff00000a103b10] __crash_kexec at ffff0000081bf200 #1 [ffff00000a103ca0] panic at ffff0000080ec93c #2 [ffff00000a103d80] watchdog_timer_fn at ffff0000081f8a14 #3 [ffff00000a103e00] __run_hrtimer at ffff00000819701c #4 [ffff00000a103e40] __hrtimer_run_queues at ffff000008197420 #5 [ffff00000a103ea0] hrtimer_interrupt at ffff00000819831c #6 [ffff00000a103f10] arch_timer_dying_cpu at ffff000008b53144 #7 [ffff00000a103f30] handle_percpu_devid_irq at ffff000008174e34 #8 [ffff00000a103f70] generic_handle_irq at ffff00000816c5e8 #9 [ffff00000a103f90] __handle_domain_irq at ffff00000816d1f4 #10 [ffff00000a103fd0] gic_handle_irq at ffff000008081860 --- <IRQ stack> --- #11 [ffff00000d6e3d50] el1_irq at ffff0000080834c8 #12 [ffff00000d6e3d60] multi_cpu_stop at ffff0000081d9964 #13 [ffff00000d6e3db0] cpu_stopper_thread at ffff0000081d9cfc #14 [ffff00000d6e3e10] smpboot_thread_fn at ffff00000811e0a8 #15 [ffff00000d6e3e70] kthread at ffff000008118988 To avoid this soft lockup, add touch_all_softlockup_watchdogs() in sysrq_timer_list_show() Signed-off-by: Yang Yingliang <yangyingliang(a)huawei.com> --- kernel/time/timer_list.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/kernel/time/timer_list.c b/kernel/time/timer_list.c index ed7d6ad694fb..3da0a2585ab4 100644 --- a/kernel/time/timer_list.c +++ b/kernel/time/timer_list.c @@ -269,13 +269,17 @@ void sysrq_timer_list_show(void) timer_list_header(NULL, now); - for_each_online_cpu(cpu) + for_each_online_cpu(cpu) { + touch_all_softlockup_watchdogs(); print_cpu(NULL, cpu, now); + } #ifdef CONFIG_GENERIC_CLOCKEVENTS timer_list_show_tickdevices_header(NULL); - for_each_online_cpu(cpu) + for_each_online_cpu(cpu) { + touch_all_softlockup_watchdogs(); print_tickdevice(NULL, tick_get_device(cpu), cpu); + } #endif return; } -- 2.25.1

2 1

[PATCH OLK-6.6] mm: Add sysctl to clear free list pages
by Yu Liao 29 Dec '23

29 Dec '23

hulk inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I8MK3O CVE: NA -------------------------------- This patch add sysctl to clear pages in free lists of each NUMA node. For each NUMA node, clear each page in the free list, these work is scheduled on a random CPU of the NUMA node. When kasan is enabled and the pages are free, the shadow memory will be filled with 0xFF, writing these free pages will cause UAF, so just disable KASAN for clear freelist. In the case of large memory, the clear freelist will hold zone lock for a long time. As a result, the process may be blocked unless clear freelist thread exit, and causing the system to be reset by the watchdog. Provide a mechanism to stop clear freelist threads when elapsed time exceeds cfp_timeout, which can be set by module_param(). Signed-off-by: Yu Liao <liaoyu15(a)huawei.com> --- .../admin-guide/kernel-parameters.txt | 3 + Documentation/admin-guide/sysctl/vm.rst | 13 ++ mm/Kconfig | 12 ++ mm/Makefile | 2 + mm/clear_freelist_page.c | 178 ++++++++++++++++++ 5 files changed, 208 insertions(+) create mode 100644 mm/clear_freelist_page.c diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt index fe2dc03538fe..14dda1178c34 100644 --- a/Documentation/admin-guide/kernel-parameters.txt +++ b/Documentation/admin-guide/kernel-parameters.txt @@ -620,6 +620,9 @@ Also note the kernel might malfunction if you disable some critical bits. + clear_freelist + Enable clear_freelist feature. + clk_ignore_unused [CLK] Prevents the clock framework from automatically gating diff --git a/Documentation/admin-guide/sysctl/vm.rst b/Documentation/admin-guide/sysctl/vm.rst index 2025b524028c..7c411b1dc3a9 100644 --- a/Documentation/admin-guide/sysctl/vm.rst +++ b/Documentation/admin-guide/sysctl/vm.rst @@ -25,6 +25,7 @@ files can be found in mm/swap.c. Currently, these files are in /proc/sys/vm: - admin_reserve_kbytes +- clear_freelist_pages - compact_memory - compaction_proactiveness - compact_unevictable_allowed @@ -110,6 +111,18 @@ On x86_64 this is about 128MB. Changing this takes effect whenever an application requests memory. +clear_freelist_pages +==================== + +Available only when CONFIG_CLEAR_FREELIST_PAGE is set. When 1 is written to the +file, all pages in free lists will be written with 0. + +Zone lock is held during clear_freelist_pages, if the execution time is too +long, RCU CPU Stall warnings will be print. For each NUMA node, +clear_freelist_pages is performed on a "random" CPU of the NUMA node. +The time consuming is related to the hardware. + + compact_memory ============== diff --git a/mm/Kconfig b/mm/Kconfig index df87fab46621..be9104ace232 100644 --- a/mm/Kconfig +++ b/mm/Kconfig @@ -1336,6 +1336,18 @@ config PAGE_CACHE_LIMIT if unsure, say N to disable the PAGE_CACHE_LIMIT. +config CLEAR_FREELIST_PAGE + bool "Support for clear free list pages" + depends on MMU && SYSCTL + default n + help + Say y here to enable the clear free list pages feature. When + writing to clear_freelist, trigger to clean up the free memory + of the buddy system. + + To enable this feature, kernel parameter "clear_freelist" also + needs to be added. + source "mm/damon/Kconfig" endmenu diff --git a/mm/Makefile b/mm/Makefile index 07cf74abd241..6759053ed782 100644 --- a/mm/Makefile +++ b/mm/Makefile @@ -7,6 +7,7 @@ KASAN_SANITIZE_slab_common.o := n KASAN_SANITIZE_slab.o := n KASAN_SANITIZE_slub.o := n KCSAN_SANITIZE_kmemleak.o := n +KASAN_SANITIZE_clear_freelist_page.o := n # These produce frequent data race reports: most of them are due to races on # the same word but accesses to different bits of that word. Re-enable KCSAN @@ -142,3 +143,4 @@ obj-$(CONFIG_SHRINKER_DEBUG) += shrinker_debug.o obj-$(CONFIG_SHARE_POOL) += share_pool.o obj-$(CONFIG_MEMCG_MEMFS_INFO) += memcg_memfs_info.o obj-$(CONFIG_PAGE_CACHE_LIMIT) += page_cache_limit.o +obj-$(CONFIG_CLEAR_FREELIST_PAGE) += clear_freelist_page.o diff --git a/mm/clear_freelist_page.c b/mm/clear_freelist_page.c new file mode 100644 index 000000000000..96c8b464b8aa --- /dev/null +++ b/mm/clear_freelist_page.c @@ -0,0 +1,178 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Support for clear free list pages. + */ + +#include <linux/mmzone.h> +#include <linux/mm_types.h> +#include <linux/mm.h> +#include <linux/sysctl.h> +#include <linux/highmem.h> +#include <linux/slab.h> +#include <linux/workqueue.h> +#include <linux/sched.h> +#include <linux/atomic.h> +#include <linux/nmi.h> +#include <linux/sched/clock.h> +#include <linux/module.h> + +#define CFP_DEFAULT_TIMEOUT 2000 +#define for_each_populated_zone_pgdat(pgdat, zone) \ + for (zone = pgdat->node_zones; \ + zone; \ + zone = next_pgdat_zone(zone)) \ + if (!populated_zone(zone)) \ + ; /* do nothing */ \ + else + +struct pgdat_entry { + struct pglist_data *pgdat; + struct work_struct work; +}; + +static DECLARE_WAIT_QUEUE_HEAD(clear_freelist_wait); +static DEFINE_MUTEX(clear_freelist_lock); +static atomic_t clear_freelist_workers; +static atomic_t clear_pages_num; +static ulong cfp_timeout_ms = CFP_DEFAULT_TIMEOUT; + +/* + * next_pgdat_zone - helper magic for for_each_populated_zone_pgdat() + */ +static struct zone *next_pgdat_zone(struct zone *zone) +{ + pg_data_t *pgdat = zone->zone_pgdat; + + if (zone < pgdat->node_zones + MAX_NR_ZONES - 1) + zone++; + else + zone = NULL; + return zone; +} + +static void clear_pgdat_freelist_pages(struct work_struct *work) +{ + struct pgdat_entry *entry = container_of(work, struct pgdat_entry, work); + u64 cfp_timeout_ns = cfp_timeout_ms * NSEC_PER_MSEC; + struct pglist_data *pgdat = entry->pgdat; + unsigned long flags, order, t; + struct page *page; + struct zone *zone; + u64 start, now; + + start = sched_clock(); + + for_each_populated_zone_pgdat(pgdat, zone) { + spin_lock_irqsave(&zone->lock, flags); + for_each_migratetype_order(order, t) { + list_for_each_entry(page, &zone->free_area[order].free_list[t], lru) { + now = sched_clock(); + if (unlikely(now - start > cfp_timeout_ns)) { + spin_unlock_irqrestore(&zone->lock, flags); + goto out; + } + +#ifdef CONFIG_KMAP_LOCAL + int i; + + /* Clear highmem by clear_highpage() */ + for (i = 0; i < (1 << order); i++) + clear_highpage(page + i); +#else + memset(page_address(page), 0, (1 << order) * PAGE_SIZE); +#endif + touch_nmi_watchdog(); + atomic_add(1 << order, &clear_pages_num); + } + } + spin_unlock_irqrestore(&zone->lock, flags); + + cond_resched(); + } + +out: + kfree(entry); + + if (atomic_dec_and_test(&clear_freelist_workers)) + wake_up(&clear_freelist_wait); +} + +static void init_clear_freelist_work(struct pglist_data *pgdat) +{ + struct pgdat_entry *entry; + + entry = kzalloc(sizeof(struct pgdat_entry), GFP_KERNEL); + if (!entry) + return; + + entry->pgdat = pgdat; + INIT_WORK(&entry->work, clear_pgdat_freelist_pages); + queue_work_node(pgdat->node_id, system_unbound_wq, &entry->work); +} + +static void clear_freelist_pages(void) +{ + struct pglist_data *pgdat; + + mutex_lock(&clear_freelist_lock); + drain_all_pages(NULL); + + for_each_online_pgdat(pgdat) { + atomic_inc(&clear_freelist_workers); + init_clear_freelist_work(pgdat); + } + + wait_event(clear_freelist_wait, atomic_read(&clear_freelist_workers) == 0); + + pr_debug("Cleared pages %d\nFree pages %lu\n", atomic_read(&clear_pages_num), + global_zone_page_state(NR_FREE_PAGES)); + atomic_set(&clear_pages_num, 0); + + mutex_unlock(&clear_freelist_lock); +} + +static int sysctl_clear_freelist_handler(struct ctl_table *table, int write, + void __user *buffer, size_t *lenp, loff_t *ppos) +{ + int ret; + int val; + + table->data = &val; + ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos); + + if (!ret && write) + clear_freelist_pages(); + + return ret; +} + +static struct ctl_table clear_freelist_table[] = { + { + .procname = "clear_freelist_pages", + .data = NULL, + .maxlen = sizeof(int), + .mode = 0200, + .proc_handler = &sysctl_clear_freelist_handler, + .extra1 = SYSCTL_ONE, + .extra2 = SYSCTL_ONE, + }, + { } +}; + +static bool clear_freelist_enabled; +static int __init setup_clear_freelist(char *str) +{ + clear_freelist_enabled = true; + return 1; +} +__setup("clear_freelist", setup_clear_freelist); + +static int __init clear_freelist_init(void) +{ + if (clear_freelist_enabled) + register_sysctl_init("vm", clear_freelist_table); + + return 0; +} +module_init(clear_freelist_init); +module_param(cfp_timeout_ms, ulong, 0644); -- 2.33.0

2 1