From 8f937c6099858eee15fae14009dcbd05177fa91d Mon Sep 17 00:00:00 2001 From: Harald Welte Date: Sun, 29 May 2005 20:23:46 -0700 Subject: [IPV4]: Primary and secondary addresses Add an option to make secondary IP addresses get promoted when primary IP addresses are removed from the device. It defaults to off to preserve existing behavior. Signed-off-by: Harald Welte Signed-off-by: David S. Miller --- include/linux/sysctl.h | 1 + 1 file changed, 1 insertion(+) (limited to 'include/linux/sysctl.h') diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h index 772998147e3e..23032d9d6071 100644 --- a/include/linux/sysctl.h +++ b/include/linux/sysctl.h @@ -399,6 +399,7 @@ enum NET_IPV4_CONF_FORCE_IGMP_VERSION=17, NET_IPV4_CONF_ARP_ANNOUNCE=18, NET_IPV4_CONF_ARP_IGNORE=19, + NET_IPV4_CONF_PROMOTE_SECONDARIES=20, __NET_IPV4_CONF_MAX }; -- cgit v1.2.3 From 1c2fb7f93cb20621772bf304f3dba0849942e5db Mon Sep 17 00:00:00 2001 From: "J. Simonetti" Date: Mon, 13 Jun 2005 15:19:03 -0700 Subject: [IPV4]: Sysctl configurable icmp error source address. This patch alows you to change the source address of icmp error messages. It applies cleanly to 2.6.11.11 and retains the default behaviour. In the old (default) behaviour icmp error messages are sent with the ip of the exiting interface. The new behaviour (when the sysctl variable is toggled on), it will send the message with the ip of the interface that received the packet that caused the icmp error. This is the behaviour network administrators will expect from a router. It makes debugging complicated network layouts much easier. Also, all 'vendor routers' I know of have the later behaviour. Signed-off-by: David S. Miller --- include/linux/sysctl.h | 1 + net/ipv4/icmp.c | 9 +++++++-- net/ipv4/sysctl_net_ipv4.c | 9 +++++++++ 3 files changed, 17 insertions(+), 2 deletions(-) (limited to 'include/linux/sysctl.h') diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h index 23032d9d6071..a17745c80a91 100644 --- a/include/linux/sysctl.h +++ b/include/linux/sysctl.h @@ -346,6 +346,7 @@ enum NET_TCP_MODERATE_RCVBUF=106, NET_TCP_TSO_WIN_DIVISOR=107, NET_TCP_BIC_BETA=108, + NET_IPV4_ICMP_ERRORS_USE_INBOUND_IFADDR=109, }; enum { diff --git a/net/ipv4/icmp.c b/net/ipv4/icmp.c index 85bf0d3e294b..cb759484979d 100644 --- a/net/ipv4/icmp.c +++ b/net/ipv4/icmp.c @@ -207,6 +207,7 @@ int sysctl_icmp_ignore_bogus_error_responses; int sysctl_icmp_ratelimit = 1 * HZ; int sysctl_icmp_ratemask = 0x1818; +int sysctl_icmp_errors_use_inbound_ifaddr; /* * ICMP control array. This specifies what to do with each ICMP. @@ -511,8 +512,12 @@ void icmp_send(struct sk_buff *skb_in, int type, int code, u32 info) */ saddr = iph->daddr; - if (!(rt->rt_flags & RTCF_LOCAL)) - saddr = 0; + if (!(rt->rt_flags & RTCF_LOCAL)) { + if (sysctl_icmp_errors_use_inbound_ifaddr) + saddr = inet_select_addr(skb_in->dev, 0, RT_SCOPE_LINK); + else + saddr = 0; + } tos = icmp_pointers[type].error ? ((iph->tos & IPTOS_TOS_MASK) | IPTOS_PREC_INTERNETCONTROL) : diff --git a/net/ipv4/sysctl_net_ipv4.c b/net/ipv4/sysctl_net_ipv4.c index 3aafb298c1c1..23068bddbf0b 100644 --- a/net/ipv4/sysctl_net_ipv4.c +++ b/net/ipv4/sysctl_net_ipv4.c @@ -23,6 +23,7 @@ extern int sysctl_ip_nonlocal_bind; extern int sysctl_icmp_echo_ignore_all; extern int sysctl_icmp_echo_ignore_broadcasts; extern int sysctl_icmp_ignore_bogus_error_responses; +extern int sysctl_icmp_errors_use_inbound_ifaddr; /* From ip_fragment.c */ extern int sysctl_ipfrag_low_thresh; @@ -395,6 +396,14 @@ ctl_table ipv4_table[] = { .mode = 0644, .proc_handler = &proc_dointvec }, + { + .ctl_name = NET_IPV4_ICMP_ERRORS_USE_INBOUND_IFADDR, + .procname = "icmp_errors_use_inbound_ifaddr", + .data = &sysctl_icmp_errors_use_inbound_ifaddr, + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = &proc_dointvec + }, { .ctl_name = NET_IPV4_ROUTE, .procname = "route", -- cgit v1.2.3 From d6e711448137ca3301512cec41a2c2ce852b3d0a Mon Sep 17 00:00:00 2001 From: Alan Cox Date: Thu, 23 Jun 2005 00:09:43 -0700 Subject: [PATCH] setuid core dump Add a new `suid_dumpable' sysctl: This value can be used to query and set the core dump mode for setuid or otherwise protected/tainted binaries. The modes are 0 - (default) - traditional behaviour. Any process which has changed privilege levels or is execute only will not be dumped 1 - (debug) - all processes dump core when possible. The core dump is owned by the current user and no security is applied. This is intended for system debugging situations only. Ptrace is unchecked. 2 - (suidsafe) - any binary which normally would not be dumped is dumped readable by root only. This allows the end user to remove such a dump but not access it directly. For security reasons core dumps in this mode will not overwrite one another or other files. This mode is appropriate when adminstrators are attempting to debug problems in a normal environment. (akpm: > > +EXPORT_SYMBOL(suid_dumpable); > > EXPORT_SYMBOL_GPL? No problem to me. > > if (current->euid == current->uid && current->egid == current->gid) > > current->mm->dumpable = 1; > > Should this be SUID_DUMP_USER? Actually the feedback I had from last time was that the SUID_ defines should go because its clearer to follow the numbers. They can go everywhere (and there are lots of places where dumpable is tested/used as a bool in untouched code) > Maybe this should be renamed to `dump_policy' or something. Doing that > would help us catch any code which isn't using the #defines, too. Fair comment. The patch was designed to be easy to maintain for Red Hat rather than for merging. Changing that field would create a gigantic diff because it is used all over the place. ) Signed-off-by: Alan Cox Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- Documentation/sysctl/kernel.txt | 20 ++++++++++++++++++++ fs/exec.c | 23 +++++++++++++++++++++-- fs/proc/base.c | 6 ++++-- include/linux/binfmts.h | 5 +++++ include/linux/sched.h | 2 +- include/linux/sysctl.h | 1 + kernel/sys.c | 22 +++++++++++----------- kernel/sysctl.c | 9 +++++++++ security/commoncap.c | 2 +- security/dummy.c | 2 +- 10 files changed, 74 insertions(+), 18 deletions(-) (limited to 'include/linux/sysctl.h') diff --git a/Documentation/sysctl/kernel.txt b/Documentation/sysctl/kernel.txt index 35159176997b..9f11d36a8c10 100644 --- a/Documentation/sysctl/kernel.txt +++ b/Documentation/sysctl/kernel.txt @@ -49,6 +49,7 @@ show up in /proc/sys/kernel: - shmmax [ sysv ipc ] - shmmni - stop-a [ SPARC only ] +- suid_dumpable - sysrq ==> Documentation/sysrq.txt - tainted - threads-max @@ -300,6 +301,25 @@ kernel. This value defaults to SHMMAX. ============================================================== +suid_dumpable: + +This value can be used to query and set the core dump mode for setuid +or otherwise protected/tainted binaries. The modes are + +0 - (default) - traditional behaviour. Any process which has changed + privilege levels or is execute only will not be dumped +1 - (debug) - all processes dump core when possible. The core dump is + owned by the current user and no security is applied. This is + intended for system debugging situations only. Ptrace is unchecked. +2 - (suidsafe) - any binary which normally would not be dumped is dumped + readable by root only. This allows the end user to remove + such a dump but not access it directly. For security reasons + core dumps in this mode will not overwrite one another or + other files. This mode is appropriate when adminstrators are + attempting to debug problems in a normal environment. + +============================================================== + tainted: Non-zero if the kernel has been tainted. Numeric values, which diff --git a/fs/exec.c b/fs/exec.c index 3a4b35a14c0d..48871917d363 100644 --- a/fs/exec.c +++ b/fs/exec.c @@ -58,6 +58,9 @@ int core_uses_pid; char core_pattern[65] = "core"; +int suid_dumpable = 0; + +EXPORT_SYMBOL(suid_dumpable); /* The maximal length of core_pattern is also specified in sysctl.c */ static struct linux_binfmt *formats; @@ -864,6 +867,9 @@ int flush_old_exec(struct linux_binprm * bprm) if (current->euid == current->uid && current->egid == current->gid) current->mm->dumpable = 1; + else + current->mm->dumpable = suid_dumpable; + name = bprm->filename; /* Copies the binary name from after last slash */ @@ -884,7 +890,7 @@ int flush_old_exec(struct linux_binprm * bprm) permission(bprm->file->f_dentry->d_inode,MAY_READ, NULL) || (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP)) { suid_keys(current); - current->mm->dumpable = 0; + current->mm->dumpable = suid_dumpable; } /* An exec changes our domain. We are no longer part of the thread @@ -1432,6 +1438,8 @@ int do_coredump(long signr, int exit_code, struct pt_regs * regs) struct inode * inode; struct file * file; int retval = 0; + int fsuid = current->fsuid; + int flag = 0; binfmt = current->binfmt; if (!binfmt || !binfmt->core_dump) @@ -1441,6 +1449,16 @@ int do_coredump(long signr, int exit_code, struct pt_regs * regs) up_write(&mm->mmap_sem); goto fail; } + + /* + * We cannot trust fsuid as being the "true" uid of the + * process nor do we know its entire history. We only know it + * was tainted so we dump it as root in mode 2. + */ + if (mm->dumpable == 2) { /* Setuid core dump mode */ + flag = O_EXCL; /* Stop rewrite attacks */ + current->fsuid = 0; /* Dump root private */ + } mm->dumpable = 0; init_completion(&mm->core_done); spin_lock_irq(¤t->sighand->siglock); @@ -1466,7 +1484,7 @@ int do_coredump(long signr, int exit_code, struct pt_regs * regs) lock_kernel(); format_corename(corename, core_pattern, signr); unlock_kernel(); - file = filp_open(corename, O_CREAT | 2 | O_NOFOLLOW | O_LARGEFILE, 0600); + file = filp_open(corename, O_CREAT | 2 | O_NOFOLLOW | O_LARGEFILE | flag, 0600); if (IS_ERR(file)) goto fail_unlock; inode = file->f_dentry->d_inode; @@ -1491,6 +1509,7 @@ int do_coredump(long signr, int exit_code, struct pt_regs * regs) close_fail: filp_close(file, NULL); fail_unlock: + current->fsuid = fsuid; complete_all(&mm->core_done); fail: return retval; diff --git a/fs/proc/base.c b/fs/proc/base.c index e31903aadd96..ace151fa4878 100644 --- a/fs/proc/base.c +++ b/fs/proc/base.c @@ -314,7 +314,7 @@ static int may_ptrace_attach(struct task_struct *task) (current->gid != task->gid)) && !capable(CAP_SYS_PTRACE)) goto out; rmb(); - if (!task->mm->dumpable && !capable(CAP_SYS_PTRACE)) + if (task->mm->dumpable != 1 && !capable(CAP_SYS_PTRACE)) goto out; if (security_ptrace(current, task)) goto out; @@ -1113,7 +1113,9 @@ static int task_dumpable(struct task_struct *task) if (mm) dumpable = mm->dumpable; task_unlock(task); - return dumpable; + if(dumpable == 1) + return 1; + return 0; } diff --git a/include/linux/binfmts.h b/include/linux/binfmts.h index 7e736e201c46..c1e82c514443 100644 --- a/include/linux/binfmts.h +++ b/include/linux/binfmts.h @@ -69,6 +69,11 @@ extern void remove_arg_zero(struct linux_binprm *); extern int search_binary_handler(struct linux_binprm *,struct pt_regs *); extern int flush_old_exec(struct linux_binprm * bprm); +extern int suid_dumpable; +#define SUID_DUMP_DISABLE 0 /* No setuid dumping */ +#define SUID_DUMP_USER 1 /* Dump as user of process */ +#define SUID_DUMP_ROOT 2 /* Dump as root */ + /* Stack area protections */ #define EXSTACK_DEFAULT 0 /* Whatever the arch defaults to */ #define EXSTACK_DISABLE_X 1 /* Disable executable stacks */ diff --git a/include/linux/sched.h b/include/linux/sched.h index b58afd97a180..901742f92389 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -246,7 +246,7 @@ struct mm_struct { unsigned long saved_auxv[42]; /* for /proc/PID/auxv */ - unsigned dumpable:1; + unsigned dumpable:2; cpumask_t cpu_vm_mask; /* Architecture-specific MM context */ diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h index a17745c80a91..614e939c78a4 100644 --- a/include/linux/sysctl.h +++ b/include/linux/sysctl.h @@ -136,6 +136,7 @@ enum KERN_UNKNOWN_NMI_PANIC=66, /* int: unknown nmi panic flag */ KERN_BOOTLOADER_TYPE=67, /* int: boot loader type */ KERN_RANDOMIZE=68, /* int: randomize virtual address space */ + KERN_SETUID_DUMPABLE=69, /* int: behaviour of dumps for setuid core */ }; diff --git a/kernel/sys.c b/kernel/sys.c index f006632c2ba7..0a2c8cda9638 100644 --- a/kernel/sys.c +++ b/kernel/sys.c @@ -525,7 +525,7 @@ asmlinkage long sys_setregid(gid_t rgid, gid_t egid) } if (new_egid != old_egid) { - current->mm->dumpable = 0; + current->mm->dumpable = suid_dumpable; smp_wmb(); } if (rgid != (gid_t) -1 || @@ -556,7 +556,7 @@ asmlinkage long sys_setgid(gid_t gid) { if(old_egid != gid) { - current->mm->dumpable=0; + current->mm->dumpable = suid_dumpable; smp_wmb(); } current->gid = current->egid = current->sgid = current->fsgid = gid; @@ -565,7 +565,7 @@ asmlinkage long sys_setgid(gid_t gid) { if(old_egid != gid) { - current->mm->dumpable=0; + current->mm->dumpable = suid_dumpable; smp_wmb(); } current->egid = current->fsgid = gid; @@ -596,7 +596,7 @@ static int set_user(uid_t new_ruid, int dumpclear) if(dumpclear) { - current->mm->dumpable = 0; + current->mm->dumpable = suid_dumpable; smp_wmb(); } current->uid = new_ruid; @@ -653,7 +653,7 @@ asmlinkage long sys_setreuid(uid_t ruid, uid_t euid) if (new_euid != old_euid) { - current->mm->dumpable=0; + current->mm->dumpable = suid_dumpable; smp_wmb(); } current->fsuid = current->euid = new_euid; @@ -703,7 +703,7 @@ asmlinkage long sys_setuid(uid_t uid) if (old_euid != uid) { - current->mm->dumpable = 0; + current->mm->dumpable = suid_dumpable; smp_wmb(); } current->fsuid = current->euid = uid; @@ -748,7 +748,7 @@ asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid) if (euid != (uid_t) -1) { if (euid != current->euid) { - current->mm->dumpable = 0; + current->mm->dumpable = suid_dumpable; smp_wmb(); } current->euid = euid; @@ -798,7 +798,7 @@ asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid) if (egid != (gid_t) -1) { if (egid != current->egid) { - current->mm->dumpable = 0; + current->mm->dumpable = suid_dumpable; smp_wmb(); } current->egid = egid; @@ -845,7 +845,7 @@ asmlinkage long sys_setfsuid(uid_t uid) { if (uid != old_fsuid) { - current->mm->dumpable = 0; + current->mm->dumpable = suid_dumpable; smp_wmb(); } current->fsuid = uid; @@ -875,7 +875,7 @@ asmlinkage long sys_setfsgid(gid_t gid) { if (gid != old_fsgid) { - current->mm->dumpable = 0; + current->mm->dumpable = suid_dumpable; smp_wmb(); } current->fsgid = gid; @@ -1652,7 +1652,7 @@ asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3, error = 1; break; case PR_SET_DUMPABLE: - if (arg2 != 0 && arg2 != 1) { + if (arg2 < 0 || arg2 > 2) { error = -EINVAL; break; } diff --git a/kernel/sysctl.c b/kernel/sysctl.c index 701d12c63068..24a4d12d5aa9 100644 --- a/kernel/sysctl.c +++ b/kernel/sysctl.c @@ -58,6 +58,7 @@ extern int sysctl_overcommit_ratio; extern int max_threads; extern int sysrq_enabled; extern int core_uses_pid; +extern int suid_dumpable; extern char core_pattern[]; extern int cad_pid; extern int pid_max; @@ -950,6 +951,14 @@ static ctl_table fs_table[] = { .proc_handler = &proc_dointvec, }, #endif + { + .ctl_name = KERN_SETUID_DUMPABLE, + .procname = "suid_dumpable", + .data = &suid_dumpable, + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = &proc_dointvec, + }, { .ctl_name = 0 } }; diff --git a/security/commoncap.c b/security/commoncap.c index 849b8c338ee8..04c12f58d656 100644 --- a/security/commoncap.c +++ b/security/commoncap.c @@ -149,7 +149,7 @@ void cap_bprm_apply_creds (struct linux_binprm *bprm, int unsafe) if (bprm->e_uid != current->uid || bprm->e_gid != current->gid || !cap_issubset (new_permitted, current->cap_permitted)) { - current->mm->dumpable = 0; + current->mm->dumpable = suid_dumpable; if (unsafe & ~LSM_UNSAFE_PTRACE_CAP) { if (!capable(CAP_SETUID)) { diff --git a/security/dummy.c b/security/dummy.c index b32eff146547..6ff887586479 100644 --- a/security/dummy.c +++ b/security/dummy.c @@ -130,7 +130,7 @@ static void dummy_bprm_free_security (struct linux_binprm *bprm) static void dummy_bprm_apply_creds (struct linux_binprm *bprm, int unsafe) { if (bprm->e_uid != current->uid || bprm->e_gid != current->gid) { - current->mm->dumpable = 0; + current->mm->dumpable = suid_dumpable; if ((unsafe & ~LSM_UNSAFE_PTRACE_CAP) && !capable(CAP_SETUID)) { bprm->e_uid = current->uid; -- cgit v1.2.3 From 317a76f9a44b437d6301718f4e5d08bd93f98da7 Mon Sep 17 00:00:00 2001 From: Stephen Hemminger Date: Thu, 23 Jun 2005 12:19:55 -0700 Subject: [TCP]: Add pluggable congestion control algorithm infrastructure. Allow TCP to have multiple pluggable congestion control algorithms. Algorithms are defined by a set of operations and can be built in or modules. The legacy "new RENO" algorithm is used as a starting point and fallback. Signed-off-by: Stephen Hemminger Signed-off-by: David S. Miller --- include/linux/sysctl.h | 9 +- include/linux/tcp.h | 49 +-- include/net/tcp.h | 237 +++++---------- net/ipv4/Makefile | 3 +- net/ipv4/sysctl_net_ipv4.c | 114 +++---- net/ipv4/tcp.c | 2 + net/ipv4/tcp_cong.c | 195 ++++++++++++ net/ipv4/tcp_diag.c | 20 +- net/ipv4/tcp_input.c | 737 ++++----------------------------------------- net/ipv4/tcp_ipv4.c | 3 + net/ipv4/tcp_minisocks.c | 4 +- net/ipv4/tcp_output.c | 23 +- net/ipv6/tcp_ipv6.c | 2 +- 13 files changed, 399 insertions(+), 999 deletions(-) create mode 100644 net/ipv4/tcp_cong.c (limited to 'include/linux/sysctl.h') diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h index 614e939c78a4..72965bfe6cfb 100644 --- a/include/linux/sysctl.h +++ b/include/linux/sysctl.h @@ -333,21 +333,14 @@ enum NET_TCP_FRTO=92, NET_TCP_LOW_LATENCY=93, NET_IPV4_IPFRAG_SECRET_INTERVAL=94, - NET_TCP_WESTWOOD=95, NET_IPV4_IGMP_MAX_MSF=96, NET_TCP_NO_METRICS_SAVE=97, - NET_TCP_VEGAS=98, - NET_TCP_VEGAS_ALPHA=99, - NET_TCP_VEGAS_BETA=100, - NET_TCP_VEGAS_GAMMA=101, - NET_TCP_BIC=102, - NET_TCP_BIC_FAST_CONVERGENCE=103, - NET_TCP_BIC_LOW_WINDOW=104, NET_TCP_DEFAULT_WIN_SCALE=105, NET_TCP_MODERATE_RCVBUF=106, NET_TCP_TSO_WIN_DIVISOR=107, NET_TCP_BIC_BETA=108, NET_IPV4_ICMP_ERRORS_USE_INBOUND_IFADDR=109, + NET_TCP_CONG_CONTROL=110, }; enum { diff --git a/include/linux/tcp.h b/include/linux/tcp.h index 97a7c9e03df5..3ea75dd6640a 100644 --- a/include/linux/tcp.h +++ b/include/linux/tcp.h @@ -203,13 +203,6 @@ struct tcp_sack_block { __u32 end_seq; }; -enum tcp_congestion_algo { - TCP_RENO=0, - TCP_VEGAS, - TCP_WESTWOOD, - TCP_BIC, -}; - struct tcp_options_received { /* PAWS/RTTM data */ long ts_recent_stamp;/* Time we stored ts_recent (for aging) */ @@ -305,7 +298,7 @@ struct tcp_sock { __u8 reordering; /* Packet reordering metric. */ __u8 frto_counter; /* Number of new acks after RTO */ - __u8 adv_cong; /* Using Vegas, Westwood, or BIC */ + __u8 unused; __u8 defer_accept; /* User waits for some data after accept() */ /* RTT measurement */ @@ -401,37 +394,10 @@ struct tcp_sock { __u32 time; } rcvq_space; -/* TCP Westwood structure */ - struct { - __u32 bw_ns_est; /* first bandwidth estimation..not too smoothed 8) */ - __u32 bw_est; /* bandwidth estimate */ - __u32 rtt_win_sx; /* here starts a new evaluation... */ - __u32 bk; - __u32 snd_una; /* used for evaluating the number of acked bytes */ - __u32 cumul_ack; - __u32 accounted; - __u32 rtt; - __u32 rtt_min; /* minimum observed RTT */ - } westwood; - -/* Vegas variables */ - struct { - __u32 beg_snd_nxt; /* right edge during last RTT */ - __u32 beg_snd_una; /* left edge during last RTT */ - __u32 beg_snd_cwnd; /* saves the size of the cwnd */ - __u8 doing_vegas_now;/* if true, do vegas for this RTT */ - __u16 cntRTT; /* # of RTTs measured within last RTT */ - __u32 minRTT; /* min of RTTs measured within last RTT (in usec) */ - __u32 baseRTT; /* the min of all Vegas RTT measurements seen (in usec) */ - } vegas; - - /* BI TCP Parameters */ - struct { - __u32 cnt; /* increase cwnd by 1 after this number of ACKs */ - __u32 last_max_cwnd; /* last maximium snd_cwnd */ - __u32 last_cwnd; /* the last snd_cwnd */ - __u32 last_stamp; /* time when updated last_cwnd */ - } bictcp; + /* Pluggable TCP congestion control hook */ + struct tcp_congestion_ops *ca_ops; + u32 ca_priv[16]; +#define TCP_CA_PRIV_SIZE (16*sizeof(u32)) }; static inline struct tcp_sock *tcp_sk(const struct sock *sk) @@ -439,6 +405,11 @@ static inline struct tcp_sock *tcp_sk(const struct sock *sk) return (struct tcp_sock *)sk; } +static inline void *tcp_ca(const struct tcp_sock *tp) +{ + return (void *) tp->ca_priv; +} + #endif #endif /* _LINUX_TCP_H */ diff --git a/include/net/tcp.h b/include/net/tcp.h index f730935b824a..e427cf35915c 100644 --- a/include/net/tcp.h +++ b/include/net/tcp.h @@ -505,25 +505,6 @@ static __inline__ int tcp_sk_listen_hashfn(struct sock *sk) #else # define TCP_TW_RECYCLE_TICK (12+2-TCP_TW_RECYCLE_SLOTS_LOG) #endif - -#define BICTCP_BETA_SCALE 1024 /* Scale factor beta calculation - * max_cwnd = snd_cwnd * beta - */ -#define BICTCP_MAX_INCREMENT 32 /* - * Limit on the amount of - * increment allowed during - * binary search. - */ -#define BICTCP_FUNC_OF_MIN_INCR 11 /* - * log(B/Smin)/log(B/(B-1))+1, - * Smin:min increment - * B:log factor - */ -#define BICTCP_B 4 /* - * In binary search, - * go to point (max+min)/N - */ - /* * TCP option */ @@ -596,16 +577,7 @@ extern int sysctl_tcp_adv_win_scale; extern int sysctl_tcp_tw_reuse; extern int sysctl_tcp_frto; extern int sysctl_tcp_low_latency; -extern int sysctl_tcp_westwood; -extern int sysctl_tcp_vegas_cong_avoid; -extern int sysctl_tcp_vegas_alpha; -extern int sysctl_tcp_vegas_beta; -extern int sysctl_tcp_vegas_gamma; extern int sysctl_tcp_nometrics_save; -extern int sysctl_tcp_bic; -extern int sysctl_tcp_bic_fast_convergence; -extern int sysctl_tcp_bic_low_window; -extern int sysctl_tcp_bic_beta; extern int sysctl_tcp_moderate_rcvbuf; extern int sysctl_tcp_tso_win_divisor; @@ -1136,6 +1108,80 @@ static inline void tcp_packets_out_dec(struct tcp_sock *tp, tp->packets_out -= tcp_skb_pcount(skb); } +/* Events passed to congestion control interface */ +enum tcp_ca_event { + CA_EVENT_TX_START, /* first transmit when no packets in flight */ + CA_EVENT_CWND_RESTART, /* congestion window restart */ + CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */ + CA_EVENT_FRTO, /* fast recovery timeout */ + CA_EVENT_LOSS, /* loss timeout */ + CA_EVENT_FAST_ACK, /* in sequence ack */ + CA_EVENT_SLOW_ACK, /* other ack */ +}; + +/* + * Interface for adding new TCP congestion control handlers + */ +#define TCP_CA_NAME_MAX 16 +struct tcp_congestion_ops { + struct list_head list; + + /* initialize private data (optional) */ + void (*init)(struct tcp_sock *tp); + /* cleanup private data (optional) */ + void (*release)(struct tcp_sock *tp); + + /* return slow start threshold (required) */ + u32 (*ssthresh)(struct tcp_sock *tp); + /* lower bound for congestion window (optional) */ + u32 (*min_cwnd)(struct tcp_sock *tp); + /* do new cwnd calculation (required) */ + void (*cong_avoid)(struct tcp_sock *tp, u32 ack, + u32 rtt, u32 in_flight, int good_ack); + /* round trip time sample per acked packet (optional) */ + void (*rtt_sample)(struct tcp_sock *tp, u32 usrtt); + /* call before changing ca_state (optional) */ + void (*set_state)(struct tcp_sock *tp, u8 new_state); + /* call when cwnd event occurs (optional) */ + void (*cwnd_event)(struct tcp_sock *tp, enum tcp_ca_event ev); + /* new value of cwnd after loss (optional) */ + u32 (*undo_cwnd)(struct tcp_sock *tp); + /* hook for packet ack accounting (optional) */ + void (*pkts_acked)(struct tcp_sock *tp, u32 num_acked); + /* get info for tcp_diag (optional) */ + void (*get_info)(struct tcp_sock *tp, u32 ext, struct sk_buff *skb); + + char name[TCP_CA_NAME_MAX]; + struct module *owner; +}; + +extern int tcp_register_congestion_control(struct tcp_congestion_ops *type); +extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type); + +extern void tcp_init_congestion_control(struct tcp_sock *tp); +extern void tcp_cleanup_congestion_control(struct tcp_sock *tp); +extern int tcp_set_default_congestion_control(const char *name); +extern void tcp_get_default_congestion_control(char *name); + +extern struct tcp_congestion_ops tcp_reno; +extern u32 tcp_reno_ssthresh(struct tcp_sock *tp); +extern void tcp_reno_cong_avoid(struct tcp_sock *tp, u32 ack, + u32 rtt, u32 in_flight, int flag); +extern u32 tcp_reno_min_cwnd(struct tcp_sock *tp); + +static inline void tcp_set_ca_state(struct tcp_sock *tp, u8 ca_state) +{ + if (tp->ca_ops->set_state) + tp->ca_ops->set_state(tp, ca_state); + tp->ca_state = ca_state; +} + +static inline void tcp_ca_event(struct tcp_sock *tp, enum tcp_ca_event event) +{ + if (tp->ca_ops->cwnd_event) + tp->ca_ops->cwnd_event(tp, event); +} + /* This determines how many packets are "in the network" to the best * of our knowledge. In many cases it is conservative, but where * detailed information is available from the receiver (via SACK @@ -1155,91 +1201,6 @@ static __inline__ unsigned int tcp_packets_in_flight(const struct tcp_sock *tp) return (tp->packets_out - tp->left_out + tp->retrans_out); } -/* - * Which congestion algorithim is in use on the connection. - */ -#define tcp_is_vegas(__tp) ((__tp)->adv_cong == TCP_VEGAS) -#define tcp_is_westwood(__tp) ((__tp)->adv_cong == TCP_WESTWOOD) -#define tcp_is_bic(__tp) ((__tp)->adv_cong == TCP_BIC) - -/* Recalculate snd_ssthresh, we want to set it to: - * - * Reno: - * one half the current congestion window, but no - * less than two segments - * - * BIC: - * behave like Reno until low_window is reached, - * then increase congestion window slowly - */ -static inline __u32 tcp_recalc_ssthresh(struct tcp_sock *tp) -{ - if (tcp_is_bic(tp)) { - if (sysctl_tcp_bic_fast_convergence && - tp->snd_cwnd < tp->bictcp.last_max_cwnd) - tp->bictcp.last_max_cwnd = (tp->snd_cwnd * - (BICTCP_BETA_SCALE - + sysctl_tcp_bic_beta)) - / (2 * BICTCP_BETA_SCALE); - else - tp->bictcp.last_max_cwnd = tp->snd_cwnd; - - if (tp->snd_cwnd > sysctl_tcp_bic_low_window) - return max((tp->snd_cwnd * sysctl_tcp_bic_beta) - / BICTCP_BETA_SCALE, 2U); - } - - return max(tp->snd_cwnd >> 1U, 2U); -} - -/* Stop taking Vegas samples for now. */ -#define tcp_vegas_disable(__tp) ((__tp)->vegas.doing_vegas_now = 0) - -static inline void tcp_vegas_enable(struct tcp_sock *tp) -{ - /* There are several situations when we must "re-start" Vegas: - * - * o when a connection is established - * o after an RTO - * o after fast recovery - * o when we send a packet and there is no outstanding - * unacknowledged data (restarting an idle connection) - * - * In these circumstances we cannot do a Vegas calculation at the - * end of the first RTT, because any calculation we do is using - * stale info -- both the saved cwnd and congestion feedback are - * stale. - * - * Instead we must wait until the completion of an RTT during - * which we actually receive ACKs. - */ - - /* Begin taking Vegas samples next time we send something. */ - tp->vegas.doing_vegas_now = 1; - - /* Set the beginning of the next send window. */ - tp->vegas.beg_snd_nxt = tp->snd_nxt; - - tp->vegas.cntRTT = 0; - tp->vegas.minRTT = 0x7fffffff; -} - -/* Should we be taking Vegas samples right now? */ -#define tcp_vegas_enabled(__tp) ((__tp)->vegas.doing_vegas_now) - -extern void tcp_ca_init(struct tcp_sock *tp); - -static inline void tcp_set_ca_state(struct tcp_sock *tp, u8 ca_state) -{ - if (tcp_is_vegas(tp)) { - if (ca_state == TCP_CA_Open) - tcp_vegas_enable(tp); - else - tcp_vegas_disable(tp); - } - tp->ca_state = ca_state; -} - /* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd. * The exception is rate halving phase, when cwnd is decreasing towards * ssthresh. @@ -1288,7 +1249,7 @@ static inline void tcp_cwnd_validate(struct sock *sk, struct tcp_sock *tp) static inline void __tcp_enter_cwr(struct tcp_sock *tp) { tp->undo_marker = 0; - tp->snd_ssthresh = tcp_recalc_ssthresh(tp); + tp->snd_ssthresh = tp->ca_ops->ssthresh(tp); tp->snd_cwnd = min(tp->snd_cwnd, tcp_packets_in_flight(tp) + 1U); tp->snd_cwnd_cnt = 0; @@ -1876,52 +1837,4 @@ struct tcp_iter_state { extern int tcp_proc_register(struct tcp_seq_afinfo *afinfo); extern void tcp_proc_unregister(struct tcp_seq_afinfo *afinfo); -/* TCP Westwood functions and constants */ - -#define TCP_WESTWOOD_INIT_RTT (20*HZ) /* maybe too conservative?! */ -#define TCP_WESTWOOD_RTT_MIN (HZ/20) /* 50ms */ - -static inline void tcp_westwood_update_rtt(struct tcp_sock *tp, __u32 rtt_seq) -{ - if (tcp_is_westwood(tp)) - tp->westwood.rtt = rtt_seq; -} - -static inline __u32 __tcp_westwood_bw_rttmin(const struct tcp_sock *tp) -{ - return max((tp->westwood.bw_est) * (tp->westwood.rtt_min) / - (__u32) (tp->mss_cache_std), - 2U); -} - -static inline __u32 tcp_westwood_bw_rttmin(const struct tcp_sock *tp) -{ - return tcp_is_westwood(tp) ? __tcp_westwood_bw_rttmin(tp) : 0; -} - -static inline int tcp_westwood_ssthresh(struct tcp_sock *tp) -{ - __u32 ssthresh = 0; - - if (tcp_is_westwood(tp)) { - ssthresh = __tcp_westwood_bw_rttmin(tp); - if (ssthresh) - tp->snd_ssthresh = ssthresh; - } - - return (ssthresh != 0); -} - -static inline int tcp_westwood_cwnd(struct tcp_sock *tp) -{ - __u32 cwnd = 0; - - if (tcp_is_westwood(tp)) { - cwnd = __tcp_westwood_bw_rttmin(tp); - if (cwnd) - tp->snd_cwnd = cwnd; - } - - return (cwnd != 0); -} #endif /* _TCP_H */ diff --git a/net/ipv4/Makefile b/net/ipv4/Makefile index 65d57d8e1add..89c0b4cb470e 100644 --- a/net/ipv4/Makefile +++ b/net/ipv4/Makefile @@ -5,7 +5,8 @@ obj-y := utils.o route.o inetpeer.o protocol.o \ ip_input.o ip_fragment.o ip_forward.o ip_options.o \ ip_output.o ip_sockglue.o \ - tcp.o tcp_input.o tcp_output.o tcp_timer.o tcp_ipv4.o tcp_minisocks.o \ + tcp.o tcp_input.o tcp_output.o tcp_timer.o tcp_ipv4.o \ + tcp_minisocks.o tcp_cong.o \ datagram.o raw.o udp.o arp.o icmp.o devinet.o af_inet.o igmp.o \ sysctl_net_ipv4.o fib_frontend.o fib_semantics.o diff --git a/net/ipv4/sysctl_net_ipv4.c b/net/ipv4/sysctl_net_ipv4.c index 23068bddbf0b..e32894532416 100644 --- a/net/ipv4/sysctl_net_ipv4.c +++ b/net/ipv4/sysctl_net_ipv4.c @@ -118,6 +118,45 @@ static int ipv4_sysctl_forward_strategy(ctl_table *table, return 1; } +static int proc_tcp_congestion_control(ctl_table *ctl, int write, struct file * filp, + void __user *buffer, size_t *lenp, loff_t *ppos) +{ + char val[TCP_CA_NAME_MAX]; + ctl_table tbl = { + .data = val, + .maxlen = TCP_CA_NAME_MAX, + }; + int ret; + + tcp_get_default_congestion_control(val); + + ret = proc_dostring(&tbl, write, filp, buffer, lenp, ppos); + if (write && ret == 0) + ret = tcp_set_default_congestion_control(val); + return ret; +} + +int sysctl_tcp_congestion_control(ctl_table *table, int __user *name, int nlen, + void __user *oldval, size_t __user *oldlenp, + void __user *newval, size_t newlen, + void **context) +{ + char val[TCP_CA_NAME_MAX]; + ctl_table tbl = { + .data = val, + .maxlen = TCP_CA_NAME_MAX, + }; + int ret; + + tcp_get_default_congestion_control(val); + ret = sysctl_string(&tbl, name, nlen, oldval, oldlenp, newval, newlen, + context); + if (ret == 0 && newval && newlen) + ret = tcp_set_default_congestion_control(val); + return ret; +} + + ctl_table ipv4_table[] = { { .ctl_name = NET_IPV4_TCP_TIMESTAMPS, @@ -611,70 +650,6 @@ ctl_table ipv4_table[] = { .mode = 0644, .proc_handler = &proc_dointvec, }, - { - .ctl_name = NET_TCP_WESTWOOD, - .procname = "tcp_westwood", - .data = &sysctl_tcp_westwood, - .maxlen = sizeof(int), - .mode = 0644, - .proc_handler = &proc_dointvec, - }, - { - .ctl_name = NET_TCP_VEGAS, - .procname = "tcp_vegas_cong_avoid", - .data = &sysctl_tcp_vegas_cong_avoid, - .maxlen = sizeof(int), - .mode = 0644, - .proc_handler = &proc_dointvec, - }, - { - .ctl_name = NET_TCP_VEGAS_ALPHA, - .procname = "tcp_vegas_alpha", - .data = &sysctl_tcp_vegas_alpha, - .maxlen = sizeof(int), - .mode = 0644, - .proc_handler = &proc_dointvec, - }, - { - .ctl_name = NET_TCP_VEGAS_BETA, - .procname = "tcp_vegas_beta", - .data = &sysctl_tcp_vegas_beta, - .maxlen = sizeof(int), - .mode = 0644, - .proc_handler = &proc_dointvec, - }, - { - .ctl_name = NET_TCP_VEGAS_GAMMA, - .procname = "tcp_vegas_gamma", - .data = &sysctl_tcp_vegas_gamma, - .maxlen = sizeof(int), - .mode = 0644, - .proc_handler = &proc_dointvec, - }, - { - .ctl_name = NET_TCP_BIC, - .procname = "tcp_bic", - .data = &sysctl_tcp_bic, - .maxlen = sizeof(int), - .mode = 0644, - .proc_handler = &proc_dointvec, - }, - { - .ctl_name = NET_TCP_BIC_FAST_CONVERGENCE, - .procname = "tcp_bic_fast_convergence", - .data = &sysctl_tcp_bic_fast_convergence, - .maxlen = sizeof(int), - .mode = 0644, - .proc_handler = &proc_dointvec, - }, - { - .ctl_name = NET_TCP_BIC_LOW_WINDOW, - .procname = "tcp_bic_low_window", - .data = &sysctl_tcp_bic_low_window, - .maxlen = sizeof(int), - .mode = 0644, - .proc_handler = &proc_dointvec, - }, { .ctl_name = NET_TCP_MODERATE_RCVBUF, .procname = "tcp_moderate_rcvbuf", @@ -692,13 +667,14 @@ ctl_table ipv4_table[] = { .proc_handler = &proc_dointvec, }, { - .ctl_name = NET_TCP_BIC_BETA, - .procname = "tcp_bic_beta", - .data = &sysctl_tcp_bic_beta, - .maxlen = sizeof(int), + .ctl_name = NET_TCP_CONG_CONTROL, + .procname = "tcp_congestion_control", .mode = 0644, - .proc_handler = &proc_dointvec, + .maxlen = TCP_CA_NAME_MAX, + .proc_handler = &proc_tcp_congestion_control, + .strategy = &sysctl_tcp_congestion_control, }, + { .ctl_name = 0 } }; diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c index 674bbd8cfd36..f3dbc8dc1263 100644 --- a/net/ipv4/tcp.c +++ b/net/ipv4/tcp.c @@ -2333,6 +2333,8 @@ void __init tcp_init(void) printk(KERN_INFO "TCP: Hash tables configured " "(established %d bind %d)\n", tcp_ehash_size << 1, tcp_bhash_size); + + tcp_register_congestion_control(&tcp_reno); } EXPORT_SYMBOL(tcp_accept); diff --git a/net/ipv4/tcp_cong.c b/net/ipv4/tcp_cong.c new file mode 100644 index 000000000000..665394a63ae4 --- /dev/null +++ b/net/ipv4/tcp_cong.c @@ -0,0 +1,195 @@ +/* + * Plugable TCP congestion control support and newReno + * congestion control. + * Based on ideas from I/O scheduler suport and Web100. + * + * Copyright (C) 2005 Stephen Hemminger + */ + +#include +#include +#include +#include +#include +#include + +static DEFINE_SPINLOCK(tcp_cong_list_lock); +static LIST_HEAD(tcp_cong_list); + +/* Simple linear search, don't expect many entries! */ +static struct tcp_congestion_ops *tcp_ca_find(const char *name) +{ + struct tcp_congestion_ops *e; + + list_for_each_entry(e, &tcp_cong_list, list) { + if (strcmp(e->name, name) == 0) + return e; + } + + return NULL; +} + +/* + * Attach new congestion control algorthim to the list + * of available options. + */ +int tcp_register_congestion_control(struct tcp_congestion_ops *ca) +{ + int ret = 0; + + /* all algorithms must implement ssthresh and cong_avoid ops */ + if (!ca->ssthresh || !ca->cong_avoid || !ca->min_cwnd) { + printk(KERN_ERR "TCP %s does not implement required ops\n", + ca->name); + return -EINVAL; + } + + spin_lock(&tcp_cong_list_lock); + if (tcp_ca_find(ca->name)) { + printk(KERN_NOTICE "TCP %s already registered\n", ca->name); + ret = -EEXIST; + } else { + list_add_rcu(&ca->list, &tcp_cong_list); + printk(KERN_INFO "TCP %s registered\n", ca->name); + } + spin_unlock(&tcp_cong_list_lock); + + return ret; +} +EXPORT_SYMBOL_GPL(tcp_register_congestion_control); + +/* + * Remove congestion control algorithm, called from + * the module's remove function. Module ref counts are used + * to ensure that this can't be done till all sockets using + * that method are closed. + */ +void tcp_unregister_congestion_control(struct tcp_congestion_ops *ca) +{ + spin_lock(&tcp_cong_list_lock); + list_del_rcu(&ca->list); + spin_unlock(&tcp_cong_list_lock); +} +EXPORT_SYMBOL_GPL(tcp_unregister_congestion_control); + +/* Assign choice of congestion control. */ +void tcp_init_congestion_control(struct tcp_sock *tp) +{ + struct tcp_congestion_ops *ca; + + rcu_read_lock(); + list_for_each_entry_rcu(ca, &tcp_cong_list, list) { + if (try_module_get(ca->owner)) { + tp->ca_ops = ca; + break; + } + + } + rcu_read_unlock(); + + if (tp->ca_ops->init) + tp->ca_ops->init(tp); +} + +/* Manage refcounts on socket close. */ +void tcp_cleanup_congestion_control(struct tcp_sock *tp) +{ + if (tp->ca_ops->release) + tp->ca_ops->release(tp); + module_put(tp->ca_ops->owner); +} + +/* Used by sysctl to change default congestion control */ +int tcp_set_default_congestion_control(const char *name) +{ + struct tcp_congestion_ops *ca; + int ret = -ENOENT; + + spin_lock(&tcp_cong_list_lock); + ca = tcp_ca_find(name); +#ifdef CONFIG_KMOD + if (!ca) { + spin_unlock(&tcp_cong_list_lock); + + request_module("tcp_%s", name); + spin_lock(&tcp_cong_list_lock); + ca = tcp_ca_find(name); + } +#endif + + if (ca) { + list_move(&ca->list, &tcp_cong_list); + ret = 0; + } + spin_unlock(&tcp_cong_list_lock); + + return ret; +} + +/* Get current default congestion control */ +void tcp_get_default_congestion_control(char *name) +{ + struct tcp_congestion_ops *ca; + /* We will always have reno... */ + BUG_ON(list_empty(&tcp_cong_list)); + + rcu_read_lock(); + ca = list_entry(tcp_cong_list.next, struct tcp_congestion_ops, list); + strncpy(name, ca->name, TCP_CA_NAME_MAX); + rcu_read_unlock(); +} + +/* + * TCP Reno congestion control + * This is special case used for fallback as well. + */ +/* This is Jacobson's slow start and congestion avoidance. + * SIGCOMM '88, p. 328. + */ +void tcp_reno_cong_avoid(struct tcp_sock *tp, u32 ack, u32 rtt, u32 in_flight, + int flag) +{ + if (in_flight < tp->snd_cwnd) + return; + + if (tp->snd_cwnd <= tp->snd_ssthresh) { + /* In "safe" area, increase. */ + if (tp->snd_cwnd < tp->snd_cwnd_clamp) + tp->snd_cwnd++; + } else { + /* In dangerous area, increase slowly. + * In theory this is tp->snd_cwnd += 1 / tp->snd_cwnd + */ + if (tp->snd_cwnd_cnt >= tp->snd_cwnd) { + if (tp->snd_cwnd < tp->snd_cwnd_clamp) + tp->snd_cwnd++; + tp->snd_cwnd_cnt = 0; + } else + tp->snd_cwnd_cnt++; + } +} +EXPORT_SYMBOL_GPL(tcp_reno_cong_avoid); + +/* Slow start threshold is half the congestion window (min 2) */ +u32 tcp_reno_ssthresh(struct tcp_sock *tp) +{ + return max(tp->snd_cwnd >> 1U, 2U); +} +EXPORT_SYMBOL_GPL(tcp_reno_ssthresh); + +/* Lower bound on congestion window. */ +u32 tcp_reno_min_cwnd(struct tcp_sock *tp) +{ + return tp->snd_ssthresh/2; +} +EXPORT_SYMBOL_GPL(tcp_reno_min_cwnd); + +struct tcp_congestion_ops tcp_reno = { + .name = "reno", + .owner = THIS_MODULE, + .ssthresh = tcp_reno_ssthresh, + .cong_avoid = tcp_reno_cong_avoid, + .min_cwnd = tcp_reno_min_cwnd, +}; + +EXPORT_SYMBOL_GPL(tcp_reno); diff --git a/net/ipv4/tcp_diag.c b/net/ipv4/tcp_diag.c index 634befc07921..867acc0f79d8 100644 --- a/net/ipv4/tcp_diag.c +++ b/net/ipv4/tcp_diag.c @@ -42,7 +42,6 @@ struct tcpdiag_entry static struct sock *tcpnl; - #define TCPDIAG_PUT(skb, attrtype, attrlen) \ ({ int rtalen = RTA_LENGTH(attrlen); \ struct rtattr *rta; \ @@ -61,7 +60,6 @@ static int tcpdiag_fill(struct sk_buff *skb, struct sock *sk, struct nlmsghdr *nlh; struct tcp_info *info = NULL; struct tcpdiag_meminfo *minfo = NULL; - struct tcpvegas_info *vinfo = NULL; unsigned char *b = skb->tail; nlh = NLMSG_PUT(skb, pid, seq, TCPDIAG_GETSOCK, sizeof(*r)); @@ -73,9 +71,6 @@ static int tcpdiag_fill(struct sk_buff *skb, struct sock *sk, if (ext & (1<<(TCPDIAG_INFO-1))) info = TCPDIAG_PUT(skb, TCPDIAG_INFO, sizeof(*info)); - if ((tcp_is_westwood(tp) || tcp_is_vegas(tp)) - && (ext & (1<<(TCPDIAG_VEGASINFO-1)))) - vinfo = TCPDIAG_PUT(skb, TCPDIAG_VEGASINFO, sizeof(*vinfo)); } r->tcpdiag_family = sk->sk_family; r->tcpdiag_state = sk->sk_state; @@ -166,19 +161,8 @@ static int tcpdiag_fill(struct sk_buff *skb, struct sock *sk, if (info) tcp_get_info(sk, info); - if (vinfo) { - if (tcp_is_vegas(tp)) { - vinfo->tcpv_enabled = tp->vegas.doing_vegas_now; - vinfo->tcpv_rttcnt = tp->vegas.cntRTT; - vinfo->tcpv_rtt = jiffies_to_usecs(tp->vegas.baseRTT); - vinfo->tcpv_minrtt = jiffies_to_usecs(tp->vegas.minRTT); - } else { - vinfo->tcpv_enabled = 0; - vinfo->tcpv_rttcnt = 0; - vinfo->tcpv_rtt = jiffies_to_usecs(tp->westwood.rtt); - vinfo->tcpv_minrtt = jiffies_to_usecs(tp->westwood.rtt_min); - } - } + if (sk->sk_state < TCP_TIME_WAIT && tp->ca_ops->get_info) + tp->ca_ops->get_info(tp, ext, skb); nlh->nlmsg_len = skb->tail - b; return skb->len; diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index 5bad504630a3..7bbbbc33eb4b 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -61,7 +61,6 @@ * Panu Kuhlberg: Experimental audit of TCP (re)transmission * engine. Lots of bugs are found. * Pasi Sarolahti: F-RTO for dealing with spurious RTOs - * Angelo Dell'Aera: TCP Westwood+ support */ #include @@ -88,23 +87,9 @@ int sysctl_tcp_rfc1337; int sysctl_tcp_max_orphans = NR_FILE; int sysctl_tcp_frto; int sysctl_tcp_nometrics_save; -int sysctl_tcp_westwood; -int sysctl_tcp_vegas_cong_avoid; int sysctl_tcp_moderate_rcvbuf = 1; -/* Default values of the Vegas variables, in fixed-point representation - * with V_PARAM_SHIFT bits to the right of the binary point. - */ -#define V_PARAM_SHIFT 1 -int sysctl_tcp_vegas_alpha = 1<snd_cwnd_stamp = tcp_time_stamp; } -static void init_bictcp(struct tcp_sock *tp) -{ - tp->bictcp.cnt = 0; - - tp->bictcp.last_max_cwnd = 0; - tp->bictcp.last_cwnd = 0; - tp->bictcp.last_stamp = 0; -} - /* 5. Recalculate window clamp after socket hit its memory bounds. */ static void tcp_clamp_window(struct sock *sk, struct tcp_sock *tp) { @@ -558,45 +534,6 @@ static void tcp_event_data_recv(struct sock *sk, struct tcp_sock *tp, struct sk_ tcp_grow_window(sk, tp, skb); } -/* When starting a new connection, pin down the current choice of - * congestion algorithm. - */ -void tcp_ca_init(struct tcp_sock *tp) -{ - if (sysctl_tcp_westwood) - tp->adv_cong = TCP_WESTWOOD; - else if (sysctl_tcp_bic) - tp->adv_cong = TCP_BIC; - else if (sysctl_tcp_vegas_cong_avoid) { - tp->adv_cong = TCP_VEGAS; - tp->vegas.baseRTT = 0x7fffffff; - tcp_vegas_enable(tp); - } -} - -/* Do RTT sampling needed for Vegas. - * Basically we: - * o min-filter RTT samples from within an RTT to get the current - * propagation delay + queuing delay (we are min-filtering to try to - * avoid the effects of delayed ACKs) - * o min-filter RTT samples from a much longer window (forever for now) - * to find the propagation delay (baseRTT) - */ -static inline void vegas_rtt_calc(struct tcp_sock *tp, __u32 rtt) -{ - __u32 vrtt = rtt + 1; /* Never allow zero rtt or baseRTT */ - - /* Filter to find propagation delay: */ - if (vrtt < tp->vegas.baseRTT) - tp->vegas.baseRTT = vrtt; - - /* Find the min RTT during the last RTT to find - * the current prop. delay + queuing delay: - */ - tp->vegas.minRTT = min(tp->vegas.minRTT, vrtt); - tp->vegas.cntRTT++; -} - /* Called to compute a smoothed rtt estimate. The data fed to this * routine either comes from timestamps, or from segments that were * known _not_ to have been retransmitted [see Karn/Partridge @@ -606,13 +543,10 @@ static inline void vegas_rtt_calc(struct tcp_sock *tp, __u32 rtt) * To save cycles in the RFC 1323 implementation it was better to break * it up into three procedures. -- erics */ -static void tcp_rtt_estimator(struct tcp_sock *tp, __u32 mrtt) +static void tcp_rtt_estimator(struct tcp_sock *tp, __u32 mrtt, u32 *usrtt) { long m = mrtt; /* RTT */ - if (tcp_vegas_enabled(tp)) - vegas_rtt_calc(tp, mrtt); - /* The following amusing code comes from Jacobson's * article in SIGCOMM '88. Note that rtt and mdev * are scaled versions of rtt and mean deviation. @@ -670,7 +604,8 @@ static void tcp_rtt_estimator(struct tcp_sock *tp, __u32 mrtt) tp->rtt_seq = tp->snd_nxt; } - tcp_westwood_update_rtt(tp, tp->srtt >> 3); + if (tp->ca_ops->rtt_sample) + tp->ca_ops->rtt_sample(tp, *usrtt); } /* Calculate rto without backoff. This is the second half of Van Jacobson's @@ -1185,8 +1120,8 @@ void tcp_enter_frto(struct sock *sk) tp->snd_una == tp->high_seq || (tp->ca_state == TCP_CA_Loss && !tp->retransmits)) { tp->prior_ssthresh = tcp_current_ssthresh(tp); - if (!tcp_westwood_ssthresh(tp)) - tp->snd_ssthresh = tcp_recalc_ssthresh(tp); + tp->snd_ssthresh = tp->ca_ops->ssthresh(tp); + tcp_ca_event(tp, CA_EVENT_FRTO); } /* Have to clear retransmission markers here to keep the bookkeeping @@ -1252,8 +1187,6 @@ static void tcp_enter_frto_loss(struct sock *sk) tcp_set_ca_state(tp, TCP_CA_Loss); tp->high_seq = tp->frto_highmark; TCP_ECN_queue_cwr(tp); - - init_bictcp(tp); } void tcp_clear_retrans(struct tcp_sock *tp) @@ -1283,7 +1216,8 @@ void tcp_enter_loss(struct sock *sk, int how) if (tp->ca_state <= TCP_CA_Disorder || tp->snd_una == tp->high_seq || (tp->ca_state == TCP_CA_Loss && !tp->retransmits)) { tp->prior_ssthresh = tcp_current_ssthresh(tp); - tp->snd_ssthresh = tcp_recalc_ssthresh(tp); + tp->snd_ssthresh = tp->ca_ops->ssthresh(tp); + tcp_ca_event(tp, CA_EVENT_LOSS); } tp->snd_cwnd = 1; tp->snd_cwnd_cnt = 0; @@ -1596,28 +1530,14 @@ static inline void tcp_moderate_cwnd(struct tcp_sock *tp) } /* Decrease cwnd each second ack. */ - static void tcp_cwnd_down(struct tcp_sock *tp) { int decr = tp->snd_cwnd_cnt + 1; - __u32 limit; - - /* - * TCP Westwood - * Here limit is evaluated as BWestimation*RTTmin (for obtaining it - * in packets we use mss_cache). If sysctl_tcp_westwood is off - * tcp_westwood_bw_rttmin() returns 0. In such case snd_ssthresh is - * still used as usual. It prevents other strange cases in which - * BWE*RTTmin could assume value 0. It should not happen but... - */ - - if (!(limit = tcp_westwood_bw_rttmin(tp))) - limit = tp->snd_ssthresh/2; tp->snd_cwnd_cnt = decr&1; decr >>= 1; - if (decr && tp->snd_cwnd > limit) + if (decr && tp->snd_cwnd > tp->ca_ops->min_cwnd(tp)) tp->snd_cwnd -= decr; tp->snd_cwnd = min(tp->snd_cwnd, tcp_packets_in_flight(tp)+1); @@ -1654,8 +1574,8 @@ static void DBGUNDO(struct sock *sk, struct tcp_sock *tp, const char *msg) static void tcp_undo_cwr(struct tcp_sock *tp, int undo) { if (tp->prior_ssthresh) { - if (tcp_is_bic(tp)) - tp->snd_cwnd = max(tp->snd_cwnd, tp->bictcp.last_max_cwnd); + if (tp->ca_ops->undo_cwnd) + tp->snd_cwnd = tp->ca_ops->undo_cwnd(tp); else tp->snd_cwnd = max(tp->snd_cwnd, tp->snd_ssthresh<<1); @@ -1767,11 +1687,9 @@ static int tcp_try_undo_loss(struct sock *sk, struct tcp_sock *tp) static inline void tcp_complete_cwr(struct tcp_sock *tp) { - if (tcp_westwood_cwnd(tp)) - tp->snd_ssthresh = tp->snd_cwnd; - else - tp->snd_cwnd = min(tp->snd_cwnd, tp->snd_ssthresh); + tp->snd_cwnd = min(tp->snd_cwnd, tp->snd_ssthresh); tp->snd_cwnd_stamp = tcp_time_stamp; + tcp_ca_event(tp, CA_EVENT_COMPLETE_CWR); } static void tcp_try_to_open(struct sock *sk, struct tcp_sock *tp, int flag) @@ -1946,7 +1864,7 @@ tcp_fastretrans_alert(struct sock *sk, u32 prior_snd_una, if (tp->ca_state < TCP_CA_CWR) { if (!(flag&FLAG_ECE)) tp->prior_ssthresh = tcp_current_ssthresh(tp); - tp->snd_ssthresh = tcp_recalc_ssthresh(tp); + tp->snd_ssthresh = tp->ca_ops->ssthresh(tp); TCP_ECN_queue_cwr(tp); } @@ -1963,7 +1881,7 @@ tcp_fastretrans_alert(struct sock *sk, u32 prior_snd_una, /* Read draft-ietf-tcplw-high-performance before mucking * with this code. (Superceeds RFC1323) */ -static void tcp_ack_saw_tstamp(struct tcp_sock *tp, int flag) +static void tcp_ack_saw_tstamp(struct tcp_sock *tp, u32 *usrtt, int flag) { __u32 seq_rtt; @@ -1983,13 +1901,13 @@ static void tcp_ack_saw_tstamp(struct tcp_sock *tp, int flag) * in window is lost... Voila. --ANK (010210) */ seq_rtt = tcp_time_stamp - tp->rx_opt.rcv_tsecr; - tcp_rtt_estimator(tp, seq_rtt); + tcp_rtt_estimator(tp, seq_rtt, usrtt); tcp_set_rto(tp); tp->backoff = 0; tcp_bound_rto(tp); } -static void tcp_ack_no_tstamp(struct tcp_sock *tp, u32 seq_rtt, int flag) +static void tcp_ack_no_tstamp(struct tcp_sock *tp, u32 seq_rtt, u32 *usrtt, int flag) { /* We don't have a timestamp. Can only use * packets that are not retransmitted to determine @@ -2003,338 +1921,29 @@ static void tcp_ack_no_tstamp(struct tcp_sock *tp, u32 seq_rtt, int flag) if (flag & FLAG_RETRANS_DATA_ACKED) return; - tcp_rtt_estimator(tp, seq_rtt); + tcp_rtt_estimator(tp, seq_rtt, usrtt); tcp_set_rto(tp); tp->backoff = 0; tcp_bound_rto(tp); } static inline void tcp_ack_update_rtt(struct tcp_sock *tp, - int flag, s32 seq_rtt) + int flag, s32 seq_rtt, u32 *usrtt) { /* Note that peer MAY send zero echo. In this case it is ignored. (rfc1323) */ if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr) - tcp_ack_saw_tstamp(tp, flag); + tcp_ack_saw_tstamp(tp, usrtt, flag); else if (seq_rtt >= 0) - tcp_ack_no_tstamp(tp, seq_rtt, flag); + tcp_ack_no_tstamp(tp, seq_rtt, usrtt, flag); } -/* - * Compute congestion window to use. - * - * This is from the implementation of BICTCP in - * Lison-Xu, Kahaled Harfoush, and Injog Rhee. - * "Binary Increase Congestion Control for Fast, Long Distance - * Networks" in InfoComm 2004 - * Available from: - * http://www.csc.ncsu.edu/faculty/rhee/export/bitcp.pdf - * - * Unless BIC is enabled and congestion window is large - * this behaves the same as the original Reno. - */ -static inline __u32 bictcp_cwnd(struct tcp_sock *tp) -{ - /* orignal Reno behaviour */ - if (!tcp_is_bic(tp)) - return tp->snd_cwnd; - - if (tp->bictcp.last_cwnd == tp->snd_cwnd && - (s32)(tcp_time_stamp - tp->bictcp.last_stamp) <= (HZ>>5)) - return tp->bictcp.cnt; - - tp->bictcp.last_cwnd = tp->snd_cwnd; - tp->bictcp.last_stamp = tcp_time_stamp; - - /* start off normal */ - if (tp->snd_cwnd <= sysctl_tcp_bic_low_window) - tp->bictcp.cnt = tp->snd_cwnd; - - /* binary increase */ - else if (tp->snd_cwnd < tp->bictcp.last_max_cwnd) { - __u32 dist = (tp->bictcp.last_max_cwnd - tp->snd_cwnd) - / BICTCP_B; - - if (dist > BICTCP_MAX_INCREMENT) - /* linear increase */ - tp->bictcp.cnt = tp->snd_cwnd / BICTCP_MAX_INCREMENT; - else if (dist <= 1U) - /* binary search increase */ - tp->bictcp.cnt = tp->snd_cwnd * BICTCP_FUNC_OF_MIN_INCR - / BICTCP_B; - else - /* binary search increase */ - tp->bictcp.cnt = tp->snd_cwnd / dist; - } else { - /* slow start amd linear increase */ - if (tp->snd_cwnd < tp->bictcp.last_max_cwnd + BICTCP_B) - /* slow start */ - tp->bictcp.cnt = tp->snd_cwnd * BICTCP_FUNC_OF_MIN_INCR - / BICTCP_B; - else if (tp->snd_cwnd < tp->bictcp.last_max_cwnd - + BICTCP_MAX_INCREMENT*(BICTCP_B-1)) - /* slow start */ - tp->bictcp.cnt = tp->snd_cwnd * (BICTCP_B-1) - / (tp->snd_cwnd-tp->bictcp.last_max_cwnd); - else - /* linear increase */ - tp->bictcp.cnt = tp->snd_cwnd / BICTCP_MAX_INCREMENT; - } - return tp->bictcp.cnt; -} - -/* This is Jacobson's slow start and congestion avoidance. - * SIGCOMM '88, p. 328. - */ -static inline void reno_cong_avoid(struct tcp_sock *tp) +static inline void tcp_cong_avoid(struct tcp_sock *tp, u32 ack, u32 rtt, + u32 in_flight, int good) { - if (tp->snd_cwnd <= tp->snd_ssthresh) { - /* In "safe" area, increase. */ - if (tp->snd_cwnd < tp->snd_cwnd_clamp) - tp->snd_cwnd++; - } else { - /* In dangerous area, increase slowly. - * In theory this is tp->snd_cwnd += 1 / tp->snd_cwnd - */ - if (tp->snd_cwnd_cnt >= bictcp_cwnd(tp)) { - if (tp->snd_cwnd < tp->snd_cwnd_clamp) - tp->snd_cwnd++; - tp->snd_cwnd_cnt=0; - } else - tp->snd_cwnd_cnt++; - } + tp->ca_ops->cong_avoid(tp, ack, rtt, in_flight, good); tp->snd_cwnd_stamp = tcp_time_stamp; } -/* This is based on the congestion detection/avoidance scheme described in - * Lawrence S. Brakmo and Larry L. Peterson. - * "TCP Vegas: End to end congestion avoidance on a global internet." - * IEEE Journal on Selected Areas in Communication, 13(8):1465--1480, - * October 1995. Available from: - * ftp://ftp.cs.arizona.edu/xkernel/Papers/jsac.ps - * - * See http://www.cs.arizona.edu/xkernel/ for their implementation. - * The main aspects that distinguish this implementation from the - * Arizona Vegas implementation are: - * o We do not change the loss detection or recovery mechanisms of - * Linux in any way. Linux already recovers from losses quite well, - * using fine-grained timers, NewReno, and FACK. - * o To avoid the performance penalty imposed by increasing cwnd - * only every-other RTT during slow start, we increase during - * every RTT during slow start, just like Reno. - * o Largely to allow continuous cwnd growth during slow start, - * we use the rate at which ACKs come back as the "actual" - * rate, rather than the rate at which data is sent. - * o To speed convergence to the right rate, we set the cwnd - * to achieve the right ("actual") rate when we exit slow start. - * o To filter out the noise caused by delayed ACKs, we use the - * minimum RTT sample observed during the last RTT to calculate - * the actual rate. - * o When the sender re-starts from idle, it waits until it has - * received ACKs for an entire flight of new data before making - * a cwnd adjustment decision. The original Vegas implementation - * assumed senders never went idle. - */ -static void vegas_cong_avoid(struct tcp_sock *tp, u32 ack, u32 seq_rtt) -{ - /* The key players are v_beg_snd_una and v_beg_snd_nxt. - * - * These are so named because they represent the approximate values - * of snd_una and snd_nxt at the beginning of the current RTT. More - * precisely, they represent the amount of data sent during the RTT. - * At the end of the RTT, when we receive an ACK for v_beg_snd_nxt, - * we will calculate that (v_beg_snd_nxt - v_beg_snd_una) outstanding - * bytes of data have been ACKed during the course of the RTT, giving - * an "actual" rate of: - * - * (v_beg_snd_nxt - v_beg_snd_una) / (rtt duration) - * - * Unfortunately, v_beg_snd_una is not exactly equal to snd_una, - * because delayed ACKs can cover more than one segment, so they - * don't line up nicely with the boundaries of RTTs. - * - * Another unfortunate fact of life is that delayed ACKs delay the - * advance of the left edge of our send window, so that the number - * of bytes we send in an RTT is often less than our cwnd will allow. - * So we keep track of our cwnd separately, in v_beg_snd_cwnd. - */ - - if (after(ack, tp->vegas.beg_snd_nxt)) { - /* Do the Vegas once-per-RTT cwnd adjustment. */ - u32 old_wnd, old_snd_cwnd; - - - /* Here old_wnd is essentially the window of data that was - * sent during the previous RTT, and has all - * been acknowledged in the course of the RTT that ended - * with the ACK we just received. Likewise, old_snd_cwnd - * is the cwnd during the previous RTT. - */ - old_wnd = (tp->vegas.beg_snd_nxt - tp->vegas.beg_snd_una) / - tp->mss_cache_std; - old_snd_cwnd = tp->vegas.beg_snd_cwnd; - - /* Save the extent of the current window so we can use this - * at the end of the next RTT. - */ - tp->vegas.beg_snd_una = tp->vegas.beg_snd_nxt; - tp->vegas.beg_snd_nxt = tp->snd_nxt; - tp->vegas.beg_snd_cwnd = tp->snd_cwnd; - - /* Take into account the current RTT sample too, to - * decrease the impact of delayed acks. This double counts - * this sample since we count it for the next window as well, - * but that's not too awful, since we're taking the min, - * rather than averaging. - */ - vegas_rtt_calc(tp, seq_rtt); - - /* We do the Vegas calculations only if we got enough RTT - * samples that we can be reasonably sure that we got - * at least one RTT sample that wasn't from a delayed ACK. - * If we only had 2 samples total, - * then that means we're getting only 1 ACK per RTT, which - * means they're almost certainly delayed ACKs. - * If we have 3 samples, we should be OK. - */ - - if (tp->vegas.cntRTT <= 2) { - /* We don't have enough RTT samples to do the Vegas - * calculation, so we'll behave like Reno. - */ - if (tp->snd_cwnd > tp->snd_ssthresh) - tp->snd_cwnd++; - } else { - u32 rtt, target_cwnd, diff; - - /* We have enough RTT samples, so, using the Vegas - * algorithm, we determine if we should increase or - * decrease cwnd, and by how much. - */ - - /* Pluck out the RTT we are using for the Vegas - * calculations. This is the min RTT seen during the - * last RTT. Taking the min filters out the effects - * of delayed ACKs, at the cost of noticing congestion - * a bit later. - */ - rtt = tp->vegas.minRTT; - - /* Calculate the cwnd we should have, if we weren't - * going too fast. - * - * This is: - * (actual rate in segments) * baseRTT - * We keep it as a fixed point number with - * V_PARAM_SHIFT bits to the right of the binary point. - */ - target_cwnd = ((old_wnd * tp->vegas.baseRTT) - << V_PARAM_SHIFT) / rtt; - - /* Calculate the difference between the window we had, - * and the window we would like to have. This quantity - * is the "Diff" from the Arizona Vegas papers. - * - * Again, this is a fixed point number with - * V_PARAM_SHIFT bits to the right of the binary - * point. - */ - diff = (old_wnd << V_PARAM_SHIFT) - target_cwnd; - - if (tp->snd_cwnd < tp->snd_ssthresh) { - /* Slow start. */ - if (diff > sysctl_tcp_vegas_gamma) { - /* Going too fast. Time to slow down - * and switch to congestion avoidance. - */ - tp->snd_ssthresh = 2; - - /* Set cwnd to match the actual rate - * exactly: - * cwnd = (actual rate) * baseRTT - * Then we add 1 because the integer - * truncation robs us of full link - * utilization. - */ - tp->snd_cwnd = min(tp->snd_cwnd, - (target_cwnd >> - V_PARAM_SHIFT)+1); - - } - } else { - /* Congestion avoidance. */ - u32 next_snd_cwnd; - - /* Figure out where we would like cwnd - * to be. - */ - if (diff > sysctl_tcp_vegas_beta) { - /* The old window was too fast, so - * we slow down. - */ - next_snd_cwnd = old_snd_cwnd - 1; - } else if (diff < sysctl_tcp_vegas_alpha) { - /* We don't have enough extra packets - * in the network, so speed up. - */ - next_snd_cwnd = old_snd_cwnd + 1; - } else { - /* Sending just as fast as we - * should be. - */ - next_snd_cwnd = old_snd_cwnd; - } - - /* Adjust cwnd upward or downward, toward the - * desired value. - */ - if (next_snd_cwnd > tp->snd_cwnd) - tp->snd_cwnd++; - else if (next_snd_cwnd < tp->snd_cwnd) - tp->snd_cwnd--; - } - } - - /* Wipe the slate clean for the next RTT. */ - tp->vegas.cntRTT = 0; - tp->vegas.minRTT = 0x7fffffff; - } - - /* The following code is executed for every ack we receive, - * except for conditions checked in should_advance_cwnd() - * before the call to tcp_cong_avoid(). Mainly this means that - * we only execute this code if the ack actually acked some - * data. - */ - - /* If we are in slow start, increase our cwnd in response to this ACK. - * (If we are not in slow start then we are in congestion avoidance, - * and adjust our congestion window only once per RTT. See the code - * above.) - */ - if (tp->snd_cwnd <= tp->snd_ssthresh) - tp->snd_cwnd++; - - /* to keep cwnd from growing without bound */ - tp->snd_cwnd = min_t(u32, tp->snd_cwnd, tp->snd_cwnd_clamp); - - /* Make sure that we are never so timid as to reduce our cwnd below - * 2 MSS. - * - * Going below 2 MSS would risk huge delayed ACKs from our receiver. - */ - tp->snd_cwnd = max(tp->snd_cwnd, 2U); - - tp->snd_cwnd_stamp = tcp_time_stamp; -} - -static inline void tcp_cong_avoid(struct tcp_sock *tp, u32 ack, u32 seq_rtt) -{ - if (tcp_vegas_enabled(tp)) - vegas_cong_avoid(tp, ack, seq_rtt); - else - reno_cong_avoid(tp); -} - /* Restart timer after forward progress on connection. * RFC2988 recommends to restart timer to now+rto. */ @@ -2415,13 +2024,18 @@ static int tcp_tso_acked(struct sock *sk, struct sk_buff *skb, /* Remove acknowledged frames from the retransmission queue. */ -static int tcp_clean_rtx_queue(struct sock *sk, __s32 *seq_rtt_p) +static int tcp_clean_rtx_queue(struct sock *sk, __s32 *seq_rtt_p, s32 *seq_usrtt) { struct tcp_sock *tp = tcp_sk(sk); struct sk_buff *skb; __u32 now = tcp_time_stamp; int acked = 0; __s32 seq_rtt = -1; + struct timeval usnow; + u32 pkts_acked = 0; + + if (seq_usrtt) + do_gettimeofday(&usnow); while ((skb = skb_peek(&sk->sk_write_queue)) && skb != sk->sk_send_head) { @@ -2448,6 +2062,7 @@ static int tcp_clean_rtx_queue(struct sock *sk, __s32 *seq_rtt_p) */ if (!(scb->flags & TCPCB_FLAG_SYN)) { acked |= FLAG_DATA_ACKED; + ++pkts_acked; } else { acked |= FLAG_SYN_ACKED; tp->retrans_stamp = 0; @@ -2461,6 +2076,10 @@ static int tcp_clean_rtx_queue(struct sock *sk, __s32 *seq_rtt_p) seq_rtt = -1; } else if (seq_rtt < 0) seq_rtt = now - scb->when; + if (seq_usrtt) + *seq_usrtt = (usnow.tv_sec - skb->stamp.tv_sec) * 1000000 + + (usnow.tv_usec - skb->stamp.tv_usec); + if (sacked & TCPCB_SACKED_ACKED) tp->sacked_out -= tcp_skb_pcount(skb); if (sacked & TCPCB_LOST) @@ -2479,8 +2098,11 @@ static int tcp_clean_rtx_queue(struct sock *sk, __s32 *seq_rtt_p) } if (acked&FLAG_ACKED) { - tcp_ack_update_rtt(tp, acked, seq_rtt); + tcp_ack_update_rtt(tp, acked, seq_rtt, seq_usrtt); tcp_ack_packets_out(sk, tp); + + if (tp->ca_ops->pkts_acked) + tp->ca_ops->pkts_acked(tp, pkts_acked); } #if FASTRETRANS_DEBUG > 0 @@ -2624,257 +2246,6 @@ static void tcp_process_frto(struct sock *sk, u32 prior_snd_una) tp->frto_counter = (tp->frto_counter + 1) % 3; } -/* - * TCP Westwood+ - */ - -/* - * @init_westwood - * This function initializes fields used in TCP Westwood+. We can't - * get no information about RTTmin at this time so we simply set it to - * TCP_WESTWOOD_INIT_RTT. This value was chosen to be too conservative - * since in this way we're sure it will be updated in a consistent - * way as soon as possible. It will reasonably happen within the first - * RTT period of the connection lifetime. - */ - -static void init_westwood(struct sock *sk) -{ - struct tcp_sock *tp = tcp_sk(sk); - - tp->westwood.bw_ns_est = 0; - tp->westwood.bw_est = 0; - tp->westwood.accounted = 0; - tp->westwood.cumul_ack = 0; - tp->westwood.rtt_win_sx = tcp_time_stamp; - tp->westwood.rtt = TCP_WESTWOOD_INIT_RTT; - tp->westwood.rtt_min = TCP_WESTWOOD_INIT_RTT; - tp->westwood.snd_una = tp->snd_una; -} - -/* - * @westwood_do_filter - * Low-pass filter. Implemented using constant coeffients. - */ - -static inline __u32 westwood_do_filter(__u32 a, __u32 b) -{ - return (((7 * a) + b) >> 3); -} - -static void westwood_filter(struct sock *sk, __u32 delta) -{ - struct tcp_sock *tp = tcp_sk(sk); - - tp->westwood.bw_ns_est = - westwood_do_filter(tp->westwood.bw_ns_est, - tp->westwood.bk / delta); - tp->westwood.bw_est = - westwood_do_filter(tp->westwood.bw_est, - tp->westwood.bw_ns_est); -} - -/* - * @westwood_update_rttmin - * It is used to update RTTmin. In this case we MUST NOT use - * WESTWOOD_RTT_MIN minimum bound since we could be on a LAN! - */ - -static inline __u32 westwood_update_rttmin(const struct sock *sk) -{ - const struct tcp_sock *tp = tcp_sk(sk); - __u32 rttmin = tp->westwood.rtt_min; - - if (tp->westwood.rtt != 0 && - (tp->westwood.rtt < tp->westwood.rtt_min || !rttmin)) - rttmin = tp->westwood.rtt; - - return rttmin; -} - -/* - * @westwood_acked - * Evaluate increases for dk. - */ - -static inline __u32 westwood_acked(const struct sock *sk) -{ - const struct tcp_sock *tp = tcp_sk(sk); - - return tp->snd_una - tp->westwood.snd_una; -} - -/* - * @westwood_new_window - * It evaluates if we are receiving data inside the same RTT window as - * when we started. - * Return value: - * It returns 0 if we are still evaluating samples in the same RTT - * window, 1 if the sample has to be considered in the next window. - */ - -static int westwood_new_window(const struct sock *sk) -{ - const struct tcp_sock *tp = tcp_sk(sk); - __u32 left_bound; - __u32 rtt; - int ret = 0; - - left_bound = tp->westwood.rtt_win_sx; - rtt = max(tp->westwood.rtt, (u32) TCP_WESTWOOD_RTT_MIN); - - /* - * A RTT-window has passed. Be careful since if RTT is less than - * 50ms we don't filter but we continue 'building the sample'. - * This minimum limit was choosen since an estimation on small - * time intervals is better to avoid... - * Obvioulsy on a LAN we reasonably will always have - * right_bound = left_bound + WESTWOOD_RTT_MIN - */ - - if ((left_bound + rtt) < tcp_time_stamp) - ret = 1; - - return ret; -} - -/* - * @westwood_update_window - * It updates RTT evaluation window if it is the right moment to do - * it. If so it calls filter for evaluating bandwidth. - */ - -static void __westwood_update_window(struct sock *sk, __u32 now) -{ - struct tcp_sock *tp = tcp_sk(sk); - __u32 delta = now - tp->westwood.rtt_win_sx; - - if (delta) { - if (tp->westwood.rtt) - westwood_filter(sk, delta); - - tp->westwood.bk = 0; - tp->westwood.rtt_win_sx = tcp_time_stamp; - } -} - - -static void westwood_update_window(struct sock *sk, __u32 now) -{ - if (westwood_new_window(sk)) - __westwood_update_window(sk, now); -} - -/* - * @__tcp_westwood_fast_bw - * It is called when we are in fast path. In particular it is called when - * header prediction is successfull. In such case infact update is - * straight forward and doesn't need any particular care. - */ - -static void __tcp_westwood_fast_bw(struct sock *sk, struct sk_buff *skb) -{ - struct tcp_sock *tp = tcp_sk(sk); - - westwood_update_window(sk, tcp_time_stamp); - - tp->westwood.bk += westwood_acked(sk); - tp->westwood.snd_una = tp->snd_una; - tp->westwood.rtt_min = westwood_update_rttmin(sk); -} - -static inline void tcp_westwood_fast_bw(struct sock *sk, struct sk_buff *skb) -{ - if (tcp_is_westwood(tcp_sk(sk))) - __tcp_westwood_fast_bw(sk, skb); -} - - -/* - * @westwood_dupack_update - * It updates accounted and cumul_ack when receiving a dupack. - */ - -static void westwood_dupack_update(struct sock *sk) -{ - struct tcp_sock *tp = tcp_sk(sk); - - tp->westwood.accounted += tp->mss_cache_std; - tp->westwood.cumul_ack = tp->mss_cache_std; -} - -static inline int westwood_may_change_cumul(struct tcp_sock *tp) -{ - return (tp->westwood.cumul_ack > tp->mss_cache_std); -} - -static inline void westwood_partial_update(struct tcp_sock *tp) -{ - tp->westwood.accounted -= tp->westwood.cumul_ack; - tp->westwood.cumul_ack = tp->mss_cache_std; -} - -static inline void westwood_complete_update(struct tcp_sock *tp) -{ - tp->westwood.cumul_ack -= tp->westwood.accounted; - tp->westwood.accounted = 0; -} - -/* - * @westwood_acked_count - * This function evaluates cumul_ack for evaluating dk in case of - * delayed or partial acks. - */ - -static inline __u32 westwood_acked_count(struct sock *sk) -{ - struct tcp_sock *tp = tcp_sk(sk); - - tp->westwood.cumul_ack = westwood_acked(sk); - - /* If cumul_ack is 0 this is a dupack since it's not moving - * tp->snd_una. - */ - if (!(tp->westwood.cumul_ack)) - westwood_dupack_update(sk); - - if (westwood_may_change_cumul(tp)) { - /* Partial or delayed ack */ - if (tp->westwood.accounted >= tp->westwood.cumul_ack) - westwood_partial_update(tp); - else - westwood_complete_update(tp); - } - - tp->westwood.snd_una = tp->snd_una; - - return tp->westwood.cumul_ack; -} - - -/* - * @__tcp_westwood_slow_bw - * It is called when something is going wrong..even if there could - * be no problems! Infact a simple delayed packet may trigger a - * dupack. But we need to be careful in such case. - */ - -static void __tcp_westwood_slow_bw(struct sock *sk, struct sk_buff *skb) -{ - struct tcp_sock *tp = tcp_sk(sk); - - westwood_update_window(sk, tcp_time_stamp); - - tp->westwood.bk += westwood_acked_count(sk); - tp->westwood.rtt_min = westwood_update_rttmin(sk); -} - -static inline void tcp_westwood_slow_bw(struct sock *sk, struct sk_buff *skb) -{ - if (tcp_is_westwood(tcp_sk(sk))) - __tcp_westwood_slow_bw(sk, skb); -} - /* This routine deals with incoming acks, but not outgoing ones. */ static int tcp_ack(struct sock *sk, struct sk_buff *skb, int flag) { @@ -2884,6 +2255,7 @@ static int tcp_ack(struct sock *sk, struct sk_buff *skb, int flag) u32 ack = TCP_SKB_CB(skb)->ack_seq; u32 prior_in_flight; s32 seq_rtt; + s32 seq_usrtt = 0; int prior_packets; /* If the ack is newer than sent or older than previous acks @@ -2902,9 +2274,10 @@ static int tcp_ack(struct sock *sk, struct sk_buff *skb, int flag) */ tcp_update_wl(tp, ack, ack_seq); tp->snd_una = ack; - tcp_westwood_fast_bw(sk, skb); flag |= FLAG_WIN_UPDATE; + tcp_ca_event(tp, CA_EVENT_FAST_ACK); + NET_INC_STATS_BH(LINUX_MIB_TCPHPACKS); } else { if (ack_seq != TCP_SKB_CB(skb)->end_seq) @@ -2920,7 +2293,7 @@ static int tcp_ack(struct sock *sk, struct sk_buff *skb, int flag) if (TCP_ECN_rcv_ecn_echo(tp, skb->h.th)) flag |= FLAG_ECE; - tcp_westwood_slow_bw(sk,skb); + tcp_ca_event(tp, CA_EVENT_SLOW_ACK); } /* We passed data and got it acked, remove any soft error @@ -2935,22 +2308,20 @@ static int tcp_ack(struct sock *sk, struct sk_buff *skb, int flag) prior_in_flight = tcp_packets_in_flight(tp); /* See if we can take anything off of the retransmit queue. */ - flag |= tcp_clean_rtx_queue(sk, &seq_rtt); + flag |= tcp_clean_rtx_queue(sk, &seq_rtt, + tp->ca_ops->rtt_sample ? &seq_usrtt : NULL); if (tp->frto_counter) tcp_process_frto(sk, prior_snd_una); if (tcp_ack_is_dubious(tp, flag)) { /* Advanve CWND, if state allows this. */ - if ((flag & FLAG_DATA_ACKED) && - (tcp_vegas_enabled(tp) || prior_in_flight >= tp->snd_cwnd) && - tcp_may_raise_cwnd(tp, flag)) - tcp_cong_avoid(tp, ack, seq_rtt); + if ((flag & FLAG_DATA_ACKED) && tcp_may_raise_cwnd(tp, flag)) + tcp_cong_avoid(tp, ack, seq_rtt, prior_in_flight, 0); tcp_fastretrans_alert(sk, prior_snd_una, prior_packets, flag); } else { - if ((flag & FLAG_DATA_ACKED) && - (tcp_vegas_enabled(tp) || prior_in_flight >= tp->snd_cwnd)) - tcp_cong_avoid(tp, ack, seq_rtt); + if ((flag & FLAG_DATA_ACKED)) + tcp_cong_avoid(tp, ack, seq_rtt, prior_in_flight, 1); } if ((flag & FLAG_FORWARD_PROGRESS) || !(flag&FLAG_NOT_DUP)) @@ -4552,6 +3923,8 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb, tcp_init_metrics(sk); + tcp_init_congestion_control(tp); + /* Prevent spurious tcp_cwnd_restart() on first data * packet. */ @@ -4708,9 +4081,6 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb, if(tp->af_specific->conn_request(sk, skb) < 0) return 1; - init_westwood(sk); - init_bictcp(tp); - /* Now we have several options: In theory there is * nothing else in the frame. KA9Q has an option to * send data with the syn, BSD accepts data with the @@ -4732,9 +4102,6 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb, goto discard; case TCP_SYN_SENT: - init_westwood(sk); - init_bictcp(tp); - queued = tcp_rcv_synsent_state_process(sk, skb, th, len); if (queued >= 0) return queued; @@ -4816,7 +4183,7 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb, */ if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr && !tp->srtt) - tcp_ack_saw_tstamp(tp, 0); + tcp_ack_saw_tstamp(tp, 0, 0); if (tp->rx_opt.tstamp_ok) tp->advmss -= TCPOLEN_TSTAMP_ALIGNED; @@ -4828,6 +4195,8 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb, tcp_init_metrics(sk); + tcp_init_congestion_control(tp); + /* Prevent spurious tcp_cwnd_restart() on * first data packet. */ diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c index 2d41d5d6ad19..9122814c13ad 100644 --- a/net/ipv4/tcp_ipv4.c +++ b/net/ipv4/tcp_ipv4.c @@ -2048,6 +2048,7 @@ static int tcp_v4_init_sock(struct sock *sk) tp->mss_cache_std = tp->mss_cache = 536; tp->reordering = sysctl_tcp_reordering; + tp->ca_ops = &tcp_reno; sk->sk_state = TCP_CLOSE; @@ -2070,6 +2071,8 @@ int tcp_v4_destroy_sock(struct sock *sk) tcp_clear_xmit_timers(sk); + tcp_cleanup_congestion_control(tp); + /* Cleanup up the write buffer. */ sk_stream_writequeue_purge(sk); diff --git a/net/ipv4/tcp_minisocks.c b/net/ipv4/tcp_minisocks.c index b3943e7562f3..f42a284164b7 100644 --- a/net/ipv4/tcp_minisocks.c +++ b/net/ipv4/tcp_minisocks.c @@ -774,6 +774,8 @@ struct sock *tcp_create_openreq_child(struct sock *sk, struct request_sock *req, newtp->frto_counter = 0; newtp->frto_highmark = 0; + newtp->ca_ops = &tcp_reno; + tcp_set_ca_state(newtp, TCP_CA_Open); tcp_init_xmit_timers(newsk); skb_queue_head_init(&newtp->out_of_order_queue); @@ -842,8 +844,6 @@ struct sock *tcp_create_openreq_child(struct sock *sk, struct request_sock *req, if (newtp->ecn_flags&TCP_ECN_OK) sock_set_flag(newsk, SOCK_NO_LARGESEND); - tcp_ca_init(newtp); - TCP_INC_STATS_BH(TCP_MIB_PASSIVEOPENS); } return newsk; diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c index f17c6577e337..0e17c244875c 100644 --- a/net/ipv4/tcp_output.c +++ b/net/ipv4/tcp_output.c @@ -111,8 +111,7 @@ static void tcp_cwnd_restart(struct tcp_sock *tp, struct dst_entry *dst) u32 restart_cwnd = tcp_init_cwnd(tp, dst); u32 cwnd = tp->snd_cwnd; - if (tcp_is_vegas(tp)) - tcp_vegas_enable(tp); + tcp_ca_event(tp, CA_EVENT_CWND_RESTART); tp->snd_ssthresh = tcp_current_ssthresh(tp); restart_cwnd = min(restart_cwnd, cwnd); @@ -280,6 +279,10 @@ static int tcp_transmit_skb(struct sock *sk, struct sk_buff *skb) #define SYSCTL_FLAG_WSCALE 0x2 #define SYSCTL_FLAG_SACK 0x4 + /* If congestion control is doing timestamping */ + if (tp->ca_ops->rtt_sample) + do_gettimeofday(&skb->stamp); + sysctl_flags = 0; if (tcb->flags & TCPCB_FLAG_SYN) { tcp_header_size = sizeof(struct tcphdr) + TCPOLEN_MSS; @@ -304,17 +307,8 @@ static int tcp_transmit_skb(struct sock *sk, struct sk_buff *skb) (tp->rx_opt.eff_sacks * TCPOLEN_SACK_PERBLOCK)); } - /* - * If the connection is idle and we are restarting, - * then we don't want to do any Vegas calculations - * until we get fresh RTT samples. So when we - * restart, we reset our Vegas state to a clean - * slate. After we get acks for this flight of - * packets, _then_ we can make Vegas calculations - * again. - */ - if (tcp_is_vegas(tp) && tcp_packets_in_flight(tp) == 0) - tcp_vegas_enable(tp); + if (tcp_packets_in_flight(tp) == 0) + tcp_ca_event(tp, CA_EVENT_TX_START); th = (struct tcphdr *) skb_push(skb, tcp_header_size); skb->h.th = th; @@ -521,6 +515,7 @@ static int tcp_fragment(struct sock *sk, struct sk_buff *skb, u32 len) * skbs, which it never sent before. --ANK */ TCP_SKB_CB(buff)->when = TCP_SKB_CB(skb)->when; + buff->stamp = skb->stamp; if (TCP_SKB_CB(skb)->sacked & TCPCB_LOST) { tp->lost_out -= tcp_skb_pcount(skb); @@ -1449,7 +1444,6 @@ static inline void tcp_connect_init(struct sock *sk) tp->window_clamp = dst_metric(dst, RTAX_WINDOW); tp->advmss = dst_metric(dst, RTAX_ADVMSS); tcp_initialize_rcv_mss(sk); - tcp_ca_init(tp); tcp_select_initial_window(tcp_full_space(sk), tp->advmss - (tp->rx_opt.ts_recent_stamp ? tp->tcp_header_len - sizeof(struct tcphdr) : 0), @@ -1503,7 +1497,6 @@ int tcp_connect(struct sock *sk) TCP_SKB_CB(buff)->end_seq = tp->write_seq; tp->snd_nxt = tp->write_seq; tp->pushed_seq = tp->write_seq; - tcp_ca_init(tp); /* Send it off. */ TCP_SKB_CB(buff)->when = tcp_time_stamp; diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c index 2414937f2a83..fce56039b0e9 100644 --- a/net/ipv6/tcp_ipv6.c +++ b/net/ipv6/tcp_ipv6.c @@ -2025,7 +2025,7 @@ static int tcp_v6_init_sock(struct sock *sk) sk->sk_state = TCP_CLOSE; tp->af_specific = &ipv6_specific; - + tp->ca_ops = &tcp_reno; sk->sk_write_space = sk_stream_write_space; sock_set_flag(sk, SOCK_USE_WRITE_QUEUE); -- cgit v1.2.3 From 51b0bdedb8e784d0d969a6b77151911130812400 Mon Sep 17 00:00:00 2001 From: Stephen Hemminger Date: Thu, 23 Jun 2005 20:14:40 -0700 Subject: [NET]: Separate two usages of netdev_max_backlog. Separate out the two uses of netdev_max_backlog. One controls the upper bound on packets processed per softirq, the new name for this is netdev_budget; the other controls the limit on packets queued via netif_rx. Increase the max_backlog default to account for faster processors. Signed-off-by: Stephen Hemminger Signed-off-by: David S. Miller --- include/linux/sysctl.h | 1 + net/core/dev.c | 6 +++--- net/core/sysctl_net_core.c | 9 +++++++++ 3 files changed, 13 insertions(+), 3 deletions(-) (limited to 'include/linux/sysctl.h') diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h index 72965bfe6cfb..ebfe1250f0a4 100644 --- a/include/linux/sysctl.h +++ b/include/linux/sysctl.h @@ -243,6 +243,7 @@ enum NET_CORE_MOD_CONG=16, NET_CORE_DEV_WEIGHT=17, NET_CORE_SOMAXCONN=18, + NET_CORE_BUDGET=19, }; /* /proc/sys/net/ethernet */ diff --git a/net/core/dev.c b/net/core/dev.c index 1a64508e527f..7016e0c36b3d 100644 --- a/net/core/dev.c +++ b/net/core/dev.c @@ -1346,7 +1346,8 @@ out: Receiver routines =======================================================================*/ -int netdev_max_backlog = 300; +int netdev_max_backlog = 1000; +int netdev_budget = 300; int weight_p = 64; /* old backlog weight */ DEFINE_PER_CPU(struct netif_rx_stats, netdev_rx_stat) = { 0, }; @@ -1695,8 +1696,7 @@ static void net_rx_action(struct softirq_action *h) { struct softnet_data *queue = &__get_cpu_var(softnet_data); unsigned long start_time = jiffies; - int budget = netdev_max_backlog; - + int budget = netdev_budget; local_irq_disable(); diff --git a/net/core/sysctl_net_core.c b/net/core/sysctl_net_core.c index fff63643a35c..8f817ad9f546 100644 --- a/net/core/sysctl_net_core.c +++ b/net/core/sysctl_net_core.c @@ -13,6 +13,7 @@ #ifdef CONFIG_SYSCTL extern int netdev_max_backlog; +extern int netdev_budget; extern int weight_p; extern int net_msg_cost; extern int net_msg_burst; @@ -124,6 +125,14 @@ ctl_table core_table[] = { .mode = 0644, .proc_handler = &proc_dointvec }, + { + .ctl_name = NET_CORE_BUDGET, + .procname = "netdev_budget", + .data = &netdev_budget, + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = &proc_dointvec + }, { .ctl_name = 0 } }; -- cgit v1.2.3