--- linux-2.4.29/include/linux/if.h 2004-08-07 16:26:06.000000000 -0700 +++ linux-2.4.29.amds/include/linux/if.h 2005-04-01 11:30:49.000000000 -0800 @@ -50,6 +50,19 @@ /* Private (from user) interface flags (netdevice->priv_flags). */ #define IFF_802_1Q_VLAN 0x1 /* 802.1Q VLAN device. */ +#define IFF_PKTGEN_RCV 0x2 /* Registered to receive & consume Pktgen skbs */ +#define IFF_ACCEPT_LOCAL_ADDRS 0x4 /** Accept pkts even if they come from a local + * address. This lets use send pkts to ourselves + * over external interfaces (when used in conjunction + * with SO_BINDTODEVICE + */ +#define IFF_ACCEPT_ALL_FRAMES 0x8 /** Accept all frames, even ones with bad CRCs. + * Should only be used in debugging/testing situations + * Do NOT enable this unless you understand the + * consequences! */ +#define IFF_SAVE_FCS 0x10 /** Save the Frame Check Sum (FCS) on receive, if + * possible. */ +#define IFF_MAC_VLAN 0x40 /* MAC VLAN device. */ #define IF_GET_IFACE 0x0001 /* for querying only */ --- linux-2.4.29/include/linux/netdevice.h 2004-11-17 03:54:22.000000000 -0800 +++ linux-2.4.29.amds/include/linux/netdevice.h 2005-04-01 11:30:49.000000000 -0800 @@ -311,7 +311,9 @@ unsigned short flags; /* interface flags (a la BSD) */ unsigned short gflags; - unsigned short priv_flags; /* Like 'flags' but invisible to userspace. */ + unsigned short priv_flags; /* Like 'flags' but invisible to userspace, + * see: if.h for flag definitions. + */ unsigned short unused_alignment_fixer; /* Because we need priv_flags, * and we want to be 32-bit aligned. */ @@ -382,7 +384,13 @@ #define NETIF_F_HW_VLAN_RX 256 /* Receive VLAN hw acceleration */ #define NETIF_F_HW_VLAN_FILTER 512 /* Receive filtering on VLAN */ #define NETIF_F_VLAN_CHALLENGED 1024 /* Device cannot handle VLAN packets */ +#define NETIF_F_RX_ALL 2048 /* Can be configured to receive all packets, even + * ones with busted CRC. May disable VLAN filtering + * in the NIC, users should NOT enable this feature + * unless they understand the consequences. */ +#define NETIF_F_SAVE_CRC 4096 /* Can save FCS in skb, last 4 bytes for ethernet */ + /* Called after device is detached from network. */ void (*uninit)(struct net_device *dev); /* Called after last user reference disappears. */ @@ -437,12 +445,25 @@ int (*neigh_setup)(struct net_device *dev, struct neigh_parms *); int (*accept_fastpath)(struct net_device *, struct dst_entry*); +#ifdef CONFIG_NET_SKB_RECYCLING + int (*skb_recycle) (struct sk_buff *skb); + void (*mem_reclaim) (struct net_device *dev); +#endif /* open/release and usage marking */ struct module *owner; + /* Callback for when the queue is woken, used by pktgen currently */ + int (*notify_queue_woken)(struct net_device *dev); + void* nqw_data; /* To be used by the method above as needed */ + + /* bridge stuff */ struct net_bridge_port *br_port; +#if defined(CONFIG_MACVLAN) || defined(CONFIG_MACVLAN_MODULE) + struct macvlan_port *macvlan_priv; +#endif + #ifdef CONFIG_NET_FASTROUTE #define NETDEV_FASTROUTE_HMASK 0xF /* Semi-private data. Keep it at the end of device struct. */ @@ -453,6 +474,7 @@ /* this will get initialized at each interface type init routine */ struct divert_blk *divert; #endif /* CONFIG_NET_DIVERT */ + }; /* 2.6 compatibility */ @@ -563,8 +585,12 @@ static inline void netif_wake_queue(struct net_device *dev) { - if (test_and_clear_bit(__LINK_STATE_XOFF, &dev->state)) + if (test_and_clear_bit(__LINK_STATE_XOFF, &dev->state)) { __netif_schedule(dev); + if (dev->notify_queue_woken) { + dev->notify_queue_woken(dev); + } + } } static inline void netif_stop_queue(struct net_device *dev) --- linux-2.4.29/net/core/dev.c 2004-04-14 06:05:41.000000000 -0700 +++ linux-2.4.29.amds/net/core/dev.c 2005-04-01 11:27:15.000000000 -0800 @@ -1,4 +1,4 @@ -/* +/* -*-linux-c-*- * NET3 Protocol independent device support routines. * * This program is free software; you can redistribute it and/or @@ -82,6 +82,7 @@ #include #include #include +#include #include #include #include @@ -109,6 +110,11 @@ #endif +#if defined(CONFIG_NET_PKTGEN) || defined(CONFIG_NET_PKTGEN_MODULE) +#include "pktgen.h" +#endif + + /* This define, if set, will randomly drop a packet when congestion * is more than moderate. It helps fairness in the multi-interface * case when one of them is a hog, but it kills performance for the @@ -1131,7 +1137,7 @@ =======================================================================*/ int netdev_max_backlog = 300; -int weight_p = 64; /* old backlog weight */ +int weight_p = 64; /* old backlog weight */ /* These numbers are selected based on intuition and some * experimentatiom, if you have more scientific way of doing this * please go ahead and fix things. @@ -1425,6 +1431,19 @@ } +#if defined(CONFIG_NET_PKTGEN) || defined(CONFIG_NET_PKTGEN_MODULE) +#warning "Compiling dev.c for pktgen."; + +int (*handle_pktgen_hook)(struct sk_buff *skb) = NULL; + +static __inline__ int handle_pktgen_rcv(struct sk_buff* skb) { + if (handle_pktgen_hook) { + return handle_pktgen_hook(skb); + } + return -1; +} +#endif + #if defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE) void (*br_handle_frame_hook)(struct sk_buff *skb) = NULL; #endif @@ -1447,6 +1466,20 @@ return ret; } +#if defined(CONFIG_MACVLAN) || defined(CONFIG_MACVLAN_MODULE) +/* Returns >= 0 if we consume the packet. Otherwise, let + * it fall through the rest of the packet processing. + */ +int (*macvlan_handle_frame_hook)(struct sk_buff *skb) = NULL; +#endif + +/* Returns >= 0 if we consume the packet. Otherwise, let + * it fall through the rest of the packet processing. + */ +static __inline__ int handle_macvlan(struct sk_buff *skb) +{ + return macvlan_handle_frame_hook(skb); +} #ifdef CONFIG_NET_DIVERT static inline int handle_diverter(struct sk_buff *skb) @@ -1499,7 +1532,8 @@ if (skb->dev->divert && skb->dev->divert->divert) ret = handle_diverter(skb); #endif /* CONFIG_NET_DIVERT */ - + + #if defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE) if (skb->dev->br_port != NULL && br_handle_frame_hook != NULL && skb->pkt_type != PACKET_LOOPBACK) { @@ -1507,6 +1541,32 @@ } #endif +#if defined(CONFIG_MACVLAN) || defined(CONFIG_MACVLAN_MODULE) + if (skb->dev->macvlan_priv != NULL && + macvlan_handle_frame_hook != NULL) { + if (handle_macvlan(skb) >= 0) { + /* consumed by mac-vlan...it would have been + * re-sent to this method with a different + * device... + */ + return 0; + } + else { + /* Let it fall through and be processed normally */ + } + } +#endif + +#if defined(CONFIG_NET_PKTGEN) || defined(CONFIG_NET_PKTGEN_MODULE) + if ((skb->dev->priv_flags & IFF_PKTGEN_RCV) && + (handle_pktgen_rcv(skb) >= 0)) { + /* Pktgen may consume the packet, no need to send + * to further protocols. + */ + return 0; + } +#endif + type = skb->protocol; for (ptype=ptype_base[ntohs(type)&15];ptype;ptype=ptype->next) { if (ptype->type == type && @@ -1622,20 +1682,45 @@ local_irq_enable(); dev = list_entry(queue->poll_list.next, struct net_device, poll_list); - +#define ORIGINAL_NAPI_ALGORITHM +#ifdef ORIGINAL_NAPI_ALGORITHM if (dev->quota <= 0 || dev->poll(dev, &budget)) { local_irq_disable(); list_del(&dev->poll_list); list_add_tail(&dev->poll_list, &queue->poll_list); if (dev->quota < 0) - dev->quota += dev->weight; - else - dev->quota = dev->weight; + dev->quota += dev->weight; + else + dev->quota = dev->weight; } else { dev_put(dev); local_irq_disable(); } - } +#else + /* This scheme should allow devices to build up 2x their weight in quota + * credit. Heavy users will only get their normal quota. This should + * help let bursty traffic get higher priority. --Ben + */ + if (dev->poll(dev, &budget)) { + /* More to do, put these guys back on the poll list */ + local_irq_disable(); + list_del(&dev->poll_list); + list_add_tail(&dev->poll_list, &queue->poll_list); + dev->quota = dev->weight; + } + else { + /* These guys are done, they come off of the poll list */ + if (dev->quota >= dev->weight) { + dev->quota = (dev->weight << 1); /* max quota of 2x weight */ + } + else { + dev->quota += dev->weight; + } + dev_put(dev); + local_irq_disable(); + } +#endif + } local_irq_enable(); br_read_unlock(BR_NETPROTO_LOCK); @@ -2187,6 +2272,34 @@ notifier_call_chain(&netdev_chain, NETDEV_CHANGENAME, dev); return 0; + case SIOCSIFWEIGHT: + if (ifr->ifr_qlen < 0) + return -EINVAL; + dev->weight = ifr->ifr_qlen; + return 0; + + case SIOCGIFWEIGHT: + ifr->ifr_qlen = dev->weight; + return 0; + + case SIOCSACCEPTLOCALADDRS: + if (ifr->ifr_flags) { + dev->priv_flags |= IFF_ACCEPT_LOCAL_ADDRS; + } + else { + dev->priv_flags &= ~IFF_ACCEPT_LOCAL_ADDRS; + } + return 0; + + case SIOCGACCEPTLOCALADDRS: + if (dev->priv_flags & IFF_ACCEPT_LOCAL_ADDRS) { + ifr->ifr_flags = 1; + } + else { + ifr->ifr_flags = 0; + } + return 0; + /* * Unknown or private ioctl */ @@ -2283,6 +2396,8 @@ case SIOCGIFMAP: case SIOCGIFINDEX: case SIOCGIFTXQLEN: + case SIOCGIFWEIGHT: + case SIOCGACCEPTLOCALADDRS: dev_load(ifr.ifr_name); read_lock(&dev_base_lock); ret = dev_ifsioc(&ifr, cmd); @@ -2359,6 +2474,8 @@ case SIOCBONDSLAVEINFOQUERY: case SIOCBONDINFOQUERY: case SIOCBONDCHANGEACTIVE: + case SIOCSIFWEIGHT: + case SIOCSACCEPTLOCALADDRS: if (!capable(CAP_NET_ADMIN)) return -EPERM; dev_load(ifr.ifr_name); --- linux-2.4.29/net/core/pktgen.c 2003-11-28 10:26:21.000000000 -0800 +++ linux-2.4.29.amds/net/core/pktgen.c 2005-04-01 11:27:15.000000000 -0800 @@ -1,9 +1,8 @@ /* -*-linux-c-*- - * $Id: pktgen.c,v 1.8 2002/07/15 19:30:17 robert Exp $ - * pktgen.c: Packet Generator for performance evaluation. * * Copyright 2001, 2002 by Robert Olsson * Uppsala University, Sweden + * 2002 Ben Greear * * A tool for loading the network with preconfigurated packets. * The tool is implemented as a linux module. Parameters are output @@ -21,34 +20,47 @@ * Added multiskb option 020301 --DaveM * Scaling of results. 020417--sigurdur@linpro.no * Significant re-work of the module: - * * Updated to support generation over multiple interfaces at once - * by creating 32 /proc/net/pg* files. Each file can be manipulated - * individually. + * * Convert to threaded model to more efficiently be able to transmit + * and receive on multiple interfaces at once. * * Converted many counters to __u64 to allow longer runs. * * Allow configuration of ranges, like min/max IP address, MACs, * and UDP-ports, for both source and destination, and can * set to use a random distribution or sequentially walk the range. - * * Can now change some values after starting. + * * Can now change most values after starting. * * Place 12-byte packet in UDP payload with magic number, - * sequence number, and timestamp. Will write receiver next. - * * The new changes seem to have a performance impact of around 1%, - * as far as I can tell. + * sequence number, and timestamp. + * * Add receiver code that detects dropped pkts, re-ordered pkts, and + * latencies (with micro-second) precision. + * * Add IOCTL interface to easily get counters & configuration. * --Ben Greear + + * Fix refcount off by one if first packet fails, potential null deref, + * memleak 030710- KJP + * + * * Added the IPMAC option to allow the MAC addresses to mirror IP addresses. + * -- (dhetheri) Dave Hetherington 03/09/29 + * * Allow the user to change the protocol field via 'pgset "prot 0"' command + * -- (dhetheri) Dave Hetherington 03/10/7 + * Integrated to 2.5.x 021029 --Lucio Maciel (luciomaciel@zipmail.com.br) + * * * Renamed multiskb to clone_skb and cleaned up sending core for two distinct * skb modes. A clone_skb=0 mode for Ben "ranges" work and a clone_skb != 0 * as a "fastpath" with a configurable number of clones after alloc's. - * * clone_skb=0 means all packets are allocated this also means ranges time * stamps etc can be used. clone_skb=100 means 1 malloc is followed by 100 * clones. * * Also moved to /proc/net/pktgen/ - * --ro + * --ro * - * Fix refcount off by one if first packet fails, potential null deref, + * Fix refcount off by one if first packet fails, potential null deref, * memleak 030710- KJP * + * + * Sept 10: Fixed threading/locking. Lots of bone-headed and more clever + * mistakes. Also merged in DaveM's patch in the -pre6 patch. + * * See Documentation/networking/pktgen.txt for how to use this. */ @@ -84,170 +96,715 @@ #include #include -#define cycles() ((u32)get_cycles()) +#include /* for lock kernel */ +#include /* do_div */ + +#include +#if defined(CONFIG_MACVLAN) || defined(CONFIG_MACVLAN_MODULE) +#include "../macvlan/macvlan.h" +#endif -#define VERSION "pktgen version 1.3" +#include "pktgen.h" + + +#define VERSION "pktgen version 1.9.2 (nospin)" static char version[] __initdata = - "pktgen.c: v1.3: Packet Generator for packet performance testing.\n"; + "pktgen.c: v1.9.2 (nospin): Packet Generator for packet performance testing.\n"; /* Used to help with determining the pkts on receive */ #define PKTGEN_MAGIC 0xbe9be955 +/* #define PG_DEBUG(a) a */ +#define PG_DEBUG(a) /* a */ -/* Keep information per interface */ -struct pktgen_info { - /* Parameters */ +/* cycles per micro-second */ +static u32 pg_cycles_per_ns; +static u32 pg_cycles_per_us; +static u32 pg_cycles_per_ms; - /* If min != max, then we will either do a linear iteration, or - * we will do a random selection from within the range. - */ - __u32 flags; +/* Module parameters, defaults. */ +static int pg_count_d = 0; /* run forever by default */ +static int pg_ipg_d = 0; +static int pg_multiskb_d = 0; +static int pg_thread_count = 1; /* Initial threads to create */ +static int debug = 0; -#define F_IPSRC_RND (1<<0) /* IP-Src Random */ -#define F_IPDST_RND (1<<1) /* IP-Dst Random */ -#define F_UDPSRC_RND (1<<2) /* UDP-Src Random */ -#define F_UDPDST_RND (1<<3) /* UDP-Dst Random */ -#define F_MACSRC_RND (1<<4) /* MAC-Src Random */ -#define F_MACDST_RND (1<<5) /* MAC-Dst Random */ -#define F_SET_SRCMAC (1<<6) /* Specify-Src-Mac - (default is to use Interface's MAC Addr) */ -#define F_SET_SRCIP (1<<7) /* Specify-Src-IP - (default is to use Interface's IP Addr) */ - - - int pkt_size; /* = ETH_ZLEN; */ - int nfrags; - __u32 ipg; /* Default Interpacket gap in nsec */ - __u64 count; /* Default No packets to send */ - __u64 sofar; /* How many pkts we've sent so far */ - __u64 errors; /* Errors when trying to transmit, pkts will be re-sent */ - struct timeval started_at; - struct timeval stopped_at; - __u64 idle_acc; - __u32 seq_num; - - int clone_skb; /* Use multiple SKBs during packet gen. If this number - * is greater than 1, then that many coppies of the same - * packet will be sent before a new packet is allocated. - * For instance, if you want to send 1024 identical packets - * before creating a new packet, set clone_skb to 1024. - */ - int busy; - int do_run_run; /* if this changes to false, the test will stop */ - - char outdev[32]; - char dst_min[32]; - char dst_max[32]; - char src_min[32]; - char src_max[32]; - /* If we're doing ranges, random or incremental, then this - * defines the min/max for those ranges. - */ - __u32 saddr_min; /* inclusive, source IP address */ - __u32 saddr_max; /* exclusive, source IP address */ - __u32 daddr_min; /* inclusive, dest IP address */ - __u32 daddr_max; /* exclusive, dest IP address */ - - __u16 udp_src_min; /* inclusive, source UDP port */ - __u16 udp_src_max; /* exclusive, source UDP port */ - __u16 udp_dst_min; /* inclusive, dest UDP port */ - __u16 udp_dst_max; /* exclusive, dest UDP port */ - - __u32 src_mac_count; /* How many MACs to iterate through */ - __u32 dst_mac_count; /* How many MACs to iterate through */ - - unsigned char dst_mac[6]; - unsigned char src_mac[6]; - - __u32 cur_dst_mac_offset; - __u32 cur_src_mac_offset; - __u32 cur_saddr; - __u32 cur_daddr; - __u16 cur_udp_dst; - __u16 cur_udp_src; - - __u8 hh[14]; - /* = { - 0x00, 0x80, 0xC8, 0x79, 0xB3, 0xCB, - - We fill in SRC address later - 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, - 0x08, 0x00 - }; - */ - __u16 pad; /* pad out the hh struct to an even 16 bytes */ - char result[512]; - /* proc file names */ - char fname[80]; - char busy_fname[80]; - - struct proc_dir_entry *proc_ent; - struct proc_dir_entry *busy_proc_ent; -}; +/* List of all running threads */ +static struct pktgen_thread_info* pktgen_threads = NULL; +spinlock_t _pg_threadlist_lock = SPIN_LOCK_UNLOCKED; + +/* Holds interfaces for all threads */ +#define PG_INFO_HASH_MAX 32 +static struct pktgen_interface_info* pg_info_hash[PG_INFO_HASH_MAX]; +spinlock_t _pg_hash_lock = SPIN_LOCK_UNLOCKED; -struct pktgen_hdr { - __u32 pgh_magic; - __u32 seq_num; - struct timeval timestamp; +#define PG_PROC_DIR "pktgen" +static struct proc_dir_entry *pg_proc_dir = NULL; + +char module_fname[128]; +struct proc_dir_entry *module_proc_ent = NULL; + + +static void init_pktgen_kthread(struct pktgen_thread_info *kthread, char *name); +static int pg_rem_interface_info(struct pktgen_thread_info* pg_thread, + struct pktgen_interface_info* i); +static int pg_add_interface_info(struct pktgen_thread_info* pg_thread, + const char* ifname); +static void exit_pktgen_kthread(struct pktgen_thread_info *kthread); +static void stop_pktgen_kthread(struct pktgen_thread_info *kthread); +static struct pktgen_thread_info* pg_find_thread(const char* name); +static int pg_add_thread_info(const char* name); +static struct pktgen_interface_info* pg_find_interface(struct pktgen_thread_info* pg_thread, + const char* ifname); +static int pktgen_device_event(struct notifier_block *, unsigned long, void *); + + +struct notifier_block pktgen_notifier_block = { + notifier_call: pktgen_device_event, }; -static int cpu_speed; -static int debug; +/* This code works around the fact that do_div cannot handle two 64-bit + numbers, and regular 64-bit division doesn't work on x86 kernels. + --Ben +*/ -/* Module parameters, defaults. */ -static int count_d = 100000; -static int ipg_d = 0; -static int clone_skb_d = 0; +#define PG_DIV 0 +#define PG_REM 1 + +/* This was emailed to LMKL by: Chris Caputo + * Function copied/adapted/optimized from: + * + * nemesis.sourceforge.net/browse/lib/static/intmath/ix86/intmath.c.html + * + * Copyright 1994, University of Cambridge Computer Laboratory + * All Rights Reserved. + * + * TODO: When running on a 64-bit CPU platform, this should no longer be + * TODO: necessary. + */ +inline static s64 divremdi3(s64 x, s64 y, int type) { + u64 a = (x < 0) ? -x : x; + u64 b = (y < 0) ? -y : y; + u64 res = 0, d = 1; + + if (b > 0) { + while (b < a) { + b <<= 1; + d <<= 1; + } + } + + do { + if ( a >= b ) { + a -= b; + res += d; + } + b >>= 1; + d >>= 1; + } + while (d); + + if (PG_DIV == type) { + return (((x ^ y) & (1ll<<63)) == 0) ? res : -(s64)res; + } + else { + return ((x & (1ll<<63)) == 0) ? a : -(s64)a; + } +}/* divremdi3 */ + +/* End of hacks to deal with 64-bit math on x86 */ + + + +inline static void pg_lock_thread_list(const char* msg) { + if (debug > 1) { + printk("before pg_lock_thread_list, msg: %s\n", msg); + } + spin_lock(&_pg_threadlist_lock); + if (debug > 1) { + printk("after pg_lock_thread_list, msg: %s\n", msg); + } +} + +inline static void pg_unlock_thread_list(const char* msg) { + if (debug > 1) { + printk("before pg_unlock_thread_list, msg: %s\n", msg); + } + spin_unlock(&_pg_threadlist_lock); + if (debug > 1) { + printk("after pg_unlock_thread_list, msg: %s\n", msg); + } +} +inline static void pg_lock_hash(const char* msg) { + if (debug > 1) { + printk("before pg_lock_hash, msg: %s\n", msg); + } + spin_lock(&_pg_hash_lock); + if (debug > 1) { + printk("before pg_lock_hash, msg: %s\n", msg); + } +} + +inline static void pg_unlock_hash(const char* msg) { + if (debug > 1) { + printk("before pg_unlock_hash, msg: %s\n", msg); + } + spin_unlock(&_pg_hash_lock); + if (debug > 1) { + printk("after pg_unlock_hash, msg: %s\n", msg); + } +} -#define MAX_PKTGEN 8 -static struct pktgen_info pginfos[MAX_PKTGEN]; +inline static void pg_lock(struct pktgen_thread_info* pg_thread, const char* msg) { + if (debug > 1) { + printk("before pg_lock thread, msg: %s\n", msg); + } + spin_lock(&(pg_thread->pg_threadlock)); + if (debug > 1) { + printk("after pg_lock thread, msg: %s\n", msg); + } +} +inline static void pg_unlock(struct pktgen_thread_info* pg_thread, const char* msg) { + if (debug > 1) { + printk("before pg_unlock thread, thread: %p msg: %s\n", + pg_thread, msg); + } + spin_unlock(&(pg_thread->pg_threadlock)); + if (debug > 1) { + printk("after pg_unlock thread, thread: %p msg: %s\n", + pg_thread, msg); + } +} /** Convert to miliseconds */ -inline __u64 tv_to_ms(const struct timeval* tv) { +static inline __u64 tv_to_ms(const struct timeval* tv) { __u64 ms = tv->tv_usec / 1000; ms += (__u64)tv->tv_sec * (__u64)1000; return ms; } -inline __u64 getCurMs(void) { + +/** Convert to micro-seconds */ +static inline __u64 tv_to_us(const struct timeval* tv) { + __u64 us = tv->tv_usec; + us += (__u64)tv->tv_sec * (__u64)1000000; + return us; +} + + +static inline __u64 pg_div(__u64 n, __u32 base) { + __u64 tmp = n; + do_div(tmp, base); + /* printk("pg_div, n: %llu base: %d rv: %llu\n", + n, base, tmp); */ + return tmp; +} + +/* Fast, not horribly accurate, since the machine started. */ +static inline __u64 getRelativeCurMs(void) { + return pg_div(get_cycles(), pg_cycles_per_ms); +} + +/* Since the epoc. More precise over long periods of time than + * getRelativeCurMs + */ +static inline __u64 getCurMs(void) { struct timeval tv; do_gettimeofday(&tv); return tv_to_ms(&tv); } -#define PG_PROC_DIR "pktgen" -static struct proc_dir_entry *proc_dir = 0; +/* Since the epoc. More precise over long periods of time than + * getRelativeCurMs + */ +static inline __u64 getCurUs(void) { + struct timeval tv; + do_gettimeofday(&tv); + return tv_to_us(&tv); +} -static struct net_device *setup_inject(struct pktgen_info* info) -{ - struct net_device *odev; +/* Since the machine booted. */ +static inline __u64 getRelativeCurUs(void) { + return pg_div(get_cycles(), pg_cycles_per_us); +} - rtnl_lock(); - odev = __dev_get_by_name(info->outdev); - if (!odev) { - sprintf(info->result, "No such netdevice: \"%s\"", info->outdev); - goto out_unlock; +/* Since the machine booted. */ +static inline __u64 getRelativeCurNs(void) { + return pg_div(get_cycles(), pg_cycles_per_ns); +} + +static inline __u64 tv_diff(const struct timeval* a, const struct timeval* b) { + return tv_to_us(a) - tv_to_us(b); +} + + + +int pktgen_proc_ioctl(struct inode* inode, struct file* file, unsigned int cmd, + unsigned long arg) { + int err = 0; + struct pktgen_ioctl_info args; + struct pktgen_thread_info* targ = NULL; + + /* + if (!capable(CAP_NET_ADMIN)){ + return -EPERM; + } + */ + + if (copy_from_user(&args, (void*)arg, sizeof(args))) { + return -EFAULT; + } + + /* Null terminate the names */ + args.thread_name[31] = 0; + args.interface_name[31] = 0; + + /* printk("pktgen: thread_name: %s interface_name: %s\n", + * args.thread_name, args.interface_name); + */ + + switch (cmd) { + case GET_PKTGEN_INTERFACE_INFO: { + targ = pg_find_thread(args.thread_name); + if (targ) { + struct pktgen_interface_info* info; + info = pg_find_interface(targ, args.interface_name); + if (info) { + memcpy(&(args.info), info, sizeof(args.info)); + if (copy_to_user((void*)(arg), &args, sizeof(args))) { + printk("ERROR: pktgen: copy_to_user failed.\n"); + err = -EFAULT; + } + else { + err = 0; + } + } + else { + /* printk("ERROR: pktgen: Could not find interface -:%s:-\n", + args.interface_name);*/ + err = -ENODEV; + } + } + else { + printk("ERROR: pktgen: Could not find thread -:%s:-.\n", + args.thread_name); + err = -ENODEV; + } + break; + } + default: + /* pass on to underlying device instead?? */ + printk("%s: Unknown pktgen IOCTL: %x \n", __FUNCTION__, + cmd); + return -EINVAL; + } + + return err; +}/* pktgen_proc_ioctl */ + +static struct file_operations pktgen_fops = { + ioctl: pktgen_proc_ioctl, +}; + +/* Runs from interrupt */ +int pg_notify_queue_woken(struct net_device* dev) { + struct pktgen_thread_info* pg_thread = dev->nqw_data; + /* printk("pg_nqw, pg_thread: %p\n", pg_thread); */ + if (pg_thread && pg_thread->sleeping) { + if (getRelativeCurNs() > (pg_thread->next_tx_ns - 1000)) { + /* See if we should wake up the thread, wake + * slightly early (1000 ns) + */ + pg_thread->sleeping = 0; + wake_up_interruptible(&(pg_thread->queue)); + } } + return 0; +} + + +static void set_nqw_hook(struct pktgen_thread_info* pg_thread, + struct pktgen_interface_info* info, + struct net_device* dev) { + /* The notify-queue-woken magic only works for physical + * devices at this time. So, apply hook to underlying + * device. + */ + struct pktgen_thread_info* pg_thread_nd; + /* printk("In set_nqw_hook, dev: %s\n", dev->name); */ + +#if defined(CONFIG_MACVLAN) || defined(CONFIG_MACVLAN_MODULE) + if (dev->priv_flags & IFF_MAC_VLAN) { + struct macvlan_vlan *vlan = dev->priv; + set_nqw_hook(pg_thread, info, vlan->lowerdev); + return; + } +#endif + + if (dev->priv_flags & IFF_802_1Q_VLAN) { + set_nqw_hook(pg_thread, info, VLAN_DEV_INFO(dev)->real_dev); + return; + } + + pg_thread_nd = (struct pktgen_thread_info*)(dev->nqw_data); + + /*printk("pg_thread_nd: %p info: %p\n", pg_thread_nd, info);*/ + + if (pg_thread_nd) { + /* Just bump a reference count, it doesn't really matter which + * info is there since we always look at the parent thread anyway + */ + atomic_inc(&(pg_thread_nd->nqw_ref_count)); + + /*printk("Incremented nqw_ref_count: %d device: %s\n", + (int)(atomic_read(&(pg_thread_nd->nqw_ref_count))), dev->name); */ + } + else { + /* Must be a real device, or at least some un-handled fake device. Can't + * hurt to add it here at any rate. + */ + atomic_inc(&(pg_thread->nqw_ref_count)); + dev->nqw_data = pg_thread; + dev->notify_queue_woken = pg_notify_queue_woken; + /*printk("Added nqw callback to device: %s, data: %p data_nd: %p\n", + dev->name, dev->nqw_data, pg_thread_nd);*/ + } +}//set_nqw_hook + + +static void clear_nqw_hook(struct pktgen_interface_info* info, + struct net_device* dev) { + /* The notify-queue-woken magic only works for physical + * devices at this time. So, apply hook to underlying + * device. + */ + struct pktgen_thread_info* pg_thread; + /* printk("In clear_nqw_hook, dev: %s\n", dev->name); */ + +#if defined(CONFIG_MACVLAN) || defined(CONFIG_MACVLAN_MODULE) + if (dev->priv_flags & IFF_MAC_VLAN) { + struct macvlan_vlan *vlan = dev->priv; + clear_nqw_hook(info, vlan->lowerdev); + return; + } +#endif + + if (dev->priv_flags & IFF_802_1Q_VLAN) { + clear_nqw_hook(info, VLAN_DEV_INFO(dev)->real_dev); + return; + } + + if (dev->nqw_data) { + pg_thread = (struct pktgen_thread_info*)(dev->nqw_data); + atomic_dec(&(pg_thread->nqw_ref_count)); + + /* printk("Decremented nqw_ref_count: %d device: %s\n", + (int)(atomic_read(&(pg_thread->nqw_ref_count))), dev->name); */ + + BUG_ON(atomic_read(&(pg_thread->nqw_ref_count)) < 0); + + if (atomic_read(&(pg_thread->nqw_ref_count)) == 0) { + /*printk("Removing nqw reference from device: %s\n", dev->name);*/ + dev->notify_queue_woken = NULL; + dev->nqw_data = NULL; + + /* Clear from all other devices too. There is an issue/bug where + * we may at times have more actual references to the pg_thread than + * our reference counter shows. In practice, I do not believe this + * will be a problem, but it will be slightly inefficient. A true + * fix would involve some sort of hash/list/map of physical interfaces + * to reference counts, stored in the pg_thread class. + * --Ben + */ + + { + int i; + struct pktgen_interface_info* p; + for (i = 0; iodev && p->odev->nqw_data == pg_thread) { + /*printk("Removing nqw reference from dev: %s\n", + p->odev->name);*/ + p->odev->notify_queue_woken = NULL; + p->odev->nqw_data = NULL; + } + p = p->next; + } + } + } + } + else { + /* Else, if we are directly attached to this device, remove the + * nqw callback then too, because we only allow one info struct + * per net-device, so this MUST be the only one. + */ + if (dev == info->odev) { + /*printk("Removing nqw reference from real device: %s\n", + dev->name);*/ + dev->notify_queue_woken = NULL; + dev->nqw_data = NULL; + } + else { + /*printk("pktgen NOTE: dev: %s != info->odev: %s\n", + dev ? dev->name : "NULL", + info->odev ? info->odev->name : "NULL");*/ + } + } + } + else { + printk("pktgen: Warning: nqw_data is null in clear_nqw_hook, dev: %s\n", + dev->name); + } +}//clear_nqw_hook + + +static void remove_pg_info_from_hash(struct pktgen_interface_info* info) { + pg_lock_hash(__FUNCTION__); + { + int device_idx = info->odev ? info->odev->ifindex : 0; + int b = device_idx % PG_INFO_HASH_MAX; + struct pktgen_interface_info* p = pg_info_hash[b]; + struct pktgen_interface_info* prev = pg_info_hash[b]; + int found_one = 0; + + PG_DEBUG(printk("remove_pg_info_from_hash, p: %p info: %p device_idx: %i\n", + p, info, device_idx)); - if (odev->type != ARPHRD_ETHER) { - sprintf(info->result, "Not ethernet device: \"%s\"", info->outdev); - goto out_unlock; + if (p != NULL) { + + if (p == info) { + pg_info_hash[b] = p->next_hash; + p->next_hash = NULL; + found_one = 1; + } + else { + while (prev->next_hash) { + p = prev->next_hash; + if (p == info) { + prev->next_hash = p->next_hash; + p->next_hash = NULL; + found_one = 1; + break; + } + prev = p; + } + } + } + + if (found_one) { + if (info->odev) { + info->odev->priv_flags &= ~(IFF_PKTGEN_RCV); + clear_nqw_hook(info, info->odev); + } + } + } + pg_unlock_hash(__FUNCTION__); +}/* remove_pg_info_from_hash */ + + +static void add_pg_info_to_hash(struct pktgen_thread_info* pg_thread, + struct pktgen_interface_info* info) { + /* First remove it, just in case it's already there. */ + remove_pg_info_from_hash(info); + + pg_lock_hash(__FUNCTION__); + { + int device_idx = info->odev ? info->odev->ifindex : 0; + int b = device_idx % PG_INFO_HASH_MAX; + + PG_DEBUG(printk("add_pg_info_from_hash, b: %i info: %p device_idx: %i\n", + b, info, device_idx)); + + info->next_hash = pg_info_hash[b]; + pg_info_hash[b] = info; + + + if (info->odev) { + set_nqw_hook(pg_thread, info, info->odev); + info->odev->priv_flags |= (IFF_PKTGEN_RCV); + } + } + pg_unlock_hash(__FUNCTION__); +}/* add_pg_info_to_hash */ + + +/* Find the pktgen_interface_info for a device idx */ +struct pktgen_interface_info* find_pg_info(int device_idx) { + struct pktgen_interface_info* p = NULL; + if (debug > 1) { + printk("in find_pg_info...\n"); + } + pg_lock_hash(__FUNCTION__); + { + int b = device_idx % PG_INFO_HASH_MAX; + p = pg_info_hash[b]; + while (p) { + if (p->odev && (p->odev->ifindex == device_idx)) { + break; + } + p = p->next_hash; + } + } + pg_unlock_hash(__FUNCTION__); + return p; +} + + +/* Remove an interface from our hash, dissassociate pktgen_interface_info + * from interface + */ +static void check_remove_device(struct pktgen_interface_info* info) { + struct pktgen_interface_info* pi = NULL; + if (info->odev) { + pi = find_pg_info(info->odev->ifindex); + if (pi != info) { + printk("ERROR: pi != info, pi: %p info: %p\n", pi, info); + } + else { + /* Remove info from our hash */ + remove_pg_info_from_hash(info); + } + + /* TODO: Wonder if we need locking here? Had rtnl_lock, but + * that can sleep and this is called with irqs disabled... + */ + info->odev->priv_flags &= ~(IFF_PKTGEN_RCV); + atomic_dec(&(info->odev->refcnt)); + info->odev = NULL; + } +}/* check_remove_device */ + + +static int pg_remove_interface_from_all_threads(const char* dev_name) { + int cnt = 0; + pg_lock_thread_list(__FUNCTION__); + { + struct pktgen_thread_info* tmp = pktgen_threads; + struct pktgen_interface_info* info = NULL; + + while (tmp) { + info = pg_find_interface(tmp, dev_name); + if (info) { + printk("pktgen: Removing interface: %s from pktgen control.\n", + dev_name); + pg_rem_interface_info(tmp, info); + cnt++; + } + else { + /* printk("pktgen: Could not find interface: %s in rem_from_all.\n", + dev_name); */ + } + tmp = tmp->next; + } + } + pg_unlock_thread_list(__FUNCTION__); + return cnt; +}/* pg_rem_interface_from_all_threads */ + + +static int pktgen_device_event(struct notifier_block *unused, unsigned long event, void *ptr) { + struct net_device *dev = (struct net_device *)(ptr); + + /* It is OK that we do not hold the group lock right now, + * as we run under the RTNL lock. + */ + + switch (event) { + case NETDEV_CHANGEADDR: + case NETDEV_GOING_DOWN: + case NETDEV_DOWN: + case NETDEV_UP: + /* Ignore for now */ + break; + + case NETDEV_UNREGISTER: + pg_remove_interface_from_all_threads(dev->name); + break; + }; + + return NOTIFY_DONE; +} + + +/* Associate pktgen_interface_info with a device. + */ +static struct net_device* pg_setup_interface(struct pktgen_thread_info* pg_thread, + struct pktgen_interface_info* info) { + struct net_device *odev; + int keep_it = 0; + + check_remove_device(info); + + odev = dev_get_by_name(info->ifname); + if (!odev) { + printk("No such netdevice: \"%s\"\n", info->ifname); } + else if (odev->type != ARPHRD_ETHER) { + printk("Not an ethernet device: \"%s\"\n", info->ifname); + } + else if (!netif_running(odev)) { + printk("Device is down: \"%s\"\n", info->ifname); + } + else if (odev->priv_flags & IFF_PKTGEN_RCV) { + printk("ERROR: Device: \"%s\" is already assigned to a pktgen interface.\n", + info->ifname); + } + else { + info->odev = odev; + info->odev->priv_flags |= (IFF_PKTGEN_RCV); + + /* Can't use multi-skb > 0 with virtual interfaces, because they change + * the skb->dev pointer (at least) and so it's really impossible to send + * the exact same pkt over and over again + */ + if ((odev->priv_flags & IFF_MAC_VLAN) || + (odev->priv_flags & IFF_802_1Q_VLAN)) { + if (info->multiskb > 0) { + printk("pktgen: WARNING: Cannot use multi-skb > 0 on virtual interfaces, setting to zero.\n"); + info->multiskb = 0; + } + } + + keep_it = 1; + } + + if (info->odev) { + add_pg_info_to_hash(pg_thread, info); + } - if (!netif_running(odev)) { - sprintf(info->result, "Device is down: \"%s\"", info->outdev); - goto out_unlock; + if ((!keep_it) && odev) { + dev_put(odev); } + + return info->odev; +}/* pg_setup_interface */ +/* Read info from the interface and set up internal pktgen_interface_info + * structure to have the right information to create/send packets + */ +static void pg_setup_inject(struct pktgen_interface_info* info) +{ + if (!info->odev) { + /* Try once more, just in case it works now. */ + pg_setup_interface(info->pg_thread, info); + } + + if (!info->odev) { + printk("ERROR: info->odev == NULL in setup_inject.\n"); + sprintf(info->result, "ERROR: info->odev == NULL in setup_inject.\n"); + return; + } + /* Default to the interface's mac if not explicitly set. */ if (!(info->flags & F_SET_SRCMAC)) { - memcpy(&(info->hh[6]), odev->dev_addr, 6); + memcpy(&(info->hh[6]), info->odev->dev_addr, 6); } else { memcpy(&(info->hh[6]), info->src_mac, 6); @@ -255,12 +812,15 @@ /* Set up Dest MAC */ memcpy(&(info->hh[0]), info->dst_mac, 6); + + /* Set up pkt size */ + info->cur_pkt_size = info->min_pkt_size; info->saddr_min = 0; info->saddr_max = 0; if (strlen(info->src_min) == 0) { - if (odev->ip_ptr) { - struct in_device *in_dev = odev->ip_ptr; + if (info->odev->ip_ptr) { + struct in_device *in_dev = info->odev->ip_ptr; if (in_dev->ifa_list) { info->saddr_min = in_dev->ifa_list->ifa_address; @@ -283,65 +843,94 @@ info->cur_daddr = info->daddr_min; info->cur_udp_dst = info->udp_dst_min; info->cur_udp_src = info->udp_src_min; - - atomic_inc(&odev->refcnt); - rtnl_unlock(); - - return odev; - -out_unlock: - rtnl_unlock(); - return NULL; } -static void nanospin(int ipg, struct pktgen_info* info) +/* delay_ns is in nano-seconds */ +static void pg_nanodelay(int delay_ns, struct pktgen_interface_info* info, + struct pktgen_thread_info* pg_thread) { - u32 idle_start, idle; - - idle_start = cycles(); + u64 idle_start = getRelativeCurNs(); + u64 last_time; + u64 itmp = idle_start; + info->nanodelays++; + info->accum_delay_ns += delay_ns; + while (info->accum_delay_ns > PG_MAX_ACCUM_DELAY_NS) { + info->sleeps++; + pg_thread->sleeping = 1; + interruptible_sleep_on_timeout(&(pg_thread->queue), 1); + pg_thread->sleeping = 0; + /* will wake after one tick */ + last_time = itmp; + itmp = getRelativeCurNs(); + info->accum_delay_ns -= (itmp - last_time); + info->idle_acc += (itmp - last_time); - for (;;) { - barrier(); - idle = cycles() - idle_start; - if (idle * 1000 >= ipg * cpu_speed) + if (!info->do_run_run) { break; - } - info->idle_acc += idle; -} + } + }/* while */ +}//pg_nanodelay + +/* Returns: cycles per micro-second */ static int calc_mhz(void) { struct timeval start, stop; - u32 start_s, elapsed; - + u64 start_s; + u64 t1, t2; + u32 elapsed; + u32 clock_time = 0; + do_gettimeofday(&start); - start_s = cycles(); + start_s = get_cycles(); + /* Spin for 50,000,000 cycles */ do { barrier(); - elapsed = cycles() - start_s; + elapsed = (u32)(get_cycles() - start_s); if (elapsed == 0) return 0; - } while (elapsed < 1000 * 50000); + } while (elapsed < 50000000); do_gettimeofday(&stop); - return elapsed/(stop.tv_usec-start.tv_usec+1000000*(stop.tv_sec-start.tv_sec)); + + t1 = tv_to_us(&start); + t2 = tv_to_us(&stop); + + clock_time = (u32)(t2 - t1); + if (clock_time == 0) { + printk("pktgen: ERROR: clock_time was zero..things may not work right, t1: %u t2: %u ...\n", + (u32)(t1), (u32)(t2)); + return 0x7FFFFFFF; + } + return elapsed / clock_time; } +/* Calibrate cycles per micro-second */ static void cycles_calibrate(void) { int i; for (i = 0; i < 3; i++) { - int res = calc_mhz(); - if (res > cpu_speed) - cpu_speed = res; + u32 res = calc_mhz(); + if (res > pg_cycles_per_us) + pg_cycles_per_us = res; } + + /* Set these up too, only need to calculate these once. */ + pg_cycles_per_ns = pg_cycles_per_us / 1000; + if (pg_cycles_per_ns == 0) { + pg_cycles_per_ns = 1; + } + pg_cycles_per_ms = pg_cycles_per_us * 1000; + + printk("pktgen: cycles_calibrate, cycles_per_ns: %d per_us: %d per_ms: %d\n", + pg_cycles_per_ns, pg_cycles_per_us, pg_cycles_per_ms); } /* Increment/randomize headers according to flags and current values * for IP src/dest, UDP src/dst port, MAC-Addr src/dst */ -static void mod_cur_headers(struct pktgen_info* info) { +static void mod_cur_headers(struct pktgen_interface_info* info) { __u32 imn; __u32 imx; @@ -431,7 +1020,7 @@ else { t = ntohl(info->cur_saddr); t++; - if (t >= imx) { + if (t > imx) { t = imn; } } @@ -446,16 +1035,54 @@ else { t = ntohl(info->cur_daddr); t++; - if (t >= imx) { + if (t > imx) { t = imn; } } info->cur_daddr = htonl(t); } + + /* dhetheri - Make MAC address = 00:00:IP address */ + if (info->flags & F_IPMAC) { + __u32 tmp; + __u32 t; + + /* SRC MAC = 00:00:IP address */ + t = ntohl(info->cur_saddr); + + tmp = info->src_mac[5] + (t & 0xFF); + info->hh[11] = tmp; + tmp = (info->src_mac[4] + ((t >> 8) & 0xFF) + (tmp >> 8)); + info->hh[10] = tmp; + tmp = (info->src_mac[3] + ((t >> 16) & 0xFF) + (tmp >> 8)); + info->hh[9] = tmp; + tmp = (info->src_mac[2] + ((t >> 24) & 0xFF) + (tmp >> 8)); + info->hh[8] = tmp; + tmp = (info->src_mac[1] + (tmp >> 8)); + info->hh[7] = tmp; + + info->cur_saddr = htonl(t); + + /* DST MAC = 00:00:IP address */ + t = ntohl(info->cur_daddr); + + tmp = info->dst_mac[5] + (t & 0xFF); + info->hh[5] = tmp; + tmp = (info->dst_mac[4] + ((t >> 8) & 0xFF) + (tmp >> 8)); + info->hh[4] = tmp; + tmp = (info->dst_mac[3] + ((t >> 16) & 0xFF) + (tmp >> 8)); + info->hh[3] = tmp; + tmp = (info->dst_mac[2] + ((t >> 24) & 0xFF) + (tmp >> 8)); + info->hh[2] = tmp; + tmp = (info->dst_mac[1] + (tmp >> 8)); + info->hh[1] = tmp; + + info->cur_daddr = htonl(t); + } /* MAC = 00:00:IP address (dhetheri) */ }/* mod_cur_headers */ -static struct sk_buff *fill_packet(struct net_device *odev, struct pktgen_info* info) +static struct sk_buff *fill_packet(struct net_device *odev, struct pktgen_interface_info* info) { struct sk_buff *skb = NULL; __u8 *eth; @@ -463,8 +1090,24 @@ int datalen, iplen; struct iphdr *iph; struct pktgen_hdr *pgh = NULL; - - skb = alloc_skb(info->pkt_size + 64 + 16, GFP_ATOMIC); + + /* dhetheri - Moved out of mod_cur_headers. */ + if (info->min_pkt_size < info->max_pkt_size) { + __u32 t; + if (info->flags & F_TXSIZE_RND) { + t = ((net_random() % (info->max_pkt_size - info->min_pkt_size)) + + info->min_pkt_size); + } + else { + t = info->cur_pkt_size + 1; + if (t > info->max_pkt_size) { + t = info->min_pkt_size; + } + } + info->cur_pkt_size = t; + } + + skb = alloc_skb(info->cur_pkt_size + 64 + 16, GFP_ATOMIC); if (!skb) { sprintf(info->result, "No memory"); return NULL; @@ -484,7 +1127,7 @@ memcpy(eth, info->hh, 14); - datalen = info->pkt_size - 14 - 20 - 8; /* Eth + IPh + UDPh */ + datalen = info->cur_pkt_size - 14 - 20 - 8; /* Eth + IPh + UDPh */ if (datalen < sizeof(struct pktgen_hdr)) { datalen = sizeof(struct pktgen_hdr); } @@ -496,9 +1139,14 @@ iph->ihl = 5; iph->version = 4; - iph->ttl = 3; + iph->ttl = 32; iph->tos = 0; - iph->protocol = IPPROTO_UDP; /* UDP */ + if (info->prot) { /* dhetheri */ + iph->protocol = info->prot; /* dhetheri */ + } + else { + iph->protocol = IPPROTO_UDP; /* UDP */ + } iph->saddr = info->cur_saddr; iph->daddr = info->cur_daddr; iph->frag_off = 0; @@ -517,7 +1165,6 @@ int frags = info->nfrags; int i; - /* TODO: Verify this is OK...it sure is ugly. --Ben */ pgh = (struct pktgen_hdr*)(((char*)(udph)) + 8); if (frags > MAX_SKB_FRAGS) @@ -565,237 +1212,911 @@ /* Stamp the time, and sequence number, convert them to network byte order */ if (pgh) { - pgh->pgh_magic = htonl(PKTGEN_MAGIC); + pgh->pgh_magic = __constant_htonl(PKTGEN_MAGIC); do_gettimeofday(&(pgh->timestamp)); pgh->timestamp.tv_usec = htonl(pgh->timestamp.tv_usec); pgh->timestamp.tv_sec = htonl(pgh->timestamp.tv_sec); pgh->seq_num = htonl(info->seq_num); } + info->seq_num++; return skb; } -static void inject(struct pktgen_info* info) -{ - struct net_device *odev = NULL; - struct sk_buff *skb = NULL; - __u64 total = 0; - __u64 idle = 0; - __u64 lcount = 0; - int nr_frags = 0; - int last_ok = 1; /* Was last skb sent? - * Or a failed transmit of some sort? This will keep - * sequence numbers in order, for example. - */ - __u64 fp = 0; - __u32 fp_tmp = 0; - - odev = setup_inject(info); - if (!odev) - return; - - info->do_run_run = 1; /* Cranke yeself! */ - info->idle_acc = 0; - info->sofar = 0; - lcount = info->count; +static void record_latency(struct pktgen_interface_info* info, int latency) { + /* NOTE: Latency can be negative */ + int div = 100; + int diff; + int vl; + int i; + info->pkts_rcvd_since_clear++; + + if (info->pkts_rcvd_since_clear < 100) { + div = info->pkts_rcvd; + if (info->pkts_rcvd_since_clear == 1) { + info->avg_latency = latency; + } + } - /* Build our initial pkt and place it as a re-try pkt. */ - skb = fill_packet(odev, info); - if (skb == NULL) goto out_reldev; + if ((div + 1) == 0) { + info->avg_latency = 0; + } + else { + info->avg_latency = ((info->avg_latency * div + latency) / (div + 1)); + } + + if (latency < info->min_latency) { + info->min_latency = latency; + } + if (latency > info->max_latency) { + info->max_latency = latency; + } + + /* Place the latency in the right 'bucket' */ + diff = (latency - info->min_latency); + for (i = 0; ilatency_bkts[i]++; + break; + } + } +}/* record latency */ - do_gettimeofday(&(info->started_at)); - while(info->do_run_run) { +/* Returns < 0 if the skb is not a pktgen buffer. */ +int pktgen_receive(struct sk_buff* skb) { + /* int i; */ /* Debugging only */ + /* unsigned char* tmp; */ + + /* dhetheri */ + //printk("pktgen receive:\n"); + //tmp=(char *)(skb->data); + //for (i=0; i<90; i++) { + // printk("%02hx ", tmp[i]); + // if (((i+1) % 15) == 0) { + // printk("\n"); + // } + //} + //printk("\n"); + /* dhetheri */ + + /* See if we have a pktgen packet */ + if ((skb->len >= (20 + 8 + sizeof(struct pktgen_hdr))) && + (skb->protocol == __constant_htons(ETH_P_IP))) { + struct pktgen_hdr* pgh; + + /* It's IP, and long enough, lets check the magic number. + * TODO: This is a hack not always guaranteed to catch the right + * packets. + */ + + /* printk("Length & protocol passed, skb->data: %p, raw: %p\n", + skb->data, skb->h.raw); */ + + pgh = (struct pktgen_hdr*)(skb->data + 20 + 8); + + /* + tmp = (char*)(skb->data); + for (i = 0; i<90; i++) { + printk("%02hx ", tmp[i]); + if (((i + 1) % 15) == 0) { + printk("\n"); + } + } + printk("\n"); + */ + + if (pgh->pgh_magic == __constant_ntohl(PKTGEN_MAGIC)) { + struct net_device* dev = skb->dev; + struct pktgen_interface_info* info = find_pg_info(dev->ifindex); + + /* Got one! */ + /* TODO: Check UDP checksum ?? */ + __u32 seq = ntohl(pgh->seq_num); + + if (!info) { + return -1; + } - /* Set a time-stamp, so build a new pkt each time */ + info->pkts_rcvd++; + info->bytes_rcvd += (skb->len + 4); /* +4 for the checksum */ + + /* Check for out-of-sequence packets */ + if (info->last_seq_rcvd == seq) { + info->dup_rcvd++; + info->dup_since_incr++; + } + else { + __s64 rx; + __s64 tx; + struct timeval txtv; + if (!skb->stamp.tv_sec) { + do_gettimeofday(&skb->stamp); + } + rx = tv_to_us(&(skb->stamp)); - if (last_ok) { - if (++fp_tmp >= info->clone_skb ) { - kfree_skb(skb); - skb = fill_packet(odev, info); - if (skb == NULL) { - goto out_reldev; + txtv.tv_usec = ntohl(pgh->timestamp.tv_usec); + txtv.tv_sec = ntohl(pgh->timestamp.tv_sec); + tx = tv_to_us(&txtv); + record_latency(info, rx - tx); + + if ((info->last_seq_rcvd + 1) == seq) { + if ((info->peer_multiskb > 1) && + (info->peer_multiskb > (info->dup_since_incr + 1))) { + + info->seq_gap_rcvd += (info->peer_multiskb - + info->dup_since_incr - 1); + } + /* Great, in order...all is well */ + } + else if (info->last_seq_rcvd < seq) { + /* sequence gap, means we dropped a pkt most likely */ + if (info->peer_multiskb > 1) { + /* We dropped more than one sequence number's worth, + * and if we're using multiskb, then this is quite + * a few. This number still will not be exact, but + * it will be closer. + */ + info->seq_gap_rcvd += (((seq - info->last_seq_rcvd) * + info->peer_multiskb) - + info->dup_since_incr); + } + else { + info->seq_gap_rcvd += (seq - info->last_seq_rcvd - 1); + } + } + else { + info->ooo_rcvd++; /* out-of-order */ } - fp++; - fp_tmp = 0; /* reset counter */ + + info->dup_since_incr = 0; } + info->last_seq_rcvd = seq; + kfree_skb(skb); + if (debug > 1) { + printk("done with pktgen_receive, free'd pkt\n"); + } + return 0; } + } + return -1; /* Let another protocol handle it, it's not for us! */ +}/* pktgen_receive */ + +static void pg_reset_latency_counters(struct pktgen_interface_info* info) { + int i; + info->avg_latency = 0; + info->min_latency = 0x7fffffff; /* largest integer */ + info->max_latency = 0x80000000; /* smallest integer */ + info->pkts_rcvd_since_clear = 0; + for (i = 0; ilatency_bkts[i] = 0; + } +} + +static void pg_clear_counters(struct pktgen_interface_info* info, int seq_too) { + info->idle_acc = 0; + info->sofar = 0; + info->tx_bytes = 0; + info->errors = 0; + info->ooo_rcvd = 0; + info->dup_rcvd = 0; + info->pkts_rcvd = 0; + info->bytes_rcvd = 0; + info->non_pg_pkts_rcvd = 0; + info->seq_gap_rcvd = 0; /* dropped */ + + /* Clear some transient state */ + info->accum_delay_ns = 0; + info->sleeps = 0; + info->nanodelays = 0; + + /* This is a bit of a hack, but it gets the dup counters + * in line so we don't have false alarms on dropped pkts. + */ + if (seq_too) { + info->dup_since_incr = info->peer_multiskb - 1; + info->seq_num = 1; + info->last_seq_rcvd = 0; + } + + pg_reset_latency_counters(info); +} + +/* Adds an interface to the thread. The interface will be in + * the stopped queue untill started. + */ +static int add_interface_to_thread(struct pktgen_thread_info* pg_thread, + struct pktgen_interface_info* info) { + int rv = 0; + /* grab lock & insert into the stopped list */ + pg_lock(pg_thread, __FUNCTION__); + + if (info->pg_thread) { + printk("pktgen: ERROR: Already assigned to a thread.\n"); + rv = -EBUSY; + goto out; + } + + info->next = pg_thread->stopped_if_infos; + pg_thread->stopped_if_infos = info; + info->pg_thread = pg_thread; + + out: + pg_unlock(pg_thread, __FUNCTION__); + return rv; +} - nr_frags = skb_shinfo(skb)->nr_frags; - - spin_lock_bh(&odev->xmit_lock); - if (!netif_queue_stopped(odev)) { +/* Set up structure for sending pkts, clear counters, add to rcv hash, + * create initial packet, and move from the stopped to the running + * interface_info list + */ +static int pg_start_interface(struct pktgen_thread_info* pg_thread, + struct pktgen_interface_info* info) { + PG_DEBUG(printk("Entering pg_start_interface..\n")); + pg_setup_inject(info); + + if (!info->odev) { + return -1; + } + + PG_DEBUG(printk("About to clean counters..\n")); + pg_clear_counters(info, 1); - atomic_inc(&skb->users); + info->do_run_run = 1; /* Cranke yeself! */ + + info->skb = NULL; - if (odev->hard_start_xmit(skb, odev)) { + info->started_at = getCurUs(); - atomic_dec(&skb->users); - if (net_ratelimit()) { - printk(KERN_INFO "Hard xmit error\n"); + pg_lock(pg_thread, __FUNCTION__); + { + /* Remove from the stopped list */ + struct pktgen_interface_info* p = pg_thread->stopped_if_infos; + if (p == info) { + pg_thread->stopped_if_infos = p->next; + p->next = NULL; + } + else { + while (p) { + if (p->next == info) { + p->next = p->next->next; + info->next = NULL; + break; } - info->errors++; - last_ok = 0; - } - else { - last_ok = 1; - info->sofar++; - info->seq_num++; + p = p->next; } - } - else { - /* Re-try it next time */ - last_ok = 0; } + + info->next_tx_ns = 0; /* Transmit immediately */ + pg_thread->next_tx_ns = 0; + + /* Move to the front of the running list */ + info->next = pg_thread->running_if_infos; + pg_thread->running_if_infos = info; + pg_thread->running_if_sz++; + } + pg_unlock(pg_thread, __FUNCTION__); + PG_DEBUG(printk("Leaving pg_start_interface..\n")); + return 0; +}/* pg_start_interface */ + + +/* set stopped-at timer, remove from running list, do counters & statistics + * NOTE: We do not remove from the rcv hash. + */ +static int pg_stop_interface(struct pktgen_thread_info* pg_thread, + struct pktgen_interface_info* info) { + __u64 total_us; + if (!info->do_run_run) { + printk("pktgen interface: %s is already stopped\n", info->ifname); + return -EINVAL; + } + + info->stopped_at = getCurMs(); + info->do_run_run = 0; + + /* The main worker loop will place it onto the stopped list if needed, + * next time this interface is asked to be re-inserted into the + * list. + */ + + total_us = info->stopped_at - info->started_at; + + { + __u64 idle = pg_div(info->idle_acc, 1000); /* convert to us */ + char *p = info->result; + __u64 pps = divremdi3(info->sofar * 1000, pg_div(total_us, 1000), PG_DIV); + __u64 bps = pps * 8 * (info->cur_pkt_size + 4); /* take 32bit ethernet CRC into account */ + p += sprintf(p, "OK: %llu(c%llu+d%llu) usec, %llu (%dbyte) %llupps %lluMb/sec (%llubps) errors: %llu", + total_us, total_us - idle, idle, + info->sofar, + info->cur_pkt_size + 4, /* Add 4 to account for the ethernet checksum */ + pps, + bps >> 20, bps, info->errors + ); + } + return 0; +}/* pg_stop_interface */ - spin_unlock_bh(&odev->xmit_lock); - if (info->ipg) { - /* Try not to busy-spin if we have larger sleep times. - * TODO: Investigate better ways to do this. - */ - if (info->ipg < 10000) { /* 10 usecs or less */ - nanospin(info->ipg, info); - } - else if (info->ipg < 10000000) { /* 10ms or less */ - udelay(info->ipg / 1000); +/* Re-inserts 'last' into the pg_thread's list. Calling code should + * make sure that 'last' is not already in the list. + */ +static struct pktgen_interface_info* pg_resort_pginfos(struct pktgen_thread_info* pg_thread, + struct pktgen_interface_info* last, + int setup_cur_if) { + struct pktgen_interface_info* rv = NULL; + + pg_lock(pg_thread, __FUNCTION__); + { + struct pktgen_interface_info* p = pg_thread->running_if_infos; + + if (last) { + if (!last->do_run_run) { + /* If this guy was stopped while 'current', then + * we'll want to place him on the stopped list + * here. + */ + last->next = pg_thread->stopped_if_infos; + pg_thread->stopped_if_infos = last; + pg_thread->running_if_sz--; } else { - mdelay(info->ipg / 1000000); + /* re-insert */ + if (!p) { + pg_thread->running_if_infos = last; + last->next = NULL; + } + else { + /* Another special case, check to see if we should go at the + * front of the queue. + */ + if (p->next_tx_ns > last->next_tx_ns) { + last->next = p; + pg_thread->running_if_infos = last; + } + else { + int inserted = 0; + while (p->next) { + if (p->next->next_tx_ns > last->next_tx_ns) { + /* Insert into the list */ + last->next = p->next; + p->next = last; + inserted = 1; + break; + } + p = p->next; + } + if (!inserted) { + /* place at the end */ + last->next = NULL; + p->next = last; + } + } + } } } - - if (signal_pending(current)) { - break; + + /* List is re-sorted, so grab the first one to return */ + rv = pg_thread->running_if_infos; + if (rv) { + /* Pop him off of the list. We do this here because we already + * have the lock. Calling code just has to be aware of this + * feature. + */ + pg_thread->running_if_infos = rv->next; } + } + + if (setup_cur_if) { + pg_thread->cur_if = rv; + } + + pg_unlock(pg_thread, __FUNCTION__); + return rv; +}/* pg_resort_pginfos */ - /* If lcount is zero, then run forever */ - if ((lcount != 0) && (--lcount == 0)) { - if (atomic_read(&skb->users) != 1) { - u32 idle_start, idle; - - idle_start = cycles(); - while (atomic_read(&skb->users) != 1) { - if (signal_pending(current)) { - break; - } - schedule(); - } - idle = cycles() - idle_start; - info->idle_acc += idle; - } - break; - } - if (netif_queue_stopped(odev) || current->need_resched) { - u32 idle_start, idle; +void pg_stop_all_ifs(struct pktgen_thread_info* pg_thread) { + struct pktgen_interface_info* next = NULL; - idle_start = cycles(); - do { - if (signal_pending(current)) { - info->do_run_run = 0; + pg_lock(pg_thread, __FUNCTION__); + if (pg_thread->cur_if) { + /* Move it onto the stopped list */ + pg_stop_interface(pg_thread, pg_thread->cur_if); + pg_thread->cur_if->next = pg_thread->stopped_if_infos; + pg_thread->stopped_if_infos = pg_thread->cur_if; + pg_thread->cur_if = NULL; + } + pg_unlock(pg_thread, __FUNCTION__); + + /* These have their own locking */ + next = pg_resort_pginfos(pg_thread, NULL, 0); + while (next) { + pg_stop_interface(pg_thread, next); + next = pg_resort_pginfos(pg_thread, NULL, 0); + } +}/* pg_stop_all_ifs */ + + +void pg_rem_all_ifs(struct pktgen_thread_info* pg_thread) { + struct pktgen_interface_info* next = NULL; + + /* Remove all interfaces, clean up memory */ + while ((next = pg_thread->stopped_if_infos)) { + int rv = pg_rem_interface_info(pg_thread, next); + if (rv >= 0) { + kfree(next); + } + else { + printk("ERROR: failed to rem_interface: %i\n", rv); + } + } +}/* pg_rem_all_ifs */ + + +void pg_rem_from_thread_list(struct pktgen_thread_info* pg_thread) { + /* Remove from the thread list */ + pg_lock_thread_list(__FUNCTION__); + { + struct pktgen_thread_info* tmp = pktgen_threads; + if (tmp == pg_thread) { + pktgen_threads = tmp->next; + } + else { + while (tmp) { + if (tmp->next == pg_thread) { + tmp->next = pg_thread->next; + pg_thread->next = NULL; break; } + tmp = tmp->next; + } + } + } + pg_unlock_thread_list(__FUNCTION__); +}/* pg_rem_from_thread_list */ + + +/* Main loop of the thread. Send pkts. + */ +void pg_thread_worker(struct pktgen_thread_info* pg_thread) { + struct net_device *odev = NULL; + __u64 idle_start = 0; + struct pktgen_interface_info* next = NULL; + u32 next_ipg = 0; + u64 now = 0; /* in nano-seconds */ + u32 tx_since_softirq = 0; + u32 queue_stopped = 0; + + /* setup the thread environment */ + init_pktgen_kthread(pg_thread, "kpktgend"); + + PG_DEBUG(printk("Starting up pktgen thread: %s\n", pg_thread->name)); + + /* an endless loop in which we are doing our work */ + while (! pg_thread->terminate) { + + if (signal_pending(current)) { + /* we received a request to terminate ourself */ + break; + } + + /* Re-sorts the list, inserting 'next' (which is really the last one + * we used). It pops the top one off of the queue and returns it. + * Calling code must make sure to re-insert the returned value + */ + next = pg_resort_pginfos(pg_thread, next, 1); + + /* Make sure the notify_queue_woken callback will act appropriately */ + if (next) { + pg_thread->next_tx_ns = next->next_tx_ns; + } + + if (queue_stopped > pg_thread->running_if_sz) { + /* All our devices are all fulled up, schedule and hope to run + * again soon. + */ + /* Take this opportunity to run the soft-irq */ + do_softirq(); + tx_since_softirq = 0; + + pg_thread->queues_stopped++; + pg_thread->sleeping = 1; + interruptible_sleep_on_timeout(&(pg_thread->queue), 1); + pg_thread->sleeping = 0; + queue_stopped = 0; + } + + if (next) { + + odev = next->odev; + + if (next->ipg || (next->accum_delay_ns > 0)) { + + now = getRelativeCurNs(); + if (now < next->next_tx_ns) { + next_ipg = (u32)(next->next_tx_ns - now); + + /* These will not actually busy-spin now. Will run as + * much as 1ms fast, and will sleep in 1ms units, assuming + * our tick is 1ms. + */ + pg_nanodelay(next_ipg, next, pg_thread); + if (!next->do_run_run) { + /* We were stopped while sleeping */ + continue; + } + } + + /* This is max IPG, this has special meaning of + * "never transmit" + */ + if (next->ipg == 0x7FFFFFFF) { + next->next_tx_ns = getRelativeCurNs() + next->ipg; + continue; + } + } + + if (need_resched()) { + idle_start = getRelativeCurNs(); + schedule(); + next->idle_acc += getRelativeCurNs() - idle_start; + } + + if (netif_queue_stopped(odev)) { + next->queue_stopped++; + queue_stopped++; + if (!netif_running(odev)) { - info->do_run_run = 0; - break; + pg_stop_interface(pg_thread, next); } - if (current->need_resched) - schedule(); - else - do_softirq(); - } while (netif_queue_stopped(odev)); - idle = cycles() - idle_start; - info->idle_acc += idle; - } - }/* while we should be running */ - do_gettimeofday(&(info->stopped_at)); + continue; /* Try the next interface */ + } + + if (next->last_ok || !next->skb) { + if ((++next->fp_tmp >= next->multiskb ) || (!next->skb)) { + /* build a new pkt */ + if (next->skb) { + kfree_skb(next->skb); + } + next->skb = fill_packet(odev, next); + if (next->skb == NULL) { + if (net_ratelimit()) { + printk(KERN_INFO "pktgen: Couldn't allocate skb in fill_packet.\n"); + } + schedule(); + next->fp_tmp--; /* back out increment, OOM */ + continue; + } + next->fp++; + next->fp_tmp = 0; /* reset counter */ + /* Not sure what good knowing nr_frags is... + next->nr_frags = skb_shinfo(skb)->nr_frags; + */ + } + atomic_inc(&(next->skb->users)); + } + + spin_lock_bh(&odev->xmit_lock); + if (!netif_queue_stopped(odev)) { + if (odev->hard_start_xmit(next->skb, odev)) { + if (net_ratelimit()) { + printk(KERN_INFO "pktgen: Hard xmit error\n"); + } + next->errors++; + next->last_ok = 0; + next->queue_stopped++; + queue_stopped++; + if (next->ntx_fudge < 10000) { + next->ntx_fudge = 10000; + } + else { + next->ntx_fudge = next->ntx_fudge << 1; + } + } + else { + queue_stopped = 0; /* reset this, we tx'd one successfully */ + next->last_ok = 1; + next->sofar++; + next->tx_bytes += (next->cur_pkt_size + 4); /* count csum */ + next->ntx_fudge = 0; + } + } + else { /* Re-try it next time */ + queue_stopped++; + next->queue_stopped++; + next->last_ok = 0; + if (next->ntx_fudge < 10000) { + next->ntx_fudge = 10000; + } + else { + next->ntx_fudge = next->ntx_fudge << 1; + } + } + spin_unlock_bh(&odev->xmit_lock); + + next->next_tx_ns = getRelativeCurNs() + next->ipg + next->ntx_fudge; - total = (info->stopped_at.tv_sec - info->started_at.tv_sec) * 1000000 + - info->stopped_at.tv_usec - info->started_at.tv_usec; + if (++tx_since_softirq > pg_thread->max_before_softirq) { + do_softirq(); + tx_since_softirq = 0; + } - idle = (__u32)(info->idle_acc)/(__u32)(cpu_speed); + /* If next->count is zero, then run forever */ + if ((next->count != 0) && (next->sofar >= next->count)) { + if (atomic_read(&(next->skb->users)) != 1) { + idle_start = getRelativeCurNs(); + while (atomic_read(&(next->skb->users)) != 1) { + if (signal_pending(current)) { + break; + } + schedule(); + } + next->idle_acc += getRelativeCurNs() - idle_start; + } + pg_stop_interface(pg_thread, next); + }/* if we're done with a particular interface. */ - { - char *p = info->result; - __u64 pps = (__u32)(info->sofar * 1000) / ((__u32)(total) / 1000); - __u64 bps = pps * 8 * (info->pkt_size + 4); /* take 32bit ethernet CRC into account */ - p += sprintf(p, "OK: %llu(c%llu+d%llu) usec, %llu (%dbyte,%dfrags) %llupps %lluMb/sec (%llubps) errors: %llu", - (unsigned long long) total, - (unsigned long long) (total - idle), - (unsigned long long) idle, - (unsigned long long) info->sofar, - skb->len + 4, /* Add 4 to account for the ethernet checksum */ - nr_frags, - (unsigned long long) pps, - (unsigned long long) (bps / (u64) 1024 / (u64) 1024), - (unsigned long long) bps, - (unsigned long long) info->errors - ); - } + }/* if could find the next interface to send on. */ + else { + /* fall asleep for a bit */ + pg_thread->sleeping = 1; + interruptible_sleep_on_timeout(&(pg_thread->queue), HZ/10); + pg_thread->sleeping = 0; + } + }//while true + + /* here we go only in case of termination of the thread */ + + PG_DEBUG(printk("pgthread: %s stopping all Interfaces.\n", pg_thread->name)); + pg_stop_all_ifs(pg_thread); + + PG_DEBUG(printk("pgthread: %s removing all Interfaces.\n", pg_thread->name)); + pg_rem_all_ifs(pg_thread); + + pg_rem_from_thread_list(pg_thread); + + /* cleanup the thread, leave */ + PG_DEBUG(printk("pgthread: %s calling exit_pktgen_kthread.\n", pg_thread->name)); + exit_pktgen_kthread(pg_thread); +} - kfree_skb(skb); +/* private functions */ +static void kthread_launcher(void *data) { + struct pktgen_thread_info *kthread = data; + kernel_thread((int (*)(void *))kthread->function, (void *)kthread, 0); +} -out_reldev: - if (odev) { - dev_put(odev); - odev = NULL; - } +/* create a new kernel thread. Called by the creator. */ +void start_pktgen_kthread(struct pktgen_thread_info *kthread) { - return; + /* initialize the semaphore: + we start with the semaphore locked. The new kernel + thread will setup its stuff and unlock it. This + control flow (the one that creates the thread) blocks + in the down operation below until the thread has reached + the up() operation. + */ + init_MUTEX_LOCKED(&kthread->startstop_sem); + + /* store the function to be executed in the data passed to + the launcher */ + kthread->function = pg_thread_worker; + + /* create the new thread by running a task through keventd */ + +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,0) + /* initialize the task queue structure */ + kthread->tq.sync = 0; + INIT_LIST_HEAD(&kthread->tq.list); + kthread->tq.routine = kthread_launcher; + kthread->tq.data = kthread; + + /* and schedule it for execution */ + schedule_task(&kthread->tq); +#else + INIT_WORK(&(kthread->wq), kthread_launcher, kthread); + /* and schedule it for execution */ + schedule_work(&kthread->wq); +#endif + + /* wait till it has reached the setup_thread routine */ + down(&kthread->startstop_sem); } -/* proc/net/pktgen/pg */ +/* stop a kernel thread. Called by the removing instance */ +static void stop_pktgen_kthread(struct pktgen_thread_info *kthread) { + printk("pgthread: %s stop_pktgen_kthread.\n", kthread->name); -static int proc_busy_read(char *buf , char **start, off_t offset, - int len, int *eof, void *data) -{ - char *p; - int idx = (int)(long)(data); - struct pktgen_info* info = NULL; + if (kthread->thread == NULL) { + printk("stop_kthread: killing non existing thread!\n"); + return; + } + + /* Stop each interface */ + pg_lock(kthread, __FUNCTION__); + { + struct pktgen_interface_info* tmp = kthread->running_if_infos; + while (tmp) { + tmp->do_run_run = 0; + tmp->next_tx_ns = 0; + tmp = tmp->next; + } + if (kthread->cur_if) { + kthread->cur_if->do_run_run = 0; + kthread->cur_if->next_tx_ns = 0; + } + } + pg_unlock(kthread, __FUNCTION__); + + printk("pktgen: stopped all interfaces...\n"); + /* Wait for everything to fully stop */ + while (1) { + pg_lock(kthread, __FUNCTION__); + if (kthread->cur_if || kthread->running_if_infos) { + pg_unlock(kthread, __FUNCTION__); + if (need_resched()) { + schedule(); + } + mdelay(1); + } + else { + pg_unlock(kthread, __FUNCTION__); + break; + } + } + + printk("pktgen: About to lock kernel.\n"); + + /* this function needs to be protected with the big + kernel lock (lock_kernel()). The lock must be + grabbed before changing the terminate + flag and released after the down() call. */ + lock_kernel(); + + /* initialize the semaphore. We lock it here, the + leave_thread call of the thread to be terminated + will unlock it. As soon as we see the semaphore + unlocked, we know that the thread has exited. + */ + init_MUTEX_LOCKED(&kthread->startstop_sem); + + /* We need to do a memory barrier here to be sure that + the flags are visible on all CPUs. + */ + mb(); + + printk("pktgen: Setting terminate to true.\n"); + + /* set flag to request thread termination */ + kthread->terminate = 1; + + /* We need to do a memory barrier here to be sure that + the flags are visible on all CPUs. + */ + mb(); + kill_proc(kthread->thread->pid, SIGKILL, 1); + + printk("pktgen: Waiting for thread to stop..\n"); + /* block till thread terminated */ + down(&kthread->startstop_sem); + kthread->in_use = 0; - if ((idx < 0) || (idx >= MAX_PKTGEN)) { - printk("ERROR: idx: %i is out of range in proc_write\n", idx); - return -EINVAL; + /* release the big kernel lock */ + unlock_kernel(); + + /* now we are sure the thread is in zombie state. We + notify keventd to clean the process up. + */ + kill_proc(2, SIGCHLD, 1); + + printk("pgthread: %s done with stop_pktgen_kthread.\n", kthread->name); +}/* stop_pktgen_kthread */ + + +/* initialize new created thread. Called by the new thread. */ +void init_pktgen_kthread(struct pktgen_thread_info *kthread, char *name) { + /* lock the kernel. A new kernel thread starts without + the big kernel lock, regardless of the lock state + of the creator (the lock level is *not* inheritated) + */ + lock_kernel(); + + /* fill in thread structure */ + kthread->thread = current; + + /* set signal mask to what we want to respond */ + siginitsetinv(¤t->blocked, sigmask(SIGKILL)|sigmask(SIGINT)|sigmask(SIGTERM)); + + /* initialise wait queue */ + init_waitqueue_head(&kthread->queue); + + /* initialise termination flag */ + kthread->terminate = 0; + + /* set name of this process (max 15 chars + 0 !) */ + sprintf(current->comm, name); + + /* let others run */ + unlock_kernel(); + + /* tell the creator that we are ready and let him continue */ + up(&kthread->startstop_sem); +}/* init_pktgen_kthread */ + +/* cleanup of thread. Called by the exiting thread. */ +static void exit_pktgen_kthread(struct pktgen_thread_info *kthread) { + /* we are terminating */ + + /* lock the kernel, the exit will unlock it */ + lock_kernel(); + kthread->thread = NULL; + mb(); + + /* Clean up proc file system */ + if (strlen(kthread->fname)) { + remove_proc_entry(kthread->fname, NULL); } - info = &(pginfos[idx]); - - p = buf; - p += sprintf(p, "%d\n", info->busy); - *eof = 1; - - return p-buf; + + /* notify the stop_kthread() routine that we are terminating. */ + up(&kthread->startstop_sem); + /* the kernel_thread that called clone() does a do_exit here. */ + + /* there is no race here between execution of the "killer" and real termination + of the thread (race window between up and do_exit), since both the + thread and the "killer" function are running with the kernel lock held. + The kernel lock will be freed after the thread exited, so the code + is really not executed anymore as soon as the unload functions gets + the kernel lock back. + The init process may not have made the cleanup of the process here, + but the cleanup can be done safely with the module unloaded. + */ +}/* exit_pktgen_kthread */ + + +/* proc/net/pg */ + +static char* pg_display_latency(struct pktgen_interface_info* info, char* p, int reset_latency) { + int i; + p += sprintf(p, " avg_latency: %dus min_lat: %dus max_lat: %dus pkts_in_sample: %llu\n", + info->avg_latency, info->min_latency, info->max_latency, + info->pkts_rcvd_since_clear); + p += sprintf(p, " Buckets(us) [ "); + for (i = 0; ilatency_bkts[i]); + } + p += sprintf(p, "]\n"); + + if (reset_latency) { + pg_reset_latency_counters(info); + } + return p; } -static int proc_read(char *buf , char **start, off_t offset, - int len, int *eof, void *data) +static int proc_pg_if_read(char *buf , char **start, off_t offset, + int len, int *eof, void *data) { char *p; int i; - int idx = (int)(long)(data); - struct pktgen_info* info = NULL; + struct pktgen_interface_info* info = (struct pktgen_interface_info*)(data); __u64 sa; __u64 stopped; - __u64 now = getCurMs(); + __u64 now = getCurUs(); + __u64 now_rel_ns = getRelativeCurNs(); - if ((idx < 0) || (idx >= MAX_PKTGEN)) { - printk("ERROR: idx: %i is out of range in proc_write\n", idx); - return -EINVAL; - } - info = &(pginfos[idx]); - p = buf; - p += sprintf(p, "%s\n", VERSION); /* Help with parsing compatibility */ - p += sprintf(p, "Params: count %llu pkt_size: %u frags: %d ipg: %u clone_skb: %d odev \"%s\"\n", - (unsigned long long) info->count, - info->pkt_size, info->nfrags, info->ipg, - info->clone_skb, info->outdev); - p += sprintf(p, " dst_min: %s dst_max: %s src_min: %s src_max: %s\n", + p += sprintf(p, "VERSION-1\n"); /* Help with parsing compatibility */ + p += sprintf(p, "Params: count %llu min_pkt_size: %u max_pkt_size: %u cur_pkt_size %u\n frags: %d ipg: %u multiskb: %d ifname: %s\n", + info->count, info->min_pkt_size, info->max_pkt_size, info->cur_pkt_size, + info->nfrags, info->ipg, info->multiskb, info->ifname); + p += sprintf(p, " dst_min: %s dst_max: %s\n src_min: %s src_max: %s\n", info->dst_min, info->dst_max, info->src_min, info->src_max); p += sprintf(p, " src_mac: "); for (i = 0; i < 6; i++) { @@ -808,14 +2129,17 @@ p += sprintf(p, " udp_src_min: %d udp_src_max: %d udp_dst_min: %d udp_dst_max: %d\n", info->udp_src_min, info->udp_src_max, info->udp_dst_min, info->udp_dst_max); - p += sprintf(p, " src_mac_count: %d dst_mac_count: %d\n Flags: ", - info->src_mac_count, info->dst_mac_count); + p += sprintf(p, " src_mac_count: %d dst_mac_count: %d peer_multiskb: %d\n Flags: ", + info->src_mac_count, info->dst_mac_count, info->peer_multiskb); if (info->flags & F_IPSRC_RND) { p += sprintf(p, "IPSRC_RND "); } if (info->flags & F_IPDST_RND) { p += sprintf(p, "IPDST_RND "); } + if (info->flags & F_TXSIZE_RND) { + p += sprintf(p, "TXSIZE_RND "); + } if (info->flags & F_UDPSRC_RND) { p += sprintf(p, "UDPSRC_RND "); } @@ -828,24 +2152,32 @@ if (info->flags & F_MACDST_RND) { p += sprintf(p, "MACDST_RND "); } + if (info->flags & F_IPMAC) { /* dhetheri */ + p += sprintf(p, "IPMAC "); + } + p += sprintf(p, "\n"); - sa = tv_to_ms(&(info->started_at)); - stopped = tv_to_ms(&(info->stopped_at)); + sa = info->started_at; + stopped = info->stopped_at; if (info->do_run_run) { stopped = now; /* not really stopped, more like last-running-at */ } - p += sprintf(p, "Current:\n pkts-sofar: %llu errors: %llu\n started: %llums stopped: %llums now: %llums idle: %lluns\n", - (unsigned long long) info->sofar, - (unsigned long long) info->errors, - (unsigned long long) sa, - (unsigned long long) stopped, - (unsigned long long) now, - (unsigned long long) info->idle_acc); + p += sprintf(p, "Current:\n pkts-sofar: %llu errors: %llu queue_stopped: %lu\naccum_delay: %lluns sleeps: %u nanodelays: %llu\n started: %lluus elapsed: %lluus\n idle: %lluns next_tx: %llu(%lli)ns\n", + info->sofar, info->errors, info->queue_stopped, + info->accum_delay_ns, info->sleeps, info->nanodelays, + sa, (stopped - sa), info->idle_acc, + info->next_tx_ns, (long long)(info->next_tx_ns) - (long long)(now_rel_ns)); p += sprintf(p, " seq_num: %d cur_dst_mac_offset: %d cur_src_mac_offset: %d\n", info->seq_num, info->cur_dst_mac_offset, info->cur_src_mac_offset); p += sprintf(p, " cur_saddr: 0x%x cur_daddr: 0x%x cur_udp_dst: %d cur_udp_src: %d\n", info->cur_saddr, info->cur_daddr, info->cur_udp_dst, info->cur_udp_src); + p += sprintf(p, " pkts_rcvd: %llu bytes_rcvd: %llu last_seq_rcvd: %d ooo_rcvd: %llu\n", + info->pkts_rcvd, info->bytes_rcvd, info->last_seq_rcvd, info->ooo_rcvd); + p += sprintf(p, " dup_rcvd: %llu seq_gap_rcvd(dropped): %llu non_pg_rcvd: %llu\n", + info->dup_rcvd, info->seq_gap_rcvd, info->non_pg_pkts_rcvd); + + p = pg_display_latency(info, p, 0); if (info->result[0]) p += sprintf(p, "Result: %s\n", info->result); @@ -856,439 +2188,992 @@ return p - buf; } -static int count_trail_chars(const char *user_buffer, unsigned int maxlen) -{ - int i; - - for (i = 0; i < maxlen; i++) { - char c; - - if (get_user(c, &user_buffer[i])) - return -EFAULT; - switch (c) { - case '\"': - case '\n': - case '\r': - case '\t': - case ' ': - case '=': - break; - default: - goto done; - }; - } -done: - return i; -} -static unsigned long num_arg(const char *user_buffer, unsigned long maxlen, - unsigned long *num) +static int proc_pg_thread_read(char *buf , char **start, off_t offset, + int len, int *eof, void *data) { - int i = 0; + char *p; + struct pktgen_thread_info* pg_thread = (struct pktgen_thread_info*)(data); + struct pktgen_interface_info* info = NULL; - *num = 0; - - for(; i < maxlen; i++) { - char c; + if (!pg_thread) { + printk("ERROR: could not find pg_thread in proc_pg_thread_read\n"); + return -EINVAL; + } + + p = buf; + p += sprintf(p, "VERSION-1 CFG_RT\n"); /* Help with parsing compatibility */ + p += sprintf(p, "PID: %i Name: %s max_before_softirq: %d queues_stopped: %u\n", + pg_thread->thread->pid, pg_thread->name, + pg_thread->max_before_softirq, pg_thread->queues_stopped); + + pg_lock(pg_thread, __FUNCTION__); + if (pg_thread->cur_if) { + p += sprintf(p, "Current: %s\n", pg_thread->cur_if->ifname); + } + else { + p += sprintf(p, "Current: NULL\n"); + } + pg_unlock(pg_thread, __FUNCTION__); + + p += sprintf(p, "Running: "); + + pg_lock(pg_thread, __FUNCTION__); + info = pg_thread->running_if_infos; + while (info) { + p += sprintf(p, "%s ", info->ifname); + info = info->next; + } + p += sprintf(p, "\nStopped: "); + info = pg_thread->stopped_if_infos; + while (info) { + p += sprintf(p, "%s ", info->ifname); + info = info->next; + } - if (get_user(c, &user_buffer[i])) - return -EFAULT; - if ((c >= '0') && (c <= '9')) { - *num *= 10; - *num += c -'0'; - } else - break; - } - return i; -} + if (pg_thread->result[0]) + p += sprintf(p, "\nResult: %s\n", pg_thread->result); + else + p += sprintf(p, "\nResult: NA\n"); + *eof = 1; -static int strn_len(const char *user_buffer, unsigned int maxlen) -{ - int i = 0; + pg_unlock(pg_thread, __FUNCTION__); - for(; i < maxlen; i++) { - char c; + return p - buf; +}/* proc_pg_thread_read */ - if (get_user(c, &user_buffer[i])) - return -EFAULT; - switch (c) { - case '\"': - case '\n': - case '\r': - case '\t': - case ' ': - goto done_str; - default: - break; - }; - } -done_str: - return i; -} -static int proc_write(struct file *file, const char *user_buffer, - unsigned long count, void *data) +static int proc_pg_ctrl_read(char *buf , char **start, off_t offset, + int len, int *eof, void *data) { - int i = 0, max, len; - char name[16], valstr[32]; - unsigned long value = 0; - int idx = (int)(long)(data); - struct pktgen_info* info = NULL; - char* result = NULL; - int tmp; + char *p; + struct pktgen_thread_info* pg_thread = NULL; - if ((idx < 0) || (idx >= MAX_PKTGEN)) { - printk("ERROR: idx: %i is out of range in proc_write\n", idx); - return -EINVAL; + p = buf; + p += sprintf(p, "VERSION-1\n"); /* Help with parsing compatibility */ + p += sprintf(p, "Threads: "); + + pg_lock_thread_list(__FUNCTION__); + pg_thread = pktgen_threads; + while (pg_thread) { + p += sprintf(p, "%s ", pg_thread->name); + pg_thread = pg_thread->next; } - info = &(pginfos[idx]); - result = &(info->result[0]); + p += sprintf(p, "\n"); + *eof = 1; + + pg_unlock_thread_list(__FUNCTION__); + return p - buf; +}/* proc_pg_ctrl_read */ + + +static int isdelim(const char c) { + switch (c) { + case '\"': + case '\n': + case '\r': + case '\t': + case ' ': + case '=': + return 1; + } + return 0; +} + + +static int count_trail_chars(const char *buf, unsigned int maxlen) { + int i; + + for (i = 0; i < maxlen; i++) { + if (!isdelim(buf[i])) { + break; + } + } + + return i; +} + + +static int strncpy_token(char* dst, const char* src, int mx) { + int i; + for (i = 0; i= '0') && (c <= '9')) { + num *= 10; + num += c - '0'; + } + else { + break; + } + } + return num; +} + +static int proc_pg_if_write(struct file *file, const char *user_buffer, + unsigned long count, void *data) +{ + char name[16]; + struct pktgen_interface_info* info = (struct pktgen_interface_info*)(data); + char* kbuf; + char* p; + char* pg_result = &(info->result[0]); + int len; + int value; + if (count < 1) { - sprintf(result, "Wrong command format"); + sprintf(pg_result, "Wrong command format"); return -EINVAL; } - max = count - i; - tmp = count_trail_chars(&user_buffer[i], max); - if (tmp < 0) - return tmp; - i += tmp; + kbuf = kmalloc(count, GFP_KERNEL); + if (copy_from_user(kbuf, user_buffer, count)) { + kfree(kbuf); + return -EFAULT; + } + + p = kbuf; + + while (p < (kbuf + count)) { + p += count_trail_chars(p, count - (p - kbuf)); - /* Read variable name */ + if (p >= (kbuf + count)) { + break; + } + + /* Read variable name */ + if (debug) { + printk("pg_thread: %s,%lu\n", name, count); + } - len = strn_len(&user_buffer[i], sizeof(name) - 1); - if (len < 0) - return len; - memset(name, 0, sizeof(name)); - if (copy_from_user(name, &user_buffer[i], len)) + len = strlen("stop"); + if (!strncmp(p, "stop", len)) { + p += len; + if (info->do_run_run) { + strcpy(pg_result, "Stopping"); + pg_stop_interface(info->pg_thread, info); + } + else { + strcpy(pg_result, "Already stopped...\n"); + } + goto foundcmd; + } + + len = strlen("min_pkt_size "); + if (!strncmp(p, "min_pkt_size ", len)) { + char f[32]; + p += len; + + p += count_trail_chars(p, count - (p - kbuf)); + + strncpy_token(f, p, min(31, (int)(count - (p - kbuf)))); + f[31] = 0; + + p += strlen(f); + value = atoui(f); + + if (value < 14+20+8) + value = 14+20+8; + if (value != info->min_pkt_size) { + info->min_pkt_size = value; + info->cur_pkt_size = value; + } + sprintf(pg_result, "OK: min_pkt_size=%u", info->min_pkt_size); + goto foundcmd; + } + + len = strlen("max_pkt_size "); + if (!strncmp(p, "max_pkt_size ", len)) { + char f[32]; + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncpy_token(f, p, min(31, (int)(count - (p - kbuf)))); + f[31] = 0; + + p += strlen(f); + value = atoui(f); + + if (value < 14+20+8) + value = 14+20+8; + if (value != info->max_pkt_size) { + info->max_pkt_size = value; + info->cur_pkt_size = value; + } + sprintf(pg_result, "OK: max_pkt_size=%u", info->max_pkt_size); + goto foundcmd; + } + + len = strlen("min_pkt_size "); + if (!strncmp(p, "min_pkt_size ", len)) { + char f[32]; + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncpy_token(f, p, min(31, (int)(count - (p - kbuf)))); + f[31] = 0; + + p += strlen(f); + debug = atoui(f); + sprintf(pg_result, "OK: debug=%u", debug); + goto foundcmd; + } + + len = strlen("frags "); + if (!strncmp(p, "frags ", len)) { + char f[32]; + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncpy_token(f, p, min(31, (int)(count - (p - kbuf)))); + f[31] = 0; + + p += strlen(f); + info->nfrags = atoui(f); + sprintf(pg_result, "OK: frags=%u", info->nfrags); + goto foundcmd; + } + + len = strlen("ipg "); + if (!strncmp(p, "ipg ", len)) { + char f[32]; + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncpy_token(f, p, min(31, (int)(count - (p - kbuf)))); + f[31] = 0; + + p += strlen(f); + info->ipg = atoui(f); + if ((getRelativeCurNs() + info->ipg) > info->next_tx_ns) { + info->next_tx_ns = getRelativeCurNs() + info->ipg; + } + sprintf(pg_result, "OK: ipg=%u", info->ipg); + goto foundcmd; + } + + len = strlen("udp_src_min "); + if (!strncmp(p, "udp_src_min ", len)) { + char f[32]; + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncpy_token(f, p, min(31, (int)(count - (p - kbuf)))); + f[31] = 0; + + p += strlen(f); + value = atoui(f); + + if (value != info->udp_src_min) { + info->udp_src_min = value; + info->cur_udp_src = value; + } + sprintf(pg_result, "OK: udp_src_min=%u", info->udp_src_min); + goto foundcmd; + } + + len = strlen("udp_dst_min "); + if (!strncmp(p, "udp_dst_min ", len)) { + char f[32]; + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncpy_token(f, p, min(31, (int)(count - (p - kbuf)))); + f[31] = 0; + + p += strlen(f); + value = atoui(f); + + if (value != info->udp_dst_min) { + info->udp_dst_min = value; + info->cur_udp_dst = value; + } + sprintf(pg_result, "OK: udp_dst_min=%u", info->udp_dst_min); + goto foundcmd; + } + + len = strlen("udp_src_max "); + if (!strncmp(p, "udp_src_max ", len)) { + char f[32]; + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncpy_token(f, p, min(31, (int)(count - (p - kbuf)))); + f[31] = 0; + + p += strlen(f); + value = atoui(f); + + if (value != info->udp_src_max) { + info->udp_src_max = value; + info->cur_udp_src = value; + } + sprintf(pg_result, "OK: udp_src_max=%u", info->udp_src_max); + goto foundcmd; + } + + len = strlen("udp_dst_max "); + if (!strncmp(p, "udp_dst_max ", len)) { + char f[32]; + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncpy_token(f, p, min(31, (int)(count - (p - kbuf)))); + f[31] = 0; + + p += strlen(f); + value = atoui(f); + + if (value != info->udp_dst_max) { + info->udp_dst_max = value; + info->cur_udp_dst = value; + } + sprintf(pg_result, "OK: udp_dst_max=%u", info->udp_dst_max); + goto foundcmd; + } + + len = strlen("multiskb "); + if (!strncmp(p, "multiskb ", len)) { + char f[32]; + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncpy_token(f, p, min(31, (int)(count - (p - kbuf)))); + f[31] = 0; + + p += strlen(f); + info->multiskb = atoui(f); + + /* Can't use multi-skb > 0 with virtual interfaces, because they change + * the skb->dev pointer (at least) and so it's really impossible to send + * the exact same pkt over and over again + */ + if (info->odev && + ((info->odev->priv_flags & IFF_MAC_VLAN) || + (info->odev->priv_flags & IFF_802_1Q_VLAN))) { + if (info->multiskb > 0) { + printk("pktgen: WARNING: Cannot use multi-skb > 0 on virtual interfaces, setting to zero.\n"); + info->multiskb = 0; + } + } + + sprintf(pg_result, "OK: multiskb=%d", info->multiskb); + goto foundcmd; + } + + len = strlen("peer_multiskb "); + if (!strncmp(p, "peer_multiskb ", len)) { + char f[32]; + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncpy_token(f, p, min(31, (int)(count - (p - kbuf)))); + f[31] = 0; + + p += strlen(f); + info->peer_multiskb = atoui(f); + + sprintf(pg_result, "OK: peer_multiskb=%d", + info->peer_multiskb); + goto foundcmd; + } + + len = strlen("count "); + if (!strncmp(p, "count ", len)) { + char f[32]; + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncpy_token(f, p, min(31, (int)(count - (p - kbuf)))); + f[31] = 0; + + p += strlen(f); + info->count = atoui(f); + + sprintf(pg_result, "OK: count=%llu", info->count); + goto foundcmd; + } + + len = strlen("prot "); + if (!strncmp(p, "prot ", len)) { + char f[32]; + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncpy_token(f, p, min(31, (int)(count - (p - kbuf)))); + f[31] = 0; + + p += strlen(f); + info->prot = atoui(f); + + sprintf(pg_result, "OK: prot=%u", info->prot); + goto foundcmd; + } + + len = strlen("src_mac_count "); + if (!strncmp(p, "src_mac_count ", len)) { + char f[32]; + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncpy_token(f, p, min(31, (int)(count - (p - kbuf)))); + f[31] = 0; + + p += strlen(f); + value = atoui(f); + + if (info->src_mac_count != value) { + info->src_mac_count = value; + info->cur_src_mac_offset = 0; + } + sprintf(pg_result, "OK: src_mac_count=%d", info->src_mac_count); + goto foundcmd; + } + + len = strlen("dst_mac_count "); + if (!strncmp(p, "dst_mac_count ", len)) { + char f[32]; + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncpy_token(f, p, min(31, (int)(count - (p - kbuf)))); + f[31] = 0; + + p += strlen(f); + value = atoui(f); + + if (info->dst_mac_count != value) { + info->dst_mac_count = value; + info->cur_dst_mac_offset = 0; + } + sprintf(pg_result, "OK: dst_mac_count=%d", info->dst_mac_count); + goto foundcmd; + } + + len = strlen("flag "); + if (!strncmp(p, "flag ", len)) { + char f[32]; + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncpy_token(f, p, min(31, (int)(count - (p - kbuf)))); + f[31] = 0; + + p += strlen(f); + + if (strcmp(f, "IPSRC_RND") == 0) { + info->flags |= F_IPSRC_RND; + } + else if (strcmp(f, "!IPSRC_RND") == 0) { + info->flags &= ~F_IPSRC_RND; + } + else if (strcmp(f, "TXSIZE_RND") == 0) { + info->flags |= F_TXSIZE_RND; + } + else if (strcmp(f, "!TXSIZE_RND") == 0) { + info->flags &= ~F_TXSIZE_RND; + } + else if (strcmp(f, "IPDST_RND") == 0) { + info->flags |= F_IPDST_RND; + } + else if (strcmp(f, "!IPDST_RND") == 0) { + info->flags &= ~F_IPDST_RND; + } + else if (strcmp(f, "UDPSRC_RND") == 0) { + info->flags |= F_UDPSRC_RND; + } + else if (strcmp(f, "!UDPSRC_RND") == 0) { + info->flags &= ~F_UDPSRC_RND; + } + else if (strcmp(f, "UDPDST_RND") == 0) { + info->flags |= F_UDPDST_RND; + } + else if (strcmp(f, "!UDPDST_RND") == 0) { + info->flags &= ~F_UDPDST_RND; + } + else if (strcmp(f, "MACSRC_RND") == 0) { + info->flags |= F_MACSRC_RND; + } + else if (strcmp(f, "!MACSRC_RND") == 0) { + info->flags &= ~F_MACSRC_RND; + } + else if (strcmp(f, "MACDST_RND") == 0) { + info->flags |= F_MACDST_RND; + } + else if (strcmp(f, "!MACDST_RND") == 0) { + info->flags &= ~F_MACDST_RND; + } + else if (strcmp(f, "IPMAC") == 0) { /* dhetheri */ + info->flags |= F_IPMAC; /* dhetheri */ + } /* dhetheri */ + else if (strcmp(f, "!IPMAC") == 0) { /* dhetheri */ + info->flags &= ~F_IPMAC; /* dhetheri */ + } /* dhetheri */ + else { + sprintf(pg_result, "Flag -:%s:- unknown\nAvailable flags, (prepend ! to un-set flag):\n%s", + f, + "IPSRC_RND, IPDST_RND, TXSIZE_RND, UDPSRC_RND, UDPDST_RND, MACSRC_RND, MACDST_RND, IPMAC\n"); + } + sprintf(pg_result, "OK: flags=0x%x", info->flags); + goto foundcmd; + } + + len = strlen("dst_min "); + if (!strncmp(p, "dst_min ", 6)) { + char f[IP_NAME_SZ]; + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncpy_token(f, p, min(IP_NAME_SZ-1, (int)(count - (p - kbuf)))); + f[IP_NAME_SZ-1] = 0; + + p += strlen(f); + if (strcmp(f, info->dst_min) != 0) { + memset(info->dst_min, 0, sizeof(info->dst_min)); + strcpy(info->dst_min, f); + info->daddr_min = in_aton(info->dst_min); + info->cur_daddr = info->daddr_min; + } + if(debug) + printk("pg: dst_min set to: %s\n", info->dst_min); + sprintf(pg_result, "OK: dst_min=%s", info->dst_min); + goto foundcmd; + } + + len = strlen("dst_max "); + if (!strncmp(p, "dst_max ", len)) { + char f[IP_NAME_SZ]; + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncpy_token(f, p, min(IP_NAME_SZ-1, (int)(count - (p - kbuf)))); + f[IP_NAME_SZ-1] = 0; + + p += strlen(f); + if (strcmp(f, info->dst_max) != 0) { + memset(info->dst_max, 0, sizeof(info->dst_max)); + strcpy(info->dst_max, f); + info->daddr_max = in_aton(info->dst_max); + info->cur_daddr = info->daddr_max; + } + if(debug) + printk("pg: dst_max set to: %s\n", info->dst_max); + sprintf(pg_result, "OK: dst_max=%s", info->dst_max); + goto foundcmd; + } + + len = strlen("src_min "); + if (!strncmp(p, "src_min ", len)) { + char f[IP_NAME_SZ]; + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncpy_token(f, p, min(IP_NAME_SZ-1, (int)(count - (p - kbuf)))); + f[IP_NAME_SZ-1] = 0; + + p += strlen(f); + if (strcmp(f, info->src_min) != 0) { + memset(info->src_min, 0, sizeof(info->src_min)); + strcpy(info->src_min, f); + info->saddr_min = in_aton(info->src_min); + info->cur_saddr = info->saddr_min; + } + if(debug) + printk("pg: src_min set to: %s\n", info->src_min); + sprintf(pg_result, "OK: src_min=%s", info->src_min); + goto foundcmd; + } + + len = strlen("src_max "); + if (!strncmp(p, "src_max ", len)) { + char f[IP_NAME_SZ]; + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncpy_token(f, p, min(IP_NAME_SZ-1, (int)(count - (p - kbuf)))); + f[IP_NAME_SZ-1] = 0; + + p += strlen(f); + if (strcmp(f, info->src_max) != 0) { + memset(info->src_max, 0, sizeof(info->src_max)); + strcpy(info->src_max, f); + info->saddr_min = in_aton(info->src_max); + info->cur_saddr = info->saddr_max; + } + if(debug) + printk("pg: src_min set to: %s\n", info->src_min); + sprintf(pg_result, "OK: src_max=%s", info->src_max); + goto foundcmd; + } + + len = strlen("dst_max "); + if (!strncmp(p, "dst_mac ", len)) { + char f[IP_NAME_SZ]; + unsigned char old_dmac[6]; + unsigned char *m = info->dst_mac; + char* v = f; + memcpy(old_dmac, info->dst_mac, 6); + + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncpy_token(f, p, min(IP_NAME_SZ-1, (int)(count - (p - kbuf)))); + f[IP_NAME_SZ-1] = 0; + + p += strlen(f); + + for(*m = 0;*v && m < info->dst_mac + 6; v++) { + if (*v >= '0' && *v <= '9') { + *m *= 16; + *m += *v - '0'; + } + if (*v >= 'A' && *v <= 'F') { + *m *= 16; + *m += *v - 'A' + 10; + } + if (*v >= 'a' && *v <= 'f') { + *m *= 16; + *m += *v - 'a' + 10; + } + if (*v == ':') { + m++; + *m = 0; + } + } + + if (memcmp(old_dmac, info->dst_mac, 6) != 0) { + /* Set up Dest MAC */ + memcpy(&(info->hh[0]), info->dst_mac, 6); + } + + sprintf(pg_result, "OK: dstmac"); + goto foundcmd; + } + + len = strlen("src_mac "); + if (!strncmp(p, "src_mac ", len)) { + char f[IP_NAME_SZ]; + char* v = f; + unsigned char old_smac[6]; + unsigned char *m = info->src_mac; + memcpy(old_smac, info->src_mac, 6); + + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncpy_token(f, p, min(IP_NAME_SZ-1, (int)(count - (p - kbuf)))); + f[IP_NAME_SZ-1] = 0; + + p += strlen(f); + + for(*m = 0;*v && m < info->src_mac + 6; v++) { + if (*v >= '0' && *v <= '9') { + *m *= 16; + *m += *v - '0'; + } + if (*v >= 'A' && *v <= 'F') { + *m *= 16; + *m += *v - 'A' + 10; + } + if (*v >= 'a' && *v <= 'f') { + *m *= 16; + *m += *v - 'a' + 10; + } + if (*v == ':') { + m++; + *m = 0; + } + } + + if (memcmp(old_smac, info->src_mac, 6) != 0) { + /* Default to the interface's mac if not explicitly set. */ + if ((!(info->flags & F_SET_SRCMAC)) && info->odev) { + memcpy(&(info->hh[6]), info->odev->dev_addr, 6); + } + else { + memcpy(&(info->hh[6]), info->src_mac, 6); + } + } + + sprintf(pg_result, "OK: srcmac"); + goto foundcmd; + } + + len = strlen("clear_counters"); + if (!strncmp(p, "clear_counters", len)) { + p += len; + + pg_clear_counters(info, 0); + sprintf(pg_result, "OK: Clearing counters...\n"); + goto foundcmd; + } + + if (!strncmp(p, "inject", 6) || !strncmp(p, "start", 5)) { + + if (strncmp(p, "start", 5) == 0) { + p += 5; + } + else { + p += 6; + } + + if (info->do_run_run) { + strcpy(info->result, "Already running...\n"); + sprintf(pg_result, "Already running...\n"); + } + else { + int rv; + if ((rv = pg_start_interface(info->pg_thread, info)) >= 0) { + strcpy(info->result, "Starting"); + sprintf(pg_result, "Starting"); + } + else { + sprintf(info->result, "Error starting: %i\n", rv); + sprintf(pg_result, "Error starting: %i\n", rv); + } + } + goto foundcmd; + } + + printk("Pktgen:pg_if_write: Unknown command -:%20s:-\n", p); + p++; + + foundcmd: + if (debug & 0x1000) { + printk("Command succeeded.\n"); + } + }/* while */ + + kfree(kbuf); + return count; +}/* proc_pg_if_write */ + + +static int proc_pg_ctrl_write(struct file *file, const char *user_buffer, + unsigned long count, void *data) +{ + char name[16]; + struct pktgen_thread_info* pg_thread = NULL; + char* kbuf; + char* p; + int len; + + if (count < 1) { + return -EINVAL; + } + + kbuf = kmalloc(count, GFP_KERNEL); + if (copy_from_user(kbuf, user_buffer, count)) { + kfree(kbuf); return -EFAULT; - i += len; - - max = count -i; - len = count_trail_chars(&user_buffer[i], max); - if (len < 0) - return len; - i += len; - - if (debug) - printk("pg: %s,%lu\n", name, count); - - if (!strcmp(name, "stop")) { - if (info->do_run_run) { - strcpy(result, "Stopping"); - } - else { - strcpy(result, "Already stopped...\n"); - } - info->do_run_run = 0; - return count; } - if (!strcmp(name, "pkt_size")) { - len = num_arg(&user_buffer[i], 10, &value); - if (len < 0) - return len; - i += len; - if (value < 14+20+8) - value = 14+20+8; - info->pkt_size = value; - sprintf(result, "OK: pkt_size=%u", info->pkt_size); - return count; - } - if (!strcmp(name, "frags")) { - len = num_arg(&user_buffer[i], 10, &value); - if (len < 0) - return len; - i += len; - info->nfrags = value; - sprintf(result, "OK: frags=%u", info->nfrags); - return count; - } - if (!strcmp(name, "ipg")) { - len = num_arg(&user_buffer[i], 10, &value); - if (len < 0) - return len; - i += len; - info->ipg = value; - sprintf(result, "OK: ipg=%u", info->ipg); - return count; - } - if (!strcmp(name, "udp_src_min")) { - len = num_arg(&user_buffer[i], 10, &value); - if (len < 0) - return len; - i += len; - info->udp_src_min = value; - sprintf(result, "OK: udp_src_min=%u", info->udp_src_min); - return count; - } - if (!strcmp(name, "udp_dst_min")) { - len = num_arg(&user_buffer[i], 10, &value); - if (len < 0) - return len; - i += len; - info->udp_dst_min = value; - sprintf(result, "OK: udp_dst_min=%u", info->udp_dst_min); - return count; - } - if (!strcmp(name, "udp_src_max")) { - len = num_arg(&user_buffer[i], 10, &value); - if (len < 0) - return len; - i += len; - info->udp_src_max = value; - sprintf(result, "OK: udp_src_max=%u", info->udp_src_max); - return count; - } - if (!strcmp(name, "udp_dst_max")) { - len = num_arg(&user_buffer[i], 10, &value); - if (len < 0) - return len; - i += len; - info->udp_dst_max = value; - sprintf(result, "OK: udp_dst_max=%u", info->udp_dst_max); - return count; - } - if (!strcmp(name, "clone_skb")) { - len = num_arg(&user_buffer[i], 10, &value); - if (len < 0) - return len; - i += len; - info->clone_skb = value; + p = kbuf; + + while (p < (kbuf + count)) { + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + if (p >= (kbuf + count)) { + break; + } + + /* Read variable name */ + if (debug) { + printk("pg_thread: %s,%lu\n", name, count); + } + + len = strlen("stop "); + if (!strncmp(p, "stop ", len)) { + char f[32]; + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncpy_token(f, p, min(31, (int)(count - (p - kbuf)))); + f[31] = 0; + + p += strlen(f); + + pg_thread = pg_find_thread(f); + if (pg_thread) { + printk("pktgen INFO: stopping thread: %s\n", + pg_thread->name); + stop_pktgen_kthread(pg_thread); + } + + goto foundcmd; + } + + len = strlen("start "); + if (!strncmp(p, "start ", len)) { + char f[32]; + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncpy_token(f, p, min(31, (int)(count - (p - kbuf)))); + f[31] = 0; + + p += strlen(f); + + pg_add_thread_info(f); + goto foundcmd; + } + + printk("Pktgen:pgctrl_write: Unknown command -:%20s:-\n", p); + p++; + + foundcmd: + if (debug & 0x1000) { + printk("Command handled successfully.\n"); + } + + }/* while */ + + kfree(kbuf); + return count; +}/* proc_pg_ctrl_write */ + + +static int proc_pg_thread_write(struct file *file, const char *user_buffer, + unsigned long count, void *data) +{ + char name[16]; + struct pktgen_thread_info* pg_thread = (struct pktgen_thread_info*)(data); + char* pg_result = &(pg_thread->result[0]); + char* kbuf; + char* p; + int len; - sprintf(result, "OK: clone_skb=%d", info->clone_skb); - return count; + if (count < 1) { + sprintf(pg_result, "Wrong command format"); + return -EINVAL; } - if (!strcmp(name, "count")) { - len = num_arg(&user_buffer[i], 10, &value); - if (len < 0) - return len; - i += len; - info->count = value; - sprintf(result, "OK: count=%llu", (unsigned long long) info->count); - return count; - } - if (!strcmp(name, "src_mac_count")) { - len = num_arg(&user_buffer[i], 10, &value); - if (len < 0) - return len; - i += len; - info->src_mac_count = value; - sprintf(result, "OK: src_mac_count=%d", info->src_mac_count); - return count; - } - if (!strcmp(name, "dst_mac_count")) { - len = num_arg(&user_buffer[i], 10, &value); - if (len < 0) - return len; - i += len; - info->dst_mac_count = value; - sprintf(result, "OK: dst_mac_count=%d", info->dst_mac_count); - return count; - } - if (!strcmp(name, "odev")) { - len = strn_len(&user_buffer[i], sizeof(info->outdev) - 1); - if (len < 0) - return len; - memset(info->outdev, 0, sizeof(info->outdev)); - if (copy_from_user(info->outdev, &user_buffer[i], len)) - return -EFAULT; - i += len; - sprintf(result, "OK: odev=%s", info->outdev); - return count; - } - if (!strcmp(name, "flag")) { - char f[32]; - len = strn_len(&user_buffer[i], sizeof(f) - 1); - if (len < 0) - return len; - memset(f, 0, 32); - if (copy_from_user(f, &user_buffer[i], len)) - return -EFAULT; - i += len; - if (strcmp(f, "IPSRC_RND") == 0) { - info->flags |= F_IPSRC_RND; - } - else if (strcmp(f, "!IPSRC_RND") == 0) { - info->flags &= ~F_IPSRC_RND; - } - else if (strcmp(f, "IPDST_RND") == 0) { - info->flags |= F_IPDST_RND; - } - else if (strcmp(f, "!IPDST_RND") == 0) { - info->flags &= ~F_IPDST_RND; - } - else if (strcmp(f, "UDPSRC_RND") == 0) { - info->flags |= F_UDPSRC_RND; - } - else if (strcmp(f, "!UDPSRC_RND") == 0) { - info->flags &= ~F_UDPSRC_RND; - } - else if (strcmp(f, "UDPDST_RND") == 0) { - info->flags |= F_UDPDST_RND; - } - else if (strcmp(f, "!UDPDST_RND") == 0) { - info->flags &= ~F_UDPDST_RND; - } - else if (strcmp(f, "MACSRC_RND") == 0) { - info->flags |= F_MACSRC_RND; - } - else if (strcmp(f, "!MACSRC_RND") == 0) { - info->flags &= ~F_MACSRC_RND; - } - else if (strcmp(f, "MACDST_RND") == 0) { - info->flags |= F_MACDST_RND; - } - else if (strcmp(f, "!MACDST_RND") == 0) { - info->flags &= ~F_MACDST_RND; - } - else { - sprintf(result, "Flag -:%s:- unknown\nAvailable flags, (prepend ! to un-set flag):\n%s", - f, - "IPSRC_RND, IPDST_RND, UDPSRC_RND, UDPDST_RND, MACSRC_RND, MACDST_RND\n"); - return count; - } - sprintf(result, "OK: flags=0x%x", info->flags); - return count; - } - if (!strcmp(name, "dst_min") || !strcmp(name, "dst")) { - len = strn_len(&user_buffer[i], sizeof(info->dst_min) - 1); - if (len < 0) - return len; - memset(info->dst_min, 0, sizeof(info->dst_min)); - if (copy_from_user(info->dst_min, &user_buffer[i], len)) - return -EFAULT; - if(debug) - printk("pg: dst_min set to: %s\n", info->dst_min); - i += len; - sprintf(result, "OK: dst_min=%s", info->dst_min); - return count; - } - if (!strcmp(name, "dst_max")) { - len = strn_len(&user_buffer[i], sizeof(info->dst_max) - 1); - if (len < 0) - return len; - memset(info->dst_max, 0, sizeof(info->dst_max)); - if (copy_from_user(info->dst_max, &user_buffer[i], len)) - return -EFAULT; - if(debug) - printk("pg: dst_max set to: %s\n", info->dst_max); - i += len; - sprintf(result, "OK: dst_max=%s", info->dst_max); - return count; - } - if (!strcmp(name, "src_min")) { - len = strn_len(&user_buffer[i], sizeof(info->src_min) - 1); - if (len < 0) - return len; - memset(info->src_min, 0, sizeof(info->src_min)); - if (copy_from_user(info->src_min, &user_buffer[i], len)) - return -EFAULT; - if(debug) - printk("pg: src_min set to: %s\n", info->src_min); - i += len; - sprintf(result, "OK: src_min=%s", info->src_min); - return count; - } - if (!strcmp(name, "src_max")) { - len = strn_len(&user_buffer[i], sizeof(info->src_max) - 1); - if (len < 0) - return len; - memset(info->src_max, 0, sizeof(info->src_max)); - if (copy_from_user(info->src_max, &user_buffer[i], len)) - return -EFAULT; - if(debug) - printk("pg: src_max set to: %s\n", info->src_max); - i += len; - sprintf(result, "OK: src_max=%s", info->src_max); - return count; - } - if (!strcmp(name, "dstmac")) { - char *v = valstr; - unsigned char *m = info->dst_mac; - - len = strn_len(&user_buffer[i], sizeof(valstr) - 1); - if (len < 0) - return len; - memset(valstr, 0, sizeof(valstr)); - if (copy_from_user(valstr, &user_buffer[i], len)) - return -EFAULT; - i += len; - - for(*m = 0;*v && m < info->dst_mac + 6; v++) { - if (*v >= '0' && *v <= '9') { - *m *= 16; - *m += *v - '0'; - } - if (*v >= 'A' && *v <= 'F') { - *m *= 16; - *m += *v - 'A' + 10; - } - if (*v >= 'a' && *v <= 'f') { - *m *= 16; - *m += *v - 'a' + 10; - } - if (*v == ':') { - m++; - *m = 0; - } - } - sprintf(result, "OK: dstmac"); - return count; - } - if (!strcmp(name, "srcmac")) { - char *v = valstr; - unsigned char *m = info->src_mac; - - len = strn_len(&user_buffer[i], sizeof(valstr) - 1); - if (len < 0) - return len; - memset(valstr, 0, sizeof(valstr)); - if (copy_from_user(valstr, &user_buffer[i], len)) - return -EFAULT; - i += len; - - for(*m = 0;*v && m < info->src_mac + 6; v++) { - if (*v >= '0' && *v <= '9') { - *m *= 16; - *m += *v - '0'; - } - if (*v >= 'A' && *v <= 'F') { - *m *= 16; - *m += *v - 'A' + 10; - } - if (*v >= 'a' && *v <= 'f') { - *m *= 16; - *m += *v - 'a' + 10; - } - if (*v == ':') { - m++; - *m = 0; - } - } - sprintf(result, "OK: srcmac"); - return count; - } - - if (!strcmp(name, "inject") || !strcmp(name, "start")) { - MOD_INC_USE_COUNT; - if (info->busy) { - strcpy(info->result, "Already running...\n"); - } - else { - info->busy = 1; - strcpy(info->result, "Starting"); - inject(info); - info->busy = 0; - } - MOD_DEC_USE_COUNT; - return count; + + kbuf = kmalloc(count, GFP_KERNEL); + if (copy_from_user(kbuf, user_buffer, count)) { + kfree(kbuf); + return -EFAULT; } - sprintf(info->result, "No such parameter \"%s\"", name); - return -EINVAL; -} + p = kbuf; + + while (p < (kbuf + count)) { + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + if (p >= (kbuf + count)) { + break; + } + + /* Read variable name */ + if (debug) { + printk("pg_thread: %s,%lu\n", name, count); + } + + len = strlen("add_interface "); + if (!strncmp(p, "add_interface ", len)) { + char f[32]; + p += len; + + p += count_trail_chars(p, (int)(count - (p - kbuf))); + + strncp