3 * Copyright 2001, 2002 by Robert Olsson <robert.olsson@its.uu.se>
4 * Uppsala University and
5 * Swedish University of Agricultural Sciences
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
8 * Ben Greear <greearb@candelatech.com>
9 * Jens Låås <jens.laas@data.slu.se>
11 * This program is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU General Public License
13 * as published by the Free Software Foundation; either version
14 * 2 of the License, or (at your option) any later version.
17 * A tool for loading the network with preconfigurated packets.
18 * The tool is implemented as a linux module. Parameters are output
19 * device, delay (to hard_xmit), number of packets, and whether
20 * to use multiple SKBs or just the same one.
21 * pktgen uses the installed interface's output routine.
23 * Additional hacking by:
25 * Jens.Laas@data.slu.se
26 * Improved by ANK. 010120.
27 * Improved by ANK even more. 010212.
28 * MAC address typo fixed. 010417 --ro
29 * Integrated. 020301 --DaveM
30 * Added multiskb option 020301 --DaveM
31 * Scaling of results. 020417--sigurdur@linpro.no
32 * Significant re-work of the module:
33 * * Convert to threaded model to more efficiently be able to transmit
34 * and receive on multiple interfaces at once.
35 * * Converted many counters to __u64 to allow longer runs.
36 * * Allow configuration of ranges, like min/max IP address, MACs,
37 * and UDP-ports, for both source and destination, and can
38 * set to use a random distribution or sequentially walk the range.
39 * * Can now change most values after starting.
40 * * Place 12-byte packet in UDP payload with magic number,
41 * sequence number, and timestamp.
42 * * Add receiver code that detects dropped pkts, re-ordered pkts, and
43 * latencies (with micro-second) precision.
44 * * Add IOCTL interface to easily get counters & configuration.
45 * --Ben Greear <greearb@candelatech.com>
47 * Renamed multiskb to clone_skb and cleaned up sending core for two distinct
48 * skb modes. A clone_skb=0 mode for Ben "ranges" work and a clone_skb != 0
49 * as a "fastpath" with a configurable number of clones after alloc's.
50 * clone_skb=0 means all packets are allocated this also means ranges time
51 * stamps etc can be used. clone_skb=100 means 1 malloc is followed by 100
54 * Also moved to /proc/net/pktgen/
57 * Sept 10: Fixed threading/locking. Lots of bone-headed and more clever
58 * mistakes. Also merged in DaveM's patch in the -pre6 patch.
59 * --Ben Greear <greearb@candelatech.com>
61 * Integrated to 2.5.x 021029 --Lucio Maciel (luciomaciel@zipmail.com.br)
64 * 021124 Finished major redesign and rewrite for new functionality.
65 * See Documentation/networking/pktgen.txt for how to use this.
68 * For each CPU one thread/process is created at start. This process checks
69 * for running devices in the if_list and sends packets until count is 0 it
70 * also the thread checks the thread->control which is used for inter-process
71 * communication. controlling process "posts" operations to the threads this
72 * way. The if_lock should be possible to remove when add/rem_device is merged
75 * By design there should only be *one* "controlling" process. In practice
76 * multiple write accesses gives unpredictable result. Understood by "write"
77 * to /proc gives result code thats should be read be the "writer".
78 * For practical use this should be no problem.
80 * Note when adding devices to a specific CPU there good idea to also assign
81 * /proc/irq/XX/smp_affinity so TX-interrupts gets bound to the same CPU.
84 * Fix refcount off by one if first packet fails, potential null deref,
87 * First "ranges" functionality for ipv6 030726 --ro
89 * Included flow support. 030802 ANK.
91 * Fixed unaligned access on IA-64 Grant Grundler <grundler@parisc-linux.org>
93 * Remove if fix from added Harald Welte <laforge@netfilter.org> 040419
94 * ia64 compilation fix from Aron Griffis <aron@hp.com> 040604
96 * New xmit() return, do_div and misc clean up by Stephen Hemminger
97 * <shemminger@osdl.org> 040923
99 * Randy Dunlap fixed u64 printk compiler waring
101 * Remove FCS from BW calculation. Lennert Buytenhek <buytenh@wantstofly.org>
102 * New time handling. Lennert Buytenhek <buytenh@wantstofly.org> 041213
104 * Corrections from Nikolai Malykh (nmalykh@bilim.com)
105 * Removed unused flags F_SET_SRCMAC & F_SET_SRCIP 041230
107 * interruptible_sleep_on_timeout() replaced Nishanth Aravamudan <nacc@us.ibm.com>
110 * MPLS support by Steven Whitehouse <steve@chygwyn.com>
112 * 802.1Q/Q-in-Q support by Francesco Fondelli (FF) <francesco.fondelli@gmail.com>
114 * Fixed src_mac command to set source mac of packet to value specified in
115 * command by Adit Ranadive <adit.262@gmail.com>
119 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
121 #include <linux/sys.h>
122 #include <linux/types.h>
123 #include <linux/module.h>
124 #include <linux/moduleparam.h>
125 #include <linux/kernel.h>
126 #include <linux/mutex.h>
127 #include <linux/sched.h>
128 #include <linux/slab.h>
129 #include <linux/vmalloc.h>
130 #include <linux/unistd.h>
131 #include <linux/string.h>
132 #include <linux/ptrace.h>
133 #include <linux/errno.h>
134 #include <linux/ioport.h>
135 #include <linux/interrupt.h>
136 #include <linux/capability.h>
137 #include <linux/hrtimer.h>
138 #include <linux/freezer.h>
139 #include <linux/delay.h>
140 #include <linux/timer.h>
141 #include <linux/list.h>
142 #include <linux/init.h>
143 #include <linux/skbuff.h>
144 #include <linux/netdevice.h>
145 #include <linux/inet.h>
146 #include <linux/inetdevice.h>
147 #include <linux/rtnetlink.h>
148 #include <linux/if_arp.h>
149 #include <linux/if_vlan.h>
150 #include <linux/in.h>
151 #include <linux/ip.h>
152 #include <linux/ipv6.h>
153 #include <linux/udp.h>
154 #include <linux/proc_fs.h>
155 #include <linux/seq_file.h>
156 #include <linux/wait.h>
157 #include <linux/etherdevice.h>
158 #include <linux/kthread.h>
159 #include <net/net_namespace.h>
160 #include <net/checksum.h>
161 #include <net/ipv6.h>
162 #include <net/addrconf.h>
164 #include <net/xfrm.h>
166 #include <asm/byteorder.h>
167 #include <linux/rcupdate.h>
168 #include <linux/bitops.h>
169 #include <linux/io.h>
170 #include <linux/timex.h>
171 #include <linux/uaccess.h>
173 #include <asm/div64.h> /* do_div */
175 #define VERSION "2.74"
176 #define IP_NAME_SZ 32
177 #define MAX_MPLS_LABELS 16 /* This is the max label stack depth */
178 #define MPLS_STACK_BOTTOM htonl(0x00000100)
180 #define func_enter() pr_debug("entering %s\n", __func__);
182 /* Device flag bits */
183 #define F_IPSRC_RND (1<<0) /* IP-Src Random */
184 #define F_IPDST_RND (1<<1) /* IP-Dst Random */
185 #define F_UDPSRC_RND (1<<2) /* UDP-Src Random */
186 #define F_UDPDST_RND (1<<3) /* UDP-Dst Random */
187 #define F_MACSRC_RND (1<<4) /* MAC-Src Random */
188 #define F_MACDST_RND (1<<5) /* MAC-Dst Random */
189 #define F_TXSIZE_RND (1<<6) /* Transmit size is random */
190 #define F_IPV6 (1<<7) /* Interface in IPV6 Mode */
191 #define F_MPLS_RND (1<<8) /* Random MPLS labels */
192 #define F_VID_RND (1<<9) /* Random VLAN ID */
193 #define F_SVID_RND (1<<10) /* Random SVLAN ID */
194 #define F_FLOW_SEQ (1<<11) /* Sequential flows */
195 #define F_IPSEC_ON (1<<12) /* ipsec on for flows */
196 #define F_QUEUE_MAP_RND (1<<13) /* queue map Random */
197 #define F_QUEUE_MAP_CPU (1<<14) /* queue map mirrors smp_processor_id() */
198 #define F_NODE (1<<15) /* Node memory alloc*/
200 /* Thread control flag bits */
201 #define T_STOP (1<<0) /* Stop run */
202 #define T_RUN (1<<1) /* Start run */
203 #define T_REMDEVALL (1<<2) /* Remove all devs */
204 #define T_REMDEV (1<<3) /* Remove one dev */
206 /* If lock -- can be removed after some work */
207 #define if_lock(t) spin_lock(&(t->if_lock));
208 #define if_unlock(t) spin_unlock(&(t->if_lock));
210 /* Used to help with determining the pkts on receive */
211 #define PKTGEN_MAGIC 0xbe9be955
212 #define PG_PROC_DIR "pktgen"
213 #define PGCTRL "pgctrl"
214 static struct proc_dir_entry *pg_proc_dir;
216 #define MAX_CFLOWS 65536
218 #define VLAN_TAG_SIZE(x) ((x)->vlan_id == 0xffff ? 0 : 4)
219 #define SVLAN_TAG_SIZE(x) ((x)->svlan_id == 0xffff ? 0 : 4)
225 struct xfrm_state *x;
231 #define F_INIT (1<<0) /* flow has been initialized */
235 * Try to keep frequent/infrequent used vars. separated.
237 struct proc_dir_entry *entry; /* proc file */
238 struct pktgen_thread *pg_thread;/* the owner */
239 struct list_head list; /* chaining in the thread's run-queue */
241 int running; /* if false, the test will stop */
243 /* If min != max, then we will either do a linear iteration, or
244 * we will do a random selection from within the range.
247 int removal_mark; /* non-zero => the device is marked for
248 * removal by worker thread */
250 int min_pkt_size; /* = ETH_ZLEN; */
251 int max_pkt_size; /* = ETH_ZLEN; */
252 int pkt_overhead; /* overhead for MPLS, VLANs, IPSEC etc */
255 u64 delay; /* nano-seconds */
257 __u64 count; /* Default No packets to send */
258 __u64 sofar; /* How many pkts we've sent so far */
259 __u64 tx_bytes; /* How many bytes we've transmitted */
260 __u64 errors; /* Errors when trying to transmit, */
262 /* runtime counters relating to clone_skb */
264 __u64 allocated_skbs;
266 int last_ok; /* Was last skb sent?
267 * Or a failed transmit of some sort?
268 * This will keep sequence numbers in order
273 u64 idle_acc; /* nano-seconds */
278 * Use multiple SKBs during packet gen.
279 * If this number is greater than 1, then
280 * that many copies of the same packet will be
281 * sent before a new packet is allocated.
282 * If you want to send 1024 identical packets
283 * before creating a new packet,
284 * set clone_skb to 1024.
287 char dst_min[IP_NAME_SZ]; /* IP, ie 1.2.3.4 */
288 char dst_max[IP_NAME_SZ]; /* IP, ie 1.2.3.4 */
289 char src_min[IP_NAME_SZ]; /* IP, ie 1.2.3.4 */
290 char src_max[IP_NAME_SZ]; /* IP, ie 1.2.3.4 */
292 struct in6_addr in6_saddr;
293 struct in6_addr in6_daddr;
294 struct in6_addr cur_in6_daddr;
295 struct in6_addr cur_in6_saddr;
297 struct in6_addr min_in6_daddr;
298 struct in6_addr max_in6_daddr;
299 struct in6_addr min_in6_saddr;
300 struct in6_addr max_in6_saddr;
302 /* If we're doing ranges, random or incremental, then this
303 * defines the min/max for those ranges.
305 __be32 saddr_min; /* inclusive, source IP address */
306 __be32 saddr_max; /* exclusive, source IP address */
307 __be32 daddr_min; /* inclusive, dest IP address */
308 __be32 daddr_max; /* exclusive, dest IP address */
310 __u16 udp_src_min; /* inclusive, source UDP port */
311 __u16 udp_src_max; /* exclusive, source UDP port */
312 __u16 udp_dst_min; /* inclusive, dest UDP port */
313 __u16 udp_dst_max; /* exclusive, dest UDP port */
316 __u8 tos; /* six MSB of (former) IPv4 TOS
317 are for dscp codepoint */
318 __u8 traffic_class; /* ditto for the (former) Traffic Class in IPv6
319 (see RFC 3260, sec. 4) */
322 unsigned nr_labels; /* Depth of stack, 0 = no MPLS */
323 __be32 labels[MAX_MPLS_LABELS];
325 /* VLAN/SVLAN (802.1Q/Q-in-Q) */
328 __u16 vlan_id; /* 0xffff means no vlan tag */
332 __u16 svlan_id; /* 0xffff means no svlan tag */
334 __u32 src_mac_count; /* How many MACs to iterate through */
335 __u32 dst_mac_count; /* How many MACs to iterate through */
337 unsigned char dst_mac[ETH_ALEN];
338 unsigned char src_mac[ETH_ALEN];
340 __u32 cur_dst_mac_offset;
341 __u32 cur_src_mac_offset;
353 0x00, 0x80, 0xC8, 0x79, 0xB3, 0xCB,
355 We fill in SRC address later
356 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
360 __u16 pad; /* pad out the hh struct to an even 16 bytes */
362 struct sk_buff *skb; /* skb we are to transmit next, used for when we
363 * are transmitting the same one multiple times
365 struct net_device *odev; /* The out-going device.
366 * Note that the device should have it's
367 * pg_info pointer pointing back to this
369 * Set when the user specifies the out-going
370 * device name (not when the inject is
371 * started as it used to do.)
374 struct flow_state *flows;
375 unsigned cflows; /* Concurrent flows (config) */
376 unsigned lflow; /* Flow length (config) */
377 unsigned nflows; /* accumulated flows (stats) */
378 unsigned curfl; /* current sequenced flow (state)*/
382 __u32 skb_priority; /* skb priority field */
383 int node; /* Memory node */
386 __u8 ipsmode; /* IPSEC mode (config) */
387 __u8 ipsproto; /* IPSEC type (config) */
399 static bool pktgen_exiting __read_mostly;
401 struct pktgen_thread {
402 spinlock_t if_lock; /* for list of devices */
403 struct list_head if_list; /* All device here */
404 struct list_head th_list;
405 struct task_struct *tsk;
408 /* Field for thread to receive "posted" events terminate,
414 wait_queue_head_t queue;
415 struct completion start_done;
421 static inline ktime_t ktime_now(void)
426 return timespec_to_ktime(ts);
429 /* This works even if 32 bit because of careful byte order choice */
430 static inline int ktime_lt(const ktime_t cmp1, const ktime_t cmp2)
432 return cmp1.tv64 < cmp2.tv64;
435 static const char version[] =
436 "Packet Generator for packet performance testing. "
437 "Version: " VERSION "\n";
439 static int pktgen_remove_device(struct pktgen_thread *t, struct pktgen_dev *i);
440 static int pktgen_add_device(struct pktgen_thread *t, const char *ifname);
441 static struct pktgen_dev *pktgen_find_dev(struct pktgen_thread *t,
442 const char *ifname, bool exact);
443 static int pktgen_device_event(struct notifier_block *, unsigned long, void *);
444 static void pktgen_run_all_threads(void);
445 static void pktgen_reset_all_threads(void);
446 static void pktgen_stop_all_threads_ifs(void);
448 static void pktgen_stop(struct pktgen_thread *t);
449 static void pktgen_clear_counters(struct pktgen_dev *pkt_dev);
451 static unsigned int scan_ip6(const char *s, char ip[16]);
453 /* Module parameters, defaults. */
454 static int pg_count_d __read_mostly = 1000;
455 static int pg_delay_d __read_mostly;
456 static int pg_clone_skb_d __read_mostly;
457 static int debug __read_mostly;
459 static DEFINE_MUTEX(pktgen_thread_lock);
460 static LIST_HEAD(pktgen_threads);
462 static struct notifier_block pktgen_notifier_block = {
463 .notifier_call = pktgen_device_event,
467 * /proc handling functions
471 static int pgctrl_show(struct seq_file *seq, void *v)
473 seq_puts(seq, version);
477 static ssize_t pgctrl_write(struct file *file, const char __user *buf,
478 size_t count, loff_t *ppos)
483 if (!capable(CAP_NET_ADMIN)) {
488 if (count > sizeof(data))
489 count = sizeof(data);
491 if (copy_from_user(data, buf, count)) {
495 data[count - 1] = 0; /* Make string */
497 if (!strcmp(data, "stop"))
498 pktgen_stop_all_threads_ifs();
500 else if (!strcmp(data, "start"))
501 pktgen_run_all_threads();
503 else if (!strcmp(data, "reset"))
504 pktgen_reset_all_threads();
507 pr_warning("Unknown command: %s\n", data);
515 static int pgctrl_open(struct inode *inode, struct file *file)
517 return single_open(file, pgctrl_show, PDE(inode)->data);
520 static const struct file_operations pktgen_fops = {
521 .owner = THIS_MODULE,
525 .write = pgctrl_write,
526 .release = single_release,
529 static int pktgen_if_show(struct seq_file *seq, void *v)
531 const struct pktgen_dev *pkt_dev = seq->private;
536 "Params: count %llu min_pkt_size: %u max_pkt_size: %u\n",
537 (unsigned long long)pkt_dev->count, pkt_dev->min_pkt_size,
538 pkt_dev->max_pkt_size);
541 " frags: %d delay: %llu clone_skb: %d ifname: %s\n",
542 pkt_dev->nfrags, (unsigned long long) pkt_dev->delay,
543 pkt_dev->clone_skb, pkt_dev->odevname);
545 seq_printf(seq, " flows: %u flowlen: %u\n", pkt_dev->cflows,
549 " queue_map_min: %u queue_map_max: %u\n",
550 pkt_dev->queue_map_min,
551 pkt_dev->queue_map_max);
553 if (pkt_dev->skb_priority)
554 seq_printf(seq, " skb_priority: %u\n",
555 pkt_dev->skb_priority);
557 if (pkt_dev->flags & F_IPV6) {
559 " saddr: %pI6c min_saddr: %pI6c max_saddr: %pI6c\n"
560 " daddr: %pI6c min_daddr: %pI6c max_daddr: %pI6c\n",
562 &pkt_dev->min_in6_saddr, &pkt_dev->max_in6_saddr,
564 &pkt_dev->min_in6_daddr, &pkt_dev->max_in6_daddr);
567 " dst_min: %s dst_max: %s\n",
568 pkt_dev->dst_min, pkt_dev->dst_max);
570 " src_min: %s src_max: %s\n",
571 pkt_dev->src_min, pkt_dev->src_max);
574 seq_puts(seq, " src_mac: ");
576 seq_printf(seq, "%pM ",
577 is_zero_ether_addr(pkt_dev->src_mac) ?
578 pkt_dev->odev->dev_addr : pkt_dev->src_mac);
580 seq_printf(seq, "dst_mac: ");
581 seq_printf(seq, "%pM\n", pkt_dev->dst_mac);
584 " udp_src_min: %d udp_src_max: %d"
585 " udp_dst_min: %d udp_dst_max: %d\n",
586 pkt_dev->udp_src_min, pkt_dev->udp_src_max,
587 pkt_dev->udp_dst_min, pkt_dev->udp_dst_max);
590 " src_mac_count: %d dst_mac_count: %d\n",
591 pkt_dev->src_mac_count, pkt_dev->dst_mac_count);
593 if (pkt_dev->nr_labels) {
595 seq_printf(seq, " mpls: ");
596 for (i = 0; i < pkt_dev->nr_labels; i++)
597 seq_printf(seq, "%08x%s", ntohl(pkt_dev->labels[i]),
598 i == pkt_dev->nr_labels-1 ? "\n" : ", ");
601 if (pkt_dev->vlan_id != 0xffff)
602 seq_printf(seq, " vlan_id: %u vlan_p: %u vlan_cfi: %u\n",
603 pkt_dev->vlan_id, pkt_dev->vlan_p,
606 if (pkt_dev->svlan_id != 0xffff)
607 seq_printf(seq, " svlan_id: %u vlan_p: %u vlan_cfi: %u\n",
608 pkt_dev->svlan_id, pkt_dev->svlan_p,
612 seq_printf(seq, " tos: 0x%02x\n", pkt_dev->tos);
614 if (pkt_dev->traffic_class)
615 seq_printf(seq, " traffic_class: 0x%02x\n", pkt_dev->traffic_class);
617 if (pkt_dev->node >= 0)
618 seq_printf(seq, " node: %d\n", pkt_dev->node);
620 seq_printf(seq, " Flags: ");
622 if (pkt_dev->flags & F_IPV6)
623 seq_printf(seq, "IPV6 ");
625 if (pkt_dev->flags & F_IPSRC_RND)
626 seq_printf(seq, "IPSRC_RND ");
628 if (pkt_dev->flags & F_IPDST_RND)
629 seq_printf(seq, "IPDST_RND ");
631 if (pkt_dev->flags & F_TXSIZE_RND)
632 seq_printf(seq, "TXSIZE_RND ");
634 if (pkt_dev->flags & F_UDPSRC_RND)
635 seq_printf(seq, "UDPSRC_RND ");
637 if (pkt_dev->flags & F_UDPDST_RND)
638 seq_printf(seq, "UDPDST_RND ");
640 if (pkt_dev->flags & F_MPLS_RND)
641 seq_printf(seq, "MPLS_RND ");
643 if (pkt_dev->flags & F_QUEUE_MAP_RND)
644 seq_printf(seq, "QUEUE_MAP_RND ");
646 if (pkt_dev->flags & F_QUEUE_MAP_CPU)
647 seq_printf(seq, "QUEUE_MAP_CPU ");
649 if (pkt_dev->cflows) {
650 if (pkt_dev->flags & F_FLOW_SEQ)
651 seq_printf(seq, "FLOW_SEQ "); /*in sequence flows*/
653 seq_printf(seq, "FLOW_RND ");
657 if (pkt_dev->flags & F_IPSEC_ON)
658 seq_printf(seq, "IPSEC ");
661 if (pkt_dev->flags & F_MACSRC_RND)
662 seq_printf(seq, "MACSRC_RND ");
664 if (pkt_dev->flags & F_MACDST_RND)
665 seq_printf(seq, "MACDST_RND ");
667 if (pkt_dev->flags & F_VID_RND)
668 seq_printf(seq, "VID_RND ");
670 if (pkt_dev->flags & F_SVID_RND)
671 seq_printf(seq, "SVID_RND ");
673 if (pkt_dev->flags & F_NODE)
674 seq_printf(seq, "NODE_ALLOC ");
678 /* not really stopped, more like last-running-at */
679 stopped = pkt_dev->running ? ktime_now() : pkt_dev->stopped_at;
680 idle = pkt_dev->idle_acc;
681 do_div(idle, NSEC_PER_USEC);
684 "Current:\n pkts-sofar: %llu errors: %llu\n",
685 (unsigned long long)pkt_dev->sofar,
686 (unsigned long long)pkt_dev->errors);
689 " started: %lluus stopped: %lluus idle: %lluus\n",
690 (unsigned long long) ktime_to_us(pkt_dev->started_at),
691 (unsigned long long) ktime_to_us(stopped),
692 (unsigned long long) idle);
695 " seq_num: %d cur_dst_mac_offset: %d cur_src_mac_offset: %d\n",
696 pkt_dev->seq_num, pkt_dev->cur_dst_mac_offset,
697 pkt_dev->cur_src_mac_offset);
699 if (pkt_dev->flags & F_IPV6) {
700 seq_printf(seq, " cur_saddr: %pI6c cur_daddr: %pI6c\n",
701 &pkt_dev->cur_in6_saddr,
702 &pkt_dev->cur_in6_daddr);
704 seq_printf(seq, " cur_saddr: 0x%x cur_daddr: 0x%x\n",
705 pkt_dev->cur_saddr, pkt_dev->cur_daddr);
707 seq_printf(seq, " cur_udp_dst: %d cur_udp_src: %d\n",
708 pkt_dev->cur_udp_dst, pkt_dev->cur_udp_src);
710 seq_printf(seq, " cur_queue_map: %u\n", pkt_dev->cur_queue_map);
712 seq_printf(seq, " flows: %u\n", pkt_dev->nflows);
714 if (pkt_dev->result[0])
715 seq_printf(seq, "Result: %s\n", pkt_dev->result);
717 seq_printf(seq, "Result: Idle\n");
723 static int hex32_arg(const char __user *user_buffer, unsigned long maxlen,
729 for (; i < maxlen; i++) {
733 if (get_user(c, &user_buffer[i]))
735 value = hex_to_bin(c);
744 static int count_trail_chars(const char __user * user_buffer,
749 for (i = 0; i < maxlen; i++) {
751 if (get_user(c, &user_buffer[i]))
769 static unsigned long num_arg(const char __user * user_buffer,
770 unsigned long maxlen, unsigned long *num)
775 for (i = 0; i < maxlen; i++) {
777 if (get_user(c, &user_buffer[i]))
779 if ((c >= '0') && (c <= '9')) {
788 static int strn_len(const char __user * user_buffer, unsigned int maxlen)
792 for (i = 0; i < maxlen; i++) {
794 if (get_user(c, &user_buffer[i]))
812 static ssize_t get_labels(const char __user *buffer, struct pktgen_dev *pkt_dev)
819 pkt_dev->nr_labels = 0;
822 len = hex32_arg(&buffer[i], 8, &tmp);
825 pkt_dev->labels[n] = htonl(tmp);
826 if (pkt_dev->labels[n] & MPLS_STACK_BOTTOM)
827 pkt_dev->flags |= F_MPLS_RND;
829 if (get_user(c, &buffer[i]))
833 if (n >= MAX_MPLS_LABELS)
837 pkt_dev->nr_labels = n;
841 static ssize_t pktgen_if_write(struct file *file,
842 const char __user * user_buffer, size_t count,
845 struct seq_file *seq = file->private_data;
846 struct pktgen_dev *pkt_dev = seq->private;
848 char name[16], valstr[32];
849 unsigned long value = 0;
850 char *pg_result = NULL;
854 pg_result = &(pkt_dev->result[0]);
857 pr_warning("wrong command format\n");
862 tmp = count_trail_chars(user_buffer, max);
864 pr_warning("illegal format\n");
869 /* Read variable name */
871 len = strn_len(&user_buffer[i], sizeof(name) - 1);
875 memset(name, 0, sizeof(name));
876 if (copy_from_user(name, &user_buffer[i], len))
881 len = count_trail_chars(&user_buffer[i], max);
888 size_t copy = min_t(size_t, count, 1023);
890 if (copy_from_user(tb, user_buffer, copy))
893 printk(KERN_DEBUG "pktgen: %s,%lu buffer -:%s:-\n", name,
894 (unsigned long)count, tb);
897 if (!strcmp(name, "min_pkt_size")) {
898 len = num_arg(&user_buffer[i], 10, &value);
903 if (value < 14 + 20 + 8)
905 if (value != pkt_dev->min_pkt_size) {
906 pkt_dev->min_pkt_size = value;
907 pkt_dev->cur_pkt_size = value;
909 sprintf(pg_result, "OK: min_pkt_size=%u",
910 pkt_dev->min_pkt_size);
914 if (!strcmp(name, "max_pkt_size")) {
915 len = num_arg(&user_buffer[i], 10, &value);
920 if (value < 14 + 20 + 8)
922 if (value != pkt_dev->max_pkt_size) {
923 pkt_dev->max_pkt_size = value;
924 pkt_dev->cur_pkt_size = value;
926 sprintf(pg_result, "OK: max_pkt_size=%u",
927 pkt_dev->max_pkt_size);
931 /* Shortcut for min = max */
933 if (!strcmp(name, "pkt_size")) {
934 len = num_arg(&user_buffer[i], 10, &value);
939 if (value < 14 + 20 + 8)
941 if (value != pkt_dev->min_pkt_size) {
942 pkt_dev->min_pkt_size = value;
943 pkt_dev->max_pkt_size = value;
944 pkt_dev->cur_pkt_size = value;
946 sprintf(pg_result, "OK: pkt_size=%u", pkt_dev->min_pkt_size);
950 if (!strcmp(name, "debug")) {
951 len = num_arg(&user_buffer[i], 10, &value);
957 sprintf(pg_result, "OK: debug=%u", debug);
961 if (!strcmp(name, "frags")) {
962 len = num_arg(&user_buffer[i], 10, &value);
967 pkt_dev->nfrags = value;
968 sprintf(pg_result, "OK: frags=%u", pkt_dev->nfrags);
971 if (!strcmp(name, "delay")) {
972 len = num_arg(&user_buffer[i], 10, &value);
977 if (value == 0x7FFFFFFF)
978 pkt_dev->delay = ULLONG_MAX;
980 pkt_dev->delay = (u64)value;
982 sprintf(pg_result, "OK: delay=%llu",
983 (unsigned long long) pkt_dev->delay);
986 if (!strcmp(name, "rate")) {
987 len = num_arg(&user_buffer[i], 10, &value);
994 pkt_dev->delay = pkt_dev->min_pkt_size*8*NSEC_PER_USEC/value;
996 pr_info("Delay set at: %llu ns\n", pkt_dev->delay);
998 sprintf(pg_result, "OK: rate=%lu", value);
1001 if (!strcmp(name, "ratep")) {
1002 len = num_arg(&user_buffer[i], 10, &value);
1009 pkt_dev->delay = NSEC_PER_SEC/value;
1011 pr_info("Delay set at: %llu ns\n", pkt_dev->delay);
1013 sprintf(pg_result, "OK: rate=%lu", value);
1016 if (!strcmp(name, "udp_src_min")) {
1017 len = num_arg(&user_buffer[i], 10, &value);
1022 if (value != pkt_dev->udp_src_min) {
1023 pkt_dev->udp_src_min = value;
1024 pkt_dev->cur_udp_src = value;
1026 sprintf(pg_result, "OK: udp_src_min=%u", pkt_dev->udp_src_min);
1029 if (!strcmp(name, "udp_dst_min")) {
1030 len = num_arg(&user_buffer[i], 10, &value);
1035 if (value != pkt_dev->udp_dst_min) {
1036 pkt_dev->udp_dst_min = value;
1037 pkt_dev->cur_udp_dst = value;
1039 sprintf(pg_result, "OK: udp_dst_min=%u", pkt_dev->udp_dst_min);
1042 if (!strcmp(name, "udp_src_max")) {
1043 len = num_arg(&user_buffer[i], 10, &value);
1048 if (value != pkt_dev->udp_src_max) {
1049 pkt_dev->udp_src_max = value;
1050 pkt_dev->cur_udp_src = value;
1052 sprintf(pg_result, "OK: udp_src_max=%u", pkt_dev->udp_src_max);
1055 if (!strcmp(name, "udp_dst_max")) {
1056 len = num_arg(&user_buffer[i], 10, &value);
1061 if (value != pkt_dev->udp_dst_max) {
1062 pkt_dev->udp_dst_max = value;
1063 pkt_dev->cur_udp_dst = value;
1065 sprintf(pg_result, "OK: udp_dst_max=%u", pkt_dev->udp_dst_max);
1068 if (!strcmp(name, "clone_skb")) {
1069 len = num_arg(&user_buffer[i], 10, &value);
1074 pkt_dev->clone_skb = value;
1076 sprintf(pg_result, "OK: clone_skb=%d", pkt_dev->clone_skb);
1079 if (!strcmp(name, "count")) {
1080 len = num_arg(&user_buffer[i], 10, &value);
1085 pkt_dev->count = value;
1086 sprintf(pg_result, "OK: count=%llu",
1087 (unsigned long long)pkt_dev->count);
1090 if (!strcmp(name, "src_mac_count")) {
1091 len = num_arg(&user_buffer[i], 10, &value);
1096 if (pkt_dev->src_mac_count != value) {
1097 pkt_dev->src_mac_count = value;
1098 pkt_dev->cur_src_mac_offset = 0;
1100 sprintf(pg_result, "OK: src_mac_count=%d",
1101 pkt_dev->src_mac_count);
1104 if (!strcmp(name, "dst_mac_count")) {
1105 len = num_arg(&user_buffer[i], 10, &value);
1110 if (pkt_dev->dst_mac_count != value) {
1111 pkt_dev->dst_mac_count = value;
1112 pkt_dev->cur_dst_mac_offset = 0;
1114 sprintf(pg_result, "OK: dst_mac_count=%d",
1115 pkt_dev->dst_mac_count);
1118 if (!strcmp(name, "node")) {
1119 len = num_arg(&user_buffer[i], 10, &value);
1125 if (node_possible(value)) {
1126 pkt_dev->node = value;
1127 sprintf(pg_result, "OK: node=%d", pkt_dev->node);
1128 if (pkt_dev->page) {
1129 put_page(pkt_dev->page);
1130 pkt_dev->page = NULL;
1134 sprintf(pg_result, "ERROR: node not possible");
1137 if (!strcmp(name, "flag")) {
1140 len = strn_len(&user_buffer[i], sizeof(f) - 1);
1144 if (copy_from_user(f, &user_buffer[i], len))
1147 if (strcmp(f, "IPSRC_RND") == 0)
1148 pkt_dev->flags |= F_IPSRC_RND;
1150 else if (strcmp(f, "!IPSRC_RND") == 0)
1151 pkt_dev->flags &= ~F_IPSRC_RND;
1153 else if (strcmp(f, "TXSIZE_RND") == 0)
1154 pkt_dev->flags |= F_TXSIZE_RND;
1156 else if (strcmp(f, "!TXSIZE_RND") == 0)
1157 pkt_dev->flags &= ~F_TXSIZE_RND;
1159 else if (strcmp(f, "IPDST_RND") == 0)
1160 pkt_dev->flags |= F_IPDST_RND;
1162 else if (strcmp(f, "!IPDST_RND") == 0)
1163 pkt_dev->flags &= ~F_IPDST_RND;
1165 else if (strcmp(f, "UDPSRC_RND") == 0)
1166 pkt_dev->flags |= F_UDPSRC_RND;
1168 else if (strcmp(f, "!UDPSRC_RND") == 0)
1169 pkt_dev->flags &= ~F_UDPSRC_RND;
1171 else if (strcmp(f, "UDPDST_RND") == 0)
1172 pkt_dev->flags |= F_UDPDST_RND;
1174 else if (strcmp(f, "!UDPDST_RND") == 0)
1175 pkt_dev->flags &= ~F_UDPDST_RND;
1177 else if (strcmp(f, "MACSRC_RND") == 0)
1178 pkt_dev->flags |= F_MACSRC_RND;
1180 else if (strcmp(f, "!MACSRC_RND") == 0)
1181 pkt_dev->flags &= ~F_MACSRC_RND;
1183 else if (strcmp(f, "MACDST_RND") == 0)
1184 pkt_dev->flags |= F_MACDST_RND;
1186 else if (strcmp(f, "!MACDST_RND") == 0)
1187 pkt_dev->flags &= ~F_MACDST_RND;
1189 else if (strcmp(f, "MPLS_RND") == 0)
1190 pkt_dev->flags |= F_MPLS_RND;
1192 else if (strcmp(f, "!MPLS_RND") == 0)
1193 pkt_dev->flags &= ~F_MPLS_RND;
1195 else if (strcmp(f, "VID_RND") == 0)
1196 pkt_dev->flags |= F_VID_RND;
1198 else if (strcmp(f, "!VID_RND") == 0)
1199 pkt_dev->flags &= ~F_VID_RND;
1201 else if (strcmp(f, "SVID_RND") == 0)
1202 pkt_dev->flags |= F_SVID_RND;
1204 else if (strcmp(f, "!SVID_RND") == 0)
1205 pkt_dev->flags &= ~F_SVID_RND;
1207 else if (strcmp(f, "FLOW_SEQ") == 0)
1208 pkt_dev->flags |= F_FLOW_SEQ;
1210 else if (strcmp(f, "QUEUE_MAP_RND") == 0)
1211 pkt_dev->flags |= F_QUEUE_MAP_RND;
1213 else if (strcmp(f, "!QUEUE_MAP_RND") == 0)
1214 pkt_dev->flags &= ~F_QUEUE_MAP_RND;
1216 else if (strcmp(f, "QUEUE_MAP_CPU") == 0)
1217 pkt_dev->flags |= F_QUEUE_MAP_CPU;
1219 else if (strcmp(f, "!QUEUE_MAP_CPU") == 0)
1220 pkt_dev->flags &= ~F_QUEUE_MAP_CPU;
1222 else if (strcmp(f, "IPSEC") == 0)
1223 pkt_dev->flags |= F_IPSEC_ON;
1226 else if (strcmp(f, "!IPV6") == 0)
1227 pkt_dev->flags &= ~F_IPV6;
1229 else if (strcmp(f, "NODE_ALLOC") == 0)
1230 pkt_dev->flags |= F_NODE;
1232 else if (strcmp(f, "!NODE_ALLOC") == 0)
1233 pkt_dev->flags &= ~F_NODE;
1237 "Flag -:%s:- unknown\nAvailable flags, (prepend ! to un-set flag):\n%s",
1239 "IPSRC_RND, IPDST_RND, UDPSRC_RND, UDPDST_RND, "
1240 "MACSRC_RND, MACDST_RND, TXSIZE_RND, IPV6, MPLS_RND, VID_RND, SVID_RND, FLOW_SEQ, IPSEC, NODE_ALLOC\n");
1243 sprintf(pg_result, "OK: flags=0x%x", pkt_dev->flags);
1246 if (!strcmp(name, "dst_min") || !strcmp(name, "dst")) {
1247 len = strn_len(&user_buffer[i], sizeof(pkt_dev->dst_min) - 1);
1251 if (copy_from_user(buf, &user_buffer[i], len))
1254 if (strcmp(buf, pkt_dev->dst_min) != 0) {
1255 memset(pkt_dev->dst_min, 0, sizeof(pkt_dev->dst_min));
1256 strncpy(pkt_dev->dst_min, buf, len);
1257 pkt_dev->daddr_min = in_aton(pkt_dev->dst_min);
1258 pkt_dev->cur_daddr = pkt_dev->daddr_min;
1261 printk(KERN_DEBUG "pktgen: dst_min set to: %s\n",
1264 sprintf(pg_result, "OK: dst_min=%s", pkt_dev->dst_min);
1267 if (!strcmp(name, "dst_max")) {
1268 len = strn_len(&user_buffer[i], sizeof(pkt_dev->dst_max) - 1);
1273 if (copy_from_user(buf, &user_buffer[i], len))
1277 if (strcmp(buf, pkt_dev->dst_max) != 0) {
1278 memset(pkt_dev->dst_max, 0, sizeof(pkt_dev->dst_max));
1279 strncpy(pkt_dev->dst_max, buf, len);
1280 pkt_dev->daddr_max = in_aton(pkt_dev->dst_max);
1281 pkt_dev->cur_daddr = pkt_dev->daddr_max;
1284 printk(KERN_DEBUG "pktgen: dst_max set to: %s\n",
1287 sprintf(pg_result, "OK: dst_max=%s", pkt_dev->dst_max);
1290 if (!strcmp(name, "dst6")) {
1291 len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1295 pkt_dev->flags |= F_IPV6;
1297 if (copy_from_user(buf, &user_buffer[i], len))
1301 scan_ip6(buf, pkt_dev->in6_daddr.s6_addr);
1302 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->in6_daddr);
1304 ipv6_addr_copy(&pkt_dev->cur_in6_daddr, &pkt_dev->in6_daddr);
1307 printk(KERN_DEBUG "pktgen: dst6 set to: %s\n", buf);
1310 sprintf(pg_result, "OK: dst6=%s", buf);
1313 if (!strcmp(name, "dst6_min")) {
1314 len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1318 pkt_dev->flags |= F_IPV6;
1320 if (copy_from_user(buf, &user_buffer[i], len))
1324 scan_ip6(buf, pkt_dev->min_in6_daddr.s6_addr);
1325 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->min_in6_daddr);
1327 ipv6_addr_copy(&pkt_dev->cur_in6_daddr,
1328 &pkt_dev->min_in6_daddr);
1330 printk(KERN_DEBUG "pktgen: dst6_min set to: %s\n", buf);
1333 sprintf(pg_result, "OK: dst6_min=%s", buf);
1336 if (!strcmp(name, "dst6_max")) {
1337 len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1341 pkt_dev->flags |= F_IPV6;
1343 if (copy_from_user(buf, &user_buffer[i], len))
1347 scan_ip6(buf, pkt_dev->max_in6_daddr.s6_addr);
1348 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->max_in6_daddr);
1351 printk(KERN_DEBUG "pktgen: dst6_max set to: %s\n", buf);
1354 sprintf(pg_result, "OK: dst6_max=%s", buf);
1357 if (!strcmp(name, "src6")) {
1358 len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1362 pkt_dev->flags |= F_IPV6;
1364 if (copy_from_user(buf, &user_buffer[i], len))
1368 scan_ip6(buf, pkt_dev->in6_saddr.s6_addr);
1369 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->in6_saddr);
1371 ipv6_addr_copy(&pkt_dev->cur_in6_saddr, &pkt_dev->in6_saddr);
1374 printk(KERN_DEBUG "pktgen: src6 set to: %s\n", buf);
1377 sprintf(pg_result, "OK: src6=%s", buf);
1380 if (!strcmp(name, "src_min")) {
1381 len = strn_len(&user_buffer[i], sizeof(pkt_dev->src_min) - 1);
1385 if (copy_from_user(buf, &user_buffer[i], len))
1388 if (strcmp(buf, pkt_dev->src_min) != 0) {
1389 memset(pkt_dev->src_min, 0, sizeof(pkt_dev->src_min));
1390 strncpy(pkt_dev->src_min, buf, len);
1391 pkt_dev->saddr_min = in_aton(pkt_dev->src_min);
1392 pkt_dev->cur_saddr = pkt_dev->saddr_min;
1395 printk(KERN_DEBUG "pktgen: src_min set to: %s\n",
1398 sprintf(pg_result, "OK: src_min=%s", pkt_dev->src_min);
1401 if (!strcmp(name, "src_max")) {
1402 len = strn_len(&user_buffer[i], sizeof(pkt_dev->src_max) - 1);
1406 if (copy_from_user(buf, &user_buffer[i], len))
1409 if (strcmp(buf, pkt_dev->src_max) != 0) {
1410 memset(pkt_dev->src_max, 0, sizeof(pkt_dev->src_max));
1411 strncpy(pkt_dev->src_max, buf, len);
1412 pkt_dev->saddr_max = in_aton(pkt_dev->src_max);
1413 pkt_dev->cur_saddr = pkt_dev->saddr_max;
1416 printk(KERN_DEBUG "pktgen: src_max set to: %s\n",
1419 sprintf(pg_result, "OK: src_max=%s", pkt_dev->src_max);
1422 if (!strcmp(name, "dst_mac")) {
1423 len = strn_len(&user_buffer[i], sizeof(valstr) - 1);
1427 memset(valstr, 0, sizeof(valstr));
1428 if (copy_from_user(valstr, &user_buffer[i], len))
1431 if (!mac_pton(valstr, pkt_dev->dst_mac))
1433 /* Set up Dest MAC */
1434 memcpy(&pkt_dev->hh[0], pkt_dev->dst_mac, ETH_ALEN);
1436 sprintf(pg_result, "OK: dstmac %pM", pkt_dev->dst_mac);
1439 if (!strcmp(name, "src_mac")) {
1440 len = strn_len(&user_buffer[i], sizeof(valstr) - 1);
1444 memset(valstr, 0, sizeof(valstr));
1445 if (copy_from_user(valstr, &user_buffer[i], len))
1448 if (!mac_pton(valstr, pkt_dev->src_mac))
1450 /* Set up Src MAC */
1451 memcpy(&pkt_dev->hh[6], pkt_dev->src_mac, ETH_ALEN);
1453 sprintf(pg_result, "OK: srcmac %pM", pkt_dev->src_mac);
1457 if (!strcmp(name, "clear_counters")) {
1458 pktgen_clear_counters(pkt_dev);
1459 sprintf(pg_result, "OK: Clearing counters.\n");
1463 if (!strcmp(name, "flows")) {
1464 len = num_arg(&user_buffer[i], 10, &value);
1469 if (value > MAX_CFLOWS)
1472 pkt_dev->cflows = value;
1473 sprintf(pg_result, "OK: flows=%u", pkt_dev->cflows);
1477 if (!strcmp(name, "flowlen")) {
1478 len = num_arg(&user_buffer[i], 10, &value);
1483 pkt_dev->lflow = value;
1484 sprintf(pg_result, "OK: flowlen=%u", pkt_dev->lflow);
1488 if (!strcmp(name, "queue_map_min")) {
1489 len = num_arg(&user_buffer[i], 5, &value);
1494 pkt_dev->queue_map_min = value;
1495 sprintf(pg_result, "OK: queue_map_min=%u", pkt_dev->queue_map_min);
1499 if (!strcmp(name, "queue_map_max")) {
1500 len = num_arg(&user_buffer[i], 5, &value);
1505 pkt_dev->queue_map_max = value;
1506 sprintf(pg_result, "OK: queue_map_max=%u", pkt_dev->queue_map_max);
1510 if (!strcmp(name, "mpls")) {
1513 len = get_labels(&user_buffer[i], pkt_dev);
1517 cnt = sprintf(pg_result, "OK: mpls=");
1518 for (n = 0; n < pkt_dev->nr_labels; n++)
1519 cnt += sprintf(pg_result + cnt,
1520 "%08x%s", ntohl(pkt_dev->labels[n]),
1521 n == pkt_dev->nr_labels-1 ? "" : ",");
1523 if (pkt_dev->nr_labels && pkt_dev->vlan_id != 0xffff) {
1524 pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1525 pkt_dev->svlan_id = 0xffff;
1528 printk(KERN_DEBUG "pktgen: VLAN/SVLAN auto turned off\n");
1533 if (!strcmp(name, "vlan_id")) {
1534 len = num_arg(&user_buffer[i], 4, &value);
1539 if (value <= 4095) {
1540 pkt_dev->vlan_id = value; /* turn on VLAN */
1543 printk(KERN_DEBUG "pktgen: VLAN turned on\n");
1545 if (debug && pkt_dev->nr_labels)
1546 printk(KERN_DEBUG "pktgen: MPLS auto turned off\n");
1548 pkt_dev->nr_labels = 0; /* turn off MPLS */
1549 sprintf(pg_result, "OK: vlan_id=%u", pkt_dev->vlan_id);
1551 pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1552 pkt_dev->svlan_id = 0xffff;
1555 printk(KERN_DEBUG "pktgen: VLAN/SVLAN turned off\n");
1560 if (!strcmp(name, "vlan_p")) {
1561 len = num_arg(&user_buffer[i], 1, &value);
1566 if ((value <= 7) && (pkt_dev->vlan_id != 0xffff)) {
1567 pkt_dev->vlan_p = value;
1568 sprintf(pg_result, "OK: vlan_p=%u", pkt_dev->vlan_p);
1570 sprintf(pg_result, "ERROR: vlan_p must be 0-7");
1575 if (!strcmp(name, "vlan_cfi")) {
1576 len = num_arg(&user_buffer[i], 1, &value);
1581 if ((value <= 1) && (pkt_dev->vlan_id != 0xffff)) {
1582 pkt_dev->vlan_cfi = value;
1583 sprintf(pg_result, "OK: vlan_cfi=%u", pkt_dev->vlan_cfi);
1585 sprintf(pg_result, "ERROR: vlan_cfi must be 0-1");
1590 if (!strcmp(name, "svlan_id")) {
1591 len = num_arg(&user_buffer[i], 4, &value);
1596 if ((value <= 4095) && ((pkt_dev->vlan_id != 0xffff))) {
1597 pkt_dev->svlan_id = value; /* turn on SVLAN */
1600 printk(KERN_DEBUG "pktgen: SVLAN turned on\n");
1602 if (debug && pkt_dev->nr_labels)
1603 printk(KERN_DEBUG "pktgen: MPLS auto turned off\n");
1605 pkt_dev->nr_labels = 0; /* turn off MPLS */
1606 sprintf(pg_result, "OK: svlan_id=%u", pkt_dev->svlan_id);
1608 pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1609 pkt_dev->svlan_id = 0xffff;
1612 printk(KERN_DEBUG "pktgen: VLAN/SVLAN turned off\n");
1617 if (!strcmp(name, "svlan_p")) {
1618 len = num_arg(&user_buffer[i], 1, &value);
1623 if ((value <= 7) && (pkt_dev->svlan_id != 0xffff)) {
1624 pkt_dev->svlan_p = value;
1625 sprintf(pg_result, "OK: svlan_p=%u", pkt_dev->svlan_p);
1627 sprintf(pg_result, "ERROR: svlan_p must be 0-7");
1632 if (!strcmp(name, "svlan_cfi")) {
1633 len = num_arg(&user_buffer[i], 1, &value);
1638 if ((value <= 1) && (pkt_dev->svlan_id != 0xffff)) {
1639 pkt_dev->svlan_cfi = value;
1640 sprintf(pg_result, "OK: svlan_cfi=%u", pkt_dev->svlan_cfi);
1642 sprintf(pg_result, "ERROR: svlan_cfi must be 0-1");
1647 if (!strcmp(name, "tos")) {
1648 __u32 tmp_value = 0;
1649 len = hex32_arg(&user_buffer[i], 2, &tmp_value);
1655 pkt_dev->tos = tmp_value;
1656 sprintf(pg_result, "OK: tos=0x%02x", pkt_dev->tos);
1658 sprintf(pg_result, "ERROR: tos must be 00-ff");
1663 if (!strcmp(name, "traffic_class")) {
1664 __u32 tmp_value = 0;
1665 len = hex32_arg(&user_buffer[i], 2, &tmp_value);
1671 pkt_dev->traffic_class = tmp_value;
1672 sprintf(pg_result, "OK: traffic_class=0x%02x", pkt_dev->traffic_class);
1674 sprintf(pg_result, "ERROR: traffic_class must be 00-ff");
1679 if (!strcmp(name, "skb_priority")) {
1680 len = num_arg(&user_buffer[i], 9, &value);
1685 pkt_dev->skb_priority = value;
1686 sprintf(pg_result, "OK: skb_priority=%i",
1687 pkt_dev->skb_priority);
1691 sprintf(pkt_dev->result, "No such parameter \"%s\"", name);
1695 static int pktgen_if_open(struct inode *inode, struct file *file)
1697 return single_open(file, pktgen_if_show, PDE(inode)->data);
1700 static const struct file_operations pktgen_if_fops = {
1701 .owner = THIS_MODULE,
1702 .open = pktgen_if_open,
1704 .llseek = seq_lseek,
1705 .write = pktgen_if_write,
1706 .release = single_release,
1709 static int pktgen_thread_show(struct seq_file *seq, void *v)
1711 struct pktgen_thread *t = seq->private;
1712 const struct pktgen_dev *pkt_dev;
1716 seq_printf(seq, "Running: ");
1719 list_for_each_entry(pkt_dev, &t->if_list, list)
1720 if (pkt_dev->running)
1721 seq_printf(seq, "%s ", pkt_dev->odevname);
1723 seq_printf(seq, "\nStopped: ");
1725 list_for_each_entry(pkt_dev, &t->if_list, list)
1726 if (!pkt_dev->running)
1727 seq_printf(seq, "%s ", pkt_dev->odevname);
1730 seq_printf(seq, "\nResult: %s\n", t->result);
1732 seq_printf(seq, "\nResult: NA\n");
1739 static ssize_t pktgen_thread_write(struct file *file,
1740 const char __user * user_buffer,
1741 size_t count, loff_t * offset)
1743 struct seq_file *seq = file->private_data;
1744 struct pktgen_thread *t = seq->private;
1745 int i, max, len, ret;
1750 // sprintf(pg_result, "Wrong command format");
1755 len = count_trail_chars(user_buffer, max);
1761 /* Read variable name */
1763 len = strn_len(&user_buffer[i], sizeof(name) - 1);
1767 memset(name, 0, sizeof(name));
1768 if (copy_from_user(name, &user_buffer[i], len))
1773 len = count_trail_chars(&user_buffer[i], max);
1780 printk(KERN_DEBUG "pktgen: t=%s, count=%lu\n",
1781 name, (unsigned long)count);
1784 pr_err("ERROR: No thread\n");
1789 pg_result = &(t->result[0]);
1791 if (!strcmp(name, "add_device")) {
1794 len = strn_len(&user_buffer[i], sizeof(f) - 1);
1799 if (copy_from_user(f, &user_buffer[i], len))
1802 mutex_lock(&pktgen_thread_lock);
1803 pktgen_add_device(t, f);
1804 mutex_unlock(&pktgen_thread_lock);
1806 sprintf(pg_result, "OK: add_device=%s", f);
1810 if (!strcmp(name, "rem_device_all")) {
1811 mutex_lock(&pktgen_thread_lock);
1812 t->control |= T_REMDEVALL;
1813 mutex_unlock(&pktgen_thread_lock);
1814 schedule_timeout_interruptible(msecs_to_jiffies(125)); /* Propagate thread->control */
1816 sprintf(pg_result, "OK: rem_device_all");
1820 if (!strcmp(name, "max_before_softirq")) {
1821 sprintf(pg_result, "OK: Note! max_before_softirq is obsoleted -- Do not use");
1831 static int pktgen_thread_open(struct inode *inode, struct file *file)
1833 return single_open(file, pktgen_thread_show, PDE(inode)->data);
1836 static const struct file_operations pktgen_thread_fops = {
1837 .owner = THIS_MODULE,
1838 .open = pktgen_thread_open,
1840 .llseek = seq_lseek,
1841 .write = pktgen_thread_write,
1842 .release = single_release,
1845 /* Think find or remove for NN */
1846 static struct pktgen_dev *__pktgen_NN_threads(const char *ifname, int remove)
1848 struct pktgen_thread *t;
1849 struct pktgen_dev *pkt_dev = NULL;
1850 bool exact = (remove == FIND);
1852 list_for_each_entry(t, &pktgen_threads, th_list) {
1853 pkt_dev = pktgen_find_dev(t, ifname, exact);
1857 pkt_dev->removal_mark = 1;
1858 t->control |= T_REMDEV;
1868 * mark a device for removal
1870 static void pktgen_mark_device(const char *ifname)
1872 struct pktgen_dev *pkt_dev = NULL;
1873 const int max_tries = 10, msec_per_try = 125;
1876 mutex_lock(&pktgen_thread_lock);
1877 pr_debug("%s: marking %s for removal\n", __func__, ifname);
1881 pkt_dev = __pktgen_NN_threads(ifname, REMOVE);
1882 if (pkt_dev == NULL)
1883 break; /* success */
1885 mutex_unlock(&pktgen_thread_lock);
1886 pr_debug("%s: waiting for %s to disappear....\n",
1888 schedule_timeout_interruptible(msecs_to_jiffies(msec_per_try));
1889 mutex_lock(&pktgen_thread_lock);
1891 if (++i >= max_tries) {
1892 pr_err("%s: timed out after waiting %d msec for device %s to be removed\n",
1893 __func__, msec_per_try * i, ifname);
1899 mutex_unlock(&pktgen_thread_lock);
1902 static void pktgen_change_name(struct net_device *dev)
1904 struct pktgen_thread *t;
1906 list_for_each_entry(t, &pktgen_threads, th_list) {
1907 struct pktgen_dev *pkt_dev;
1909 list_for_each_entry(pkt_dev, &t->if_list, list) {
1910 if (pkt_dev->odev != dev)
1913 remove_proc_entry(pkt_dev->entry->name, pg_proc_dir);
1915 pkt_dev->entry = proc_create_data(dev->name, 0600,
1919 if (!pkt_dev->entry)
1920 pr_err("can't move proc entry for '%s'\n",
1927 static int pktgen_device_event(struct notifier_block *unused,
1928 unsigned long event, void *ptr)
1930 struct net_device *dev = ptr;
1932 if (!net_eq(dev_net(dev), &init_net))
1935 /* It is OK that we do not hold the group lock right now,
1936 * as we run under the RTNL lock.
1940 case NETDEV_CHANGENAME:
1941 pktgen_change_name(dev);
1944 case NETDEV_UNREGISTER:
1945 pktgen_mark_device(dev->name);
1952 static struct net_device *pktgen_dev_get_by_name(struct pktgen_dev *pkt_dev,
1958 for (i = 0; ifname[i] != '@'; i++) {
1966 return dev_get_by_name(&init_net, b);
1970 /* Associate pktgen_dev with a device. */
1972 static int pktgen_setup_dev(struct pktgen_dev *pkt_dev, const char *ifname)
1974 struct net_device *odev;
1977 /* Clean old setups */
1978 if (pkt_dev->odev) {
1979 dev_put(pkt_dev->odev);
1980 pkt_dev->odev = NULL;
1983 odev = pktgen_dev_get_by_name(pkt_dev, ifname);
1985 pr_err("no such netdevice: \"%s\"\n", ifname);
1989 if (odev->type != ARPHRD_ETHER) {
1990 pr_err("not an ethernet device: \"%s\"\n", ifname);
1992 } else if (!netif_running(odev)) {
1993 pr_err("device is down: \"%s\"\n", ifname);
1996 pkt_dev->odev = odev;
2004 /* Read pkt_dev from the interface and set up internal pktgen_dev
2005 * structure to have the right information to create/send packets
2007 static void pktgen_setup_inject(struct pktgen_dev *pkt_dev)
2011 if (!pkt_dev->odev) {
2012 pr_err("ERROR: pkt_dev->odev == NULL in setup_inject\n");
2013 sprintf(pkt_dev->result,
2014 "ERROR: pkt_dev->odev == NULL in setup_inject.\n");
2018 /* make sure that we don't pick a non-existing transmit queue */
2019 ntxq = pkt_dev->odev->real_num_tx_queues;
2021 if (ntxq <= pkt_dev->queue_map_min) {
2022 pr_warning("WARNING: Requested queue_map_min (zero-based) (%d) exceeds valid range [0 - %d] for (%d) queues on %s, resetting\n",
2023 pkt_dev->queue_map_min, (ntxq ?: 1) - 1, ntxq,
2025 pkt_dev->queue_map_min = ntxq - 1;
2027 if (pkt_dev->queue_map_max >= ntxq) {
2028 pr_warning("WARNING: Requested queue_map_max (zero-based) (%d) exceeds valid range [0 - %d] for (%d) queues on %s, resetting\n",
2029 pkt_dev->queue_map_max, (ntxq ?: 1) - 1, ntxq,
2031 pkt_dev->queue_map_max = ntxq - 1;
2034 /* Default to the interface's mac if not explicitly set. */
2036 if (is_zero_ether_addr(pkt_dev->src_mac))
2037 memcpy(&(pkt_dev->hh[6]), pkt_dev->odev->dev_addr, ETH_ALEN);
2039 /* Set up Dest MAC */
2040 memcpy(&(pkt_dev->hh[0]), pkt_dev->dst_mac, ETH_ALEN);
2042 /* Set up pkt size */
2043 pkt_dev->cur_pkt_size = pkt_dev->min_pkt_size;
2045 if (pkt_dev->flags & F_IPV6) {
2047 * Skip this automatic address setting until locks or functions
2052 int i, set = 0, err = 1;
2053 struct inet6_dev *idev;
2055 for (i = 0; i < IN6_ADDR_HSIZE; i++)
2056 if (pkt_dev->cur_in6_saddr.s6_addr[i]) {
2064 * Use linklevel address if unconfigured.
2066 * use ipv6_get_lladdr if/when it's get exported
2070 idev = __in6_dev_get(pkt_dev->odev);
2072 struct inet6_ifaddr *ifp;
2074 read_lock_bh(&idev->lock);
2075 for (ifp = idev->addr_list; ifp;
2076 ifp = ifp->if_next) {
2077 if (ifp->scope == IFA_LINK &&
2078 !(ifp->flags & IFA_F_TENTATIVE)) {
2079 ipv6_addr_copy(&pkt_dev->
2086 read_unlock_bh(&idev->lock);
2090 pr_err("ERROR: IPv6 link address not available\n");
2094 pkt_dev->saddr_min = 0;
2095 pkt_dev->saddr_max = 0;
2096 if (strlen(pkt_dev->src_min) == 0) {
2098 struct in_device *in_dev;
2101 in_dev = __in_dev_get_rcu(pkt_dev->odev);
2103 if (in_dev->ifa_list) {
2104 pkt_dev->saddr_min =
2105 in_dev->ifa_list->ifa_address;
2106 pkt_dev->saddr_max = pkt_dev->saddr_min;
2111 pkt_dev->saddr_min = in_aton(pkt_dev->src_min);
2112 pkt_dev->saddr_max = in_aton(pkt_dev->src_max);
2115 pkt_dev->daddr_min = in_aton(pkt_dev->dst_min);
2116 pkt_dev->daddr_max = in_aton(pkt_dev->dst_max);
2118 /* Initialize current values. */
2119 pkt_dev->cur_dst_mac_offset = 0;
2120 pkt_dev->cur_src_mac_offset = 0;
2121 pkt_dev->cur_saddr = pkt_dev->saddr_min;
2122 pkt_dev->cur_daddr = pkt_dev->daddr_min;
2123 pkt_dev->cur_udp_dst = pkt_dev->udp_dst_min;
2124 pkt_dev->cur_udp_src = pkt_dev->udp_src_min;
2125 pkt_dev->nflows = 0;
2129 static void spin(struct pktgen_dev *pkt_dev, ktime_t spin_until)
2131 ktime_t start_time, end_time;
2133 struct hrtimer_sleeper t;
2135 hrtimer_init_on_stack(&t.timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS);
2136 hrtimer_set_expires(&t.timer, spin_until);
2138 remaining = ktime_to_ns(hrtimer_expires_remaining(&t.timer));
2139 if (remaining <= 0) {
2140 pkt_dev->next_tx = ktime_add_ns(spin_until, pkt_dev->delay);
2144 start_time = ktime_now();
2145 if (remaining < 100000)
2146 ndelay(remaining); /* really small just spin */
2148 /* see do_nanosleep */
2149 hrtimer_init_sleeper(&t, current);
2151 set_current_state(TASK_INTERRUPTIBLE);
2152 hrtimer_start_expires(&t.timer, HRTIMER_MODE_ABS);
2153 if (!hrtimer_active(&t.timer))
2159 hrtimer_cancel(&t.timer);
2160 } while (t.task && pkt_dev->running && !signal_pending(current));
2161 __set_current_state(TASK_RUNNING);
2163 end_time = ktime_now();
2165 pkt_dev->idle_acc += ktime_to_ns(ktime_sub(end_time, start_time));
2166 pkt_dev->next_tx = ktime_add_ns(spin_until, pkt_dev->delay);
2169 static inline void set_pkt_overhead(struct pktgen_dev *pkt_dev)
2171 pkt_dev->pkt_overhead = 0;
2172 pkt_dev->pkt_overhead += pkt_dev->nr_labels*sizeof(u32);
2173 pkt_dev->pkt_overhead += VLAN_TAG_SIZE(pkt_dev);
2174 pkt_dev->pkt_overhead += SVLAN_TAG_SIZE(pkt_dev);
2177 static inline int f_seen(const struct pktgen_dev *pkt_dev, int flow)
2179 return !!(pkt_dev->flows[flow].flags & F_INIT);
2182 static inline int f_pick(struct pktgen_dev *pkt_dev)
2184 int flow = pkt_dev->curfl;
2186 if (pkt_dev->flags & F_FLOW_SEQ) {
2187 if (pkt_dev->flows[flow].count >= pkt_dev->lflow) {
2189 pkt_dev->flows[flow].count = 0;
2190 pkt_dev->flows[flow].flags = 0;
2191 pkt_dev->curfl += 1;
2192 if (pkt_dev->curfl >= pkt_dev->cflows)
2193 pkt_dev->curfl = 0; /*reset */
2196 flow = random32() % pkt_dev->cflows;
2197 pkt_dev->curfl = flow;
2199 if (pkt_dev->flows[flow].count > pkt_dev->lflow) {
2200 pkt_dev->flows[flow].count = 0;
2201 pkt_dev->flows[flow].flags = 0;
2205 return pkt_dev->curfl;
2210 /* If there was already an IPSEC SA, we keep it as is, else
2211 * we go look for it ...
2213 #define DUMMY_MARK 0
2214 static void get_ipsec_sa(struct pktgen_dev *pkt_dev, int flow)
2216 struct xfrm_state *x = pkt_dev->flows[flow].x;
2218 /*slow path: we dont already have xfrm_state*/
2219 x = xfrm_stateonly_find(&init_net, DUMMY_MARK,
2220 (xfrm_address_t *)&pkt_dev->cur_daddr,
2221 (xfrm_address_t *)&pkt_dev->cur_saddr,
2224 pkt_dev->ipsproto, 0);
2226 pkt_dev->flows[flow].x = x;
2227 set_pkt_overhead(pkt_dev);
2228 pkt_dev->pkt_overhead += x->props.header_len;
2234 static void set_cur_queue_map(struct pktgen_dev *pkt_dev)
2237 if (pkt_dev->flags & F_QUEUE_MAP_CPU)
2238 pkt_dev->cur_queue_map = smp_processor_id();
2240 else if (pkt_dev->queue_map_min <= pkt_dev->queue_map_max) {
2242 if (pkt_dev->flags & F_QUEUE_MAP_RND) {
2244 (pkt_dev->queue_map_max -
2245 pkt_dev->queue_map_min + 1)
2246 + pkt_dev->queue_map_min;
2248 t = pkt_dev->cur_queue_map + 1;
2249 if (t > pkt_dev->queue_map_max)
2250 t = pkt_dev->queue_map_min;
2252 pkt_dev->cur_queue_map = t;
2254 pkt_dev->cur_queue_map = pkt_dev->cur_queue_map % pkt_dev->odev->real_num_tx_queues;
2257 /* Increment/randomize headers according to flags and current values
2258 * for IP src/dest, UDP src/dst port, MAC-Addr src/dst
2260 static void mod_cur_headers(struct pktgen_dev *pkt_dev)
2266 if (pkt_dev->cflows)
2267 flow = f_pick(pkt_dev);
2269 /* Deal with source MAC */
2270 if (pkt_dev->src_mac_count > 1) {
2274 if (pkt_dev->flags & F_MACSRC_RND)
2275 mc = random32() % pkt_dev->src_mac_count;
2277 mc = pkt_dev->cur_src_mac_offset++;
2278 if (pkt_dev->cur_src_mac_offset >=
2279 pkt_dev->src_mac_count)
2280 pkt_dev->cur_src_mac_offset = 0;
2283 tmp = pkt_dev->src_mac[5] + (mc & 0xFF);
2284 pkt_dev->hh[11] = tmp;
2285 tmp = (pkt_dev->src_mac[4] + ((mc >> 8) & 0xFF) + (tmp >> 8));
2286 pkt_dev->hh[10] = tmp;
2287 tmp = (pkt_dev->src_mac[3] + ((mc >> 16) & 0xFF) + (tmp >> 8));
2288 pkt_dev->hh[9] = tmp;
2289 tmp = (pkt_dev->src_mac[2] + ((mc >> 24) & 0xFF) + (tmp >> 8));
2290 pkt_dev->hh[8] = tmp;
2291 tmp = (pkt_dev->src_mac[1] + (tmp >> 8));
2292 pkt_dev->hh[7] = tmp;
2295 /* Deal with Destination MAC */
2296 if (pkt_dev->dst_mac_count > 1) {
2300 if (pkt_dev->flags & F_MACDST_RND)
2301 mc = random32() % pkt_dev->dst_mac_count;
2304 mc = pkt_dev->cur_dst_mac_offset++;
2305 if (pkt_dev->cur_dst_mac_offset >=
2306 pkt_dev->dst_mac_count) {
2307 pkt_dev->cur_dst_mac_offset = 0;
2311 tmp = pkt_dev->dst_mac[5] + (mc & 0xFF);
2312 pkt_dev->hh[5] = tmp;
2313 tmp = (pkt_dev->dst_mac[4] + ((mc >> 8) & 0xFF) + (tmp >> 8));
2314 pkt_dev->hh[4] = tmp;
2315 tmp = (pkt_dev->dst_mac[3] + ((mc >> 16) & 0xFF) + (tmp >> 8));
2316 pkt_dev->hh[3] = tmp;
2317 tmp = (pkt_dev->dst_mac[2] + ((mc >> 24) & 0xFF) + (tmp >> 8));
2318 pkt_dev->hh[2] = tmp;
2319 tmp = (pkt_dev->dst_mac[1] + (tmp >> 8));
2320 pkt_dev->hh[1] = tmp;
2323 if (pkt_dev->flags & F_MPLS_RND) {
2325 for (i = 0; i < pkt_dev->nr_labels; i++)
2326 if (pkt_dev->labels[i] & MPLS_STACK_BOTTOM)
2327 pkt_dev->labels[i] = MPLS_STACK_BOTTOM |
2328 ((__force __be32)random32() &
2332 if ((pkt_dev->flags & F_VID_RND) && (pkt_dev->vlan_id != 0xffff)) {
2333 pkt_dev->vlan_id = random32() & (4096-1);
2336 if ((pkt_dev->flags & F_SVID_RND) && (pkt_dev->svlan_id != 0xffff)) {
2337 pkt_dev->svlan_id = random32() & (4096 - 1);
2340 if (pkt_dev->udp_src_min < pkt_dev->udp_src_max) {
2341 if (pkt_dev->flags & F_UDPSRC_RND)
2342 pkt_dev->cur_udp_src = random32() %
2343 (pkt_dev->udp_src_max - pkt_dev->udp_src_min)
2344 + pkt_dev->udp_src_min;
2347 pkt_dev->cur_udp_src++;
2348 if (pkt_dev->cur_udp_src >= pkt_dev->udp_src_max)
2349 pkt_dev->cur_udp_src = pkt_dev->udp_src_min;
2353 if (pkt_dev->udp_dst_min < pkt_dev->udp_dst_max) {
2354 if (pkt_dev->flags & F_UDPDST_RND) {
2355 pkt_dev->cur_udp_dst = random32() %
2356 (pkt_dev->udp_dst_max - pkt_dev->udp_dst_min)
2357 + pkt_dev->udp_dst_min;
2359 pkt_dev->cur_udp_dst++;
2360 if (pkt_dev->cur_udp_dst >= pkt_dev->udp_dst_max)
2361 pkt_dev->cur_udp_dst = pkt_dev->udp_dst_min;
2365 if (!(pkt_dev->flags & F_IPV6)) {
2367 imn = ntohl(pkt_dev->saddr_min);
2368 imx = ntohl(pkt_dev->saddr_max);
2371 if (pkt_dev->flags & F_IPSRC_RND)
2372 t = random32() % (imx - imn) + imn;
2374 t = ntohl(pkt_dev->cur_saddr);
2380 pkt_dev->cur_saddr = htonl(t);
2383 if (pkt_dev->cflows && f_seen(pkt_dev, flow)) {
2384 pkt_dev->cur_daddr = pkt_dev->flows[flow].cur_daddr;
2386 imn = ntohl(pkt_dev->daddr_min);
2387 imx = ntohl(pkt_dev->daddr_max);
2391 if (pkt_dev->flags & F_IPDST_RND) {
2393 t = random32() % (imx - imn) + imn;
2396 while (ipv4_is_loopback(s) ||
2397 ipv4_is_multicast(s) ||
2398 ipv4_is_lbcast(s) ||
2399 ipv4_is_zeronet(s) ||
2400 ipv4_is_local_multicast(s)) {
2401 t = random32() % (imx - imn) + imn;
2404 pkt_dev->cur_daddr = s;
2406 t = ntohl(pkt_dev->cur_daddr);
2411 pkt_dev->cur_daddr = htonl(t);
2414 if (pkt_dev->cflows) {
2415 pkt_dev->flows[flow].flags |= F_INIT;
2416 pkt_dev->flows[flow].cur_daddr =
2419 if (pkt_dev->flags & F_IPSEC_ON)
2420 get_ipsec_sa(pkt_dev, flow);
2425 } else { /* IPV6 * */
2427 if (pkt_dev->min_in6_daddr.s6_addr32[0] == 0 &&
2428 pkt_dev->min_in6_daddr.s6_addr32[1] == 0 &&
2429 pkt_dev->min_in6_daddr.s6_addr32[2] == 0 &&
2430 pkt_dev->min_in6_daddr.s6_addr32[3] == 0) ;
2434 /* Only random destinations yet */
2436 for (i = 0; i < 4; i++) {
2437 pkt_dev->cur_in6_daddr.s6_addr32[i] =
2438 (((__force __be32)random32() |
2439 pkt_dev->min_in6_daddr.s6_addr32[i]) &
2440 pkt_dev->max_in6_daddr.s6_addr32[i]);
2445 if (pkt_dev->min_pkt_size < pkt_dev->max_pkt_size) {
2447 if (pkt_dev->flags & F_TXSIZE_RND) {
2449 (pkt_dev->max_pkt_size - pkt_dev->min_pkt_size)
2450 + pkt_dev->min_pkt_size;
2452 t = pkt_dev->cur_pkt_size + 1;
2453 if (t > pkt_dev->max_pkt_size)
2454 t = pkt_dev->min_pkt_size;
2456 pkt_dev->cur_pkt_size = t;
2459 set_cur_queue_map(pkt_dev);
2461 pkt_dev->flows[flow].count++;
2466 static int pktgen_output_ipsec(struct sk_buff *skb, struct pktgen_dev *pkt_dev)
2468 struct xfrm_state *x = pkt_dev->flows[pkt_dev->curfl].x;
2473 /* XXX: we dont support tunnel mode for now until
2474 * we resolve the dst issue */
2475 if (x->props.mode != XFRM_MODE_TRANSPORT)
2478 spin_lock(&x->lock);
2480 err = x->outer_mode->output(x, skb);
2483 err = x->type->output(x, skb);
2487 x->curlft.bytes += skb->len;
2488 x->curlft.packets++;
2490 spin_unlock(&x->lock);
2494 static void free_SAs(struct pktgen_dev *pkt_dev)
2496 if (pkt_dev->cflows) {
2497 /* let go of the SAs if we have them */
2499 for (i = 0; i < pkt_dev->cflows; i++) {
2500 struct xfrm_state *x = pkt_dev->flows[i].x;
2503 pkt_dev->flows[i].x = NULL;
2509 static int process_ipsec(struct pktgen_dev *pkt_dev,
2510 struct sk_buff *skb, __be16 protocol)
2512 if (pkt_dev->flags & F_IPSEC_ON) {
2513 struct xfrm_state *x = pkt_dev->flows[pkt_dev->curfl].x;
2518 nhead = x->props.header_len - skb_headroom(skb);
2520 ret = pskb_expand_head(skb, nhead, 0, GFP_ATOMIC);
2522 pr_err("Error expanding ipsec packet %d\n",
2528 /* ipsec is not expecting ll header */
2529 skb_pull(skb, ETH_HLEN);
2530 ret = pktgen_output_ipsec(skb, pkt_dev);
2532 pr_err("Error creating ipsec packet %d\n", ret);
2536 eth = (__u8 *) skb_push(skb, ETH_HLEN);
2537 memcpy(eth, pkt_dev->hh, 12);
2538 *(u16 *) ð[12] = protocol;
2548 static void mpls_push(__be32 *mpls, struct pktgen_dev *pkt_dev)
2551 for (i = 0; i < pkt_dev->nr_labels; i++)
2552 *mpls++ = pkt_dev->labels[i] & ~MPLS_STACK_BOTTOM;
2555 *mpls |= MPLS_STACK_BOTTOM;
2558 static inline __be16 build_tci(unsigned int id, unsigned int cfi,
2561 return htons(id | (cfi << 12) | (prio << 13));
2564 static void pktgen_finalize_skb(struct pktgen_dev *pkt_dev, struct sk_buff *skb,
2567 struct timeval timestamp;
2568 struct pktgen_hdr *pgh;
2570 pgh = (struct pktgen_hdr *)skb_put(skb, sizeof(*pgh));
2571 datalen -= sizeof(*pgh);
2573 if (pkt_dev->nfrags <= 0) {
2574 memset(skb_put(skb, datalen), 0, datalen);
2576 int frags = pkt_dev->nfrags;
2581 if (frags > MAX_SKB_FRAGS)
2582 frags = MAX_SKB_FRAGS;
2583 len = datalen - frags * PAGE_SIZE;
2585 memset(skb_put(skb, len), 0, len);
2586 datalen = frags * PAGE_SIZE;
2590 frag_len = (datalen/frags) < PAGE_SIZE ?
2591 (datalen/frags) : PAGE_SIZE;
2592 while (datalen > 0) {
2593 if (unlikely(!pkt_dev->page)) {
2594 int node = numa_node_id();
2596 if (pkt_dev->node >= 0 && (pkt_dev->flags & F_NODE))
2597 node = pkt_dev->node;
2598 pkt_dev->page = alloc_pages_node(node, GFP_KERNEL | __GFP_ZERO, 0);
2602 skb_shinfo(skb)->frags[i].page = pkt_dev->page;
2603 get_page(pkt_dev->page);
2604 skb_shinfo(skb)->frags[i].page_offset = 0;
2605 /*last fragment, fill rest of data*/
2606 if (i == (frags - 1))
2607 skb_shinfo(skb)->frags[i].size =
2608 (datalen < PAGE_SIZE ? datalen : PAGE_SIZE);
2610 skb_shinfo(skb)->frags[i].size = frag_len;
2611 datalen -= skb_shinfo(skb)->frags[i].size;
2612 skb->len += skb_shinfo(skb)->frags[i].size;
2613 skb->data_len += skb_shinfo(skb)->frags[i].size;
2615 skb_shinfo(skb)->nr_frags = i;
2619 /* Stamp the time, and sequence number,
2620 * convert them to network byte order
2622 pgh->pgh_magic = htonl(PKTGEN_MAGIC);
2623 pgh->seq_num = htonl(pkt_dev->seq_num);
2625 do_gettimeofday(×tamp);
2626 pgh->tv_sec = htonl(timestamp.tv_sec);
2627 pgh->tv_usec = htonl(timestamp.tv_usec);
2630 static struct sk_buff *fill_packet_ipv4(struct net_device *odev,
2631 struct pktgen_dev *pkt_dev)
2633 struct sk_buff *skb = NULL;
2635 struct udphdr *udph;
2638 __be16 protocol = htons(ETH_P_IP);
2640 __be16 *vlan_tci = NULL; /* Encapsulates priority and VLAN ID */
2641 __be16 *vlan_encapsulated_proto = NULL; /* packet type ID field (or len) for VLAN tag */
2642 __be16 *svlan_tci = NULL; /* Encapsulates priority and SVLAN ID */
2643 __be16 *svlan_encapsulated_proto = NULL; /* packet type ID field (or len) for SVLAN tag */
2646 if (pkt_dev->nr_labels)
2647 protocol = htons(ETH_P_MPLS_UC);
2649 if (pkt_dev->vlan_id != 0xffff)
2650 protocol = htons(ETH_P_8021Q);
2652 /* Update any of the values, used when we're incrementing various
2655 mod_cur_headers(pkt_dev);
2656 queue_map = pkt_dev->cur_queue_map;
2658 datalen = (odev->hard_header_len + 16) & ~0xf;
2660 if (pkt_dev->flags & F_NODE) {
2663 if (pkt_dev->node >= 0)
2664 node = pkt_dev->node;
2666 node = numa_node_id();
2668 skb = __alloc_skb(NET_SKB_PAD + pkt_dev->cur_pkt_size + 64
2669 + datalen + pkt_dev->pkt_overhead, GFP_NOWAIT, 0, node);
2671 skb_reserve(skb, NET_SKB_PAD);
2676 skb = __netdev_alloc_skb(odev,
2677 pkt_dev->cur_pkt_size + 64
2678 + datalen + pkt_dev->pkt_overhead, GFP_NOWAIT);
2681 sprintf(pkt_dev->result, "No memory");
2684 prefetchw(skb->data);
2686 skb_reserve(skb, datalen);
2688 /* Reserve for ethernet and IP header */
2689 eth = (__u8 *) skb_push(skb, 14);
2690 mpls = (__be32 *)skb_put(skb, pkt_dev->nr_labels*sizeof(__u32));
2691 if (pkt_dev->nr_labels)
2692 mpls_push(mpls, pkt_dev);
2694 if (pkt_dev->vlan_id != 0xffff) {
2695 if (pkt_dev->svlan_id != 0xffff) {
2696 svlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2697 *svlan_tci = build_tci(pkt_dev->svlan_id,
2700 svlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2701 *svlan_encapsulated_proto = htons(ETH_P_8021Q);
2703 vlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2704 *vlan_tci = build_tci(pkt_dev->vlan_id,
2707 vlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2708 *vlan_encapsulated_proto = htons(ETH_P_IP);
2711 skb->network_header = skb->tail;
2712 skb->transport_header = skb->network_header + sizeof(struct iphdr);
2713 skb_put(skb, sizeof(struct iphdr) + sizeof(struct udphdr));
2714 skb_set_queue_mapping(skb, queue_map);
2715 skb->priority = pkt_dev->skb_priority;
2718 udph = udp_hdr(skb);
2720 memcpy(eth, pkt_dev->hh, 12);
2721 *(__be16 *) & eth[12] = protocol;
2723 /* Eth + IPh + UDPh + mpls */
2724 datalen = pkt_dev->cur_pkt_size - 14 - 20 - 8 -
2725 pkt_dev->pkt_overhead;
2726 if (datalen < sizeof(struct pktgen_hdr))
2727 datalen = sizeof(struct pktgen_hdr);
2729 udph->source = htons(pkt_dev->cur_udp_src);
2730 udph->dest = htons(pkt_dev->cur_udp_dst);
2731 udph->len = htons(datalen + 8); /* DATA + udphdr */
2732 udph->check = 0; /* No checksum */
2737 iph->tos = pkt_dev->tos;
2738 iph->protocol = IPPROTO_UDP; /* UDP */
2739 iph->saddr = pkt_dev->cur_saddr;
2740 iph->daddr = pkt_dev->cur_daddr;
2741 iph->id = htons(pkt_dev->ip_id);
2744 iplen = 20 + 8 + datalen;
2745 iph->tot_len = htons(iplen);
2747 iph->check = ip_fast_csum((void *)iph, iph->ihl);
2748 skb->protocol = protocol;
2749 skb->mac_header = (skb->network_header - ETH_HLEN -
2750 pkt_dev->pkt_overhead);
2752 skb->pkt_type = PACKET_HOST;
2753 pktgen_finalize_skb(pkt_dev, skb, datalen);
2756 if (!process_ipsec(pkt_dev, skb, protocol))
2764 * scan_ip6, fmt_ip taken from dietlibc-0.21
2765 * Author Felix von Leitner <felix-dietlibc@fefe.de>
2767 * Slightly modified for kernel.
2768 * Should be candidate for net/ipv4/utils.c
2772 static unsigned int scan_ip6(const char *s, char ip[16])
2775 unsigned int len = 0;
2778 unsigned int prefixlen = 0;
2779 unsigned int suffixlen = 0;
2783 for (i = 0; i < 16; i++)
2789 if (s[1] == ':') { /* Found "::", skip to part 2 */
2797 u = simple_strtoul(s, &pos, 16);
2801 if (prefixlen == 12 && s[i] == '.') {
2803 /* the last 4 bytes may be written as IPv4 address */
2806 memcpy((struct in_addr *)(ip + 12), &tmp, sizeof(tmp));
2809 ip[prefixlen++] = (u >> 8);
2810 ip[prefixlen++] = (u & 255);
2813 if (prefixlen == 16)
2817 /* part 2, after "::" */
2824 } else if (suffixlen != 0)
2827 u = simple_strtol(s, &pos, 16);
2834 if (suffixlen + prefixlen <= 12 && s[i] == '.') {
2836 memcpy((struct in_addr *)(suffix + suffixlen), &tmp,
2842 suffix[suffixlen++] = (u >> 8);
2843 suffix[suffixlen++] = (u & 255);
2846 if (prefixlen + suffixlen == 16)
2849 for (i = 0; i < suffixlen; i++)
2850 ip[16 - suffixlen + i] = suffix[i];
2854 static struct sk_buff *fill_packet_ipv6(struct net_device *odev,
2855 struct pktgen_dev *pkt_dev)
2857 struct sk_buff *skb = NULL;
2859 struct udphdr *udph;
2861 struct ipv6hdr *iph;
2862 __be16 protocol = htons(ETH_P_IPV6);
2864 __be16 *vlan_tci = NULL; /* Encapsulates priority and VLAN ID */
2865 __be16 *vlan_encapsulated_proto = NULL; /* packet type ID field (or len) for VLAN tag */
2866 __be16 *svlan_tci = NULL; /* Encapsulates priority and SVLAN ID */
2867 __be16 *svlan_encapsulated_proto = NULL; /* packet type ID field (or len) for SVLAN tag */
2870 if (pkt_dev->nr_labels)
2871 protocol = htons(ETH_P_MPLS_UC);
2873 if (pkt_dev->vlan_id != 0xffff)
2874 protocol = htons(ETH_P_8021Q);
2876 /* Update any of the values, used when we're incrementing various
2879 mod_cur_headers(pkt_dev);
2880 queue_map = pkt_dev->cur_queue_map;
2882 skb = __netdev_alloc_skb(odev,
2883 pkt_dev->cur_pkt_size + 64
2884 + 16 + pkt_dev->pkt_overhead, GFP_NOWAIT);
2886 sprintf(pkt_dev->result, "No memory");
2889 prefetchw(skb->data);
2891 skb_reserve(skb, 16);
2893 /* Reserve for ethernet and IP header */
2894 eth = (__u8 *) skb_push(skb, 14);
2895 mpls = (__be32 *)skb_put(skb, pkt_dev->nr_labels*sizeof(__u32));
2896 if (pkt_dev->nr_labels)
2897 mpls_push(mpls, pkt_dev);
2899 if (pkt_dev->vlan_id != 0xffff) {
2900 if (pkt_dev->svlan_id != 0xffff) {
2901 svlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2902 *svlan_tci = build_tci(pkt_dev->svlan_id,
2905 svlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2906 *svlan_encapsulated_proto = htons(ETH_P_8021Q);
2908 vlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2909 *vlan_tci = build_tci(pkt_dev->vlan_id,
2912 vlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2913 *vlan_encapsulated_proto = htons(ETH_P_IPV6);
2916 skb->network_header = skb->tail;
2917 skb->transport_header = skb->network_header + sizeof(struct ipv6hdr);
2918 skb_put(skb, sizeof(struct ipv6hdr) + sizeof(struct udphdr));
2919 skb_set_queue_mapping(skb, queue_map);
2920 skb->priority = pkt_dev->skb_priority;
2921 iph = ipv6_hdr(skb);
2922 udph = udp_hdr(skb);
2924 memcpy(eth, pkt_dev->hh, 12);
2925 *(__be16 *) ð[12] = protocol;
2927 /* Eth + IPh + UDPh + mpls */
2928 datalen = pkt_dev->cur_pkt_size - 14 -
2929 sizeof(struct ipv6hdr) - sizeof(struct udphdr) -
2930 pkt_dev->pkt_overhead;
2932 if (datalen < sizeof(struct pktgen_hdr)) {
2933 datalen = sizeof(struct pktgen_hdr);
2934 if (net_ratelimit())
2935 pr_info("increased datalen to %d\n", datalen);
2938 udph->source = htons(pkt_dev->cur_udp_src);
2939 udph->dest = htons(pkt_dev->cur_udp_dst);
2940 udph->len = htons(datalen + sizeof(struct udphdr));
2941 udph->check = 0; /* No checksum */
2943 *(__be32 *) iph = htonl(0x60000000); /* Version + flow */
2945 if (pkt_dev->traffic_class) {
2946 /* Version + traffic class + flow (0) */
2947 *(__be32 *)iph |= htonl(0x60000000 | (pkt_dev->traffic_class << 20));
2950 iph->hop_limit = 32;
2952 iph->payload_len = htons(sizeof(struct udphdr) + datalen);
2953 iph->nexthdr = IPPROTO_UDP;
2955 ipv6_addr_copy(&iph->daddr, &pkt_dev->cur_in6_daddr);
2956 ipv6_addr_copy(&iph->saddr, &pkt_dev->cur_in6_saddr);
2958 skb->mac_header = (skb->network_header - ETH_HLEN -
2959 pkt_dev->pkt_overhead);
2960 skb->protocol = protocol;
2962 skb->pkt_type = PACKET_HOST;
2964 pktgen_finalize_skb(pkt_dev, skb, datalen);
2969 static struct sk_buff *fill_packet(struct net_device *odev,
2970 struct pktgen_dev *pkt_dev)
2972 if (pkt_dev->flags & F_IPV6)
2973 return fill_packet_ipv6(odev, pkt_dev);
2975 return fill_packet_ipv4(odev, pkt_dev);
2978 static void pktgen_clear_counters(struct pktgen_dev *pkt_dev)
2980 pkt_dev->seq_num = 1;
2981 pkt_dev->idle_acc = 0;
2983 pkt_dev->tx_bytes = 0;
2984 pkt_dev->errors = 0;
2987 /* Set up structure for sending pkts, clear counters */
2989 static void pktgen_run(struct pktgen_thread *t)
2991 struct pktgen_dev *pkt_dev;
2997 list_for_each_entry(pkt_dev, &t->if_list, list) {
3000 * setup odev and create initial packet.
3002 pktgen_setup_inject(pkt_dev);
3004 if (pkt_dev->odev) {
3005 pktgen_clear_counters(pkt_dev);
3006 pkt_dev->running = 1; /* Cranke yeself! */
3007 pkt_dev->skb = NULL;
3008 pkt_dev->started_at =
3009 pkt_dev->next_tx = ktime_now();
3011 set_pkt_overhead(pkt_dev);
3013 strcpy(pkt_dev->result, "Starting");
3016 strcpy(pkt_dev->result, "Error starting");
3020 t->control &= ~(T_STOP);
3023 static void pktgen_stop_all_threads_ifs(void)
3025 struct pktgen_thread *t;
3029 mutex_lock(&pktgen_thread_lock);
3031 list_for_each_entry(t, &pktgen_threads, th_list)
3032 t->control |= T_STOP;
3034 mutex_unlock(&pktgen_thread_lock);
3037 static int thread_is_running(const struct pktgen_thread *t)
3039 const struct pktgen_dev *pkt_dev;
3041 list_for_each_entry(pkt_dev, &t->if_list, list)
3042 if (pkt_dev->running)
3047 static int pktgen_wait_thread_run(struct pktgen_thread *t)
3051 while (thread_is_running(t)) {
3055 msleep_interruptible(100);
3057 if (signal_pending(current))
3067 static int pktgen_wait_all_threads_run(void)
3069 struct pktgen_thread *t;
3072 mutex_lock(&pktgen_thread_lock);
3074 list_for_each_entry(t, &pktgen_threads, th_list) {
3075 sig = pktgen_wait_thread_run(t);
3081 list_for_each_entry(t, &pktgen_threads, th_list)
3082 t->control |= (T_STOP);
3084 mutex_unlock(&pktgen_thread_lock);
3088 static void pktgen_run_all_threads(void)
3090 struct pktgen_thread *t;
3094 mutex_lock(&pktgen_thread_lock);
3096 list_for_each_entry(t, &pktgen_threads, th_list)
3097 t->control |= (T_RUN);
3099 mutex_unlock(&pktgen_thread_lock);
3101 /* Propagate thread->control */
3102 schedule_timeout_interruptible(msecs_to_jiffies(125));
3104 pktgen_wait_all_threads_run();
3107 static void pktgen_reset_all_threads(void)
3109 struct pktgen_thread *t;
3113 mutex_lock(&pktgen_thread_lock);
3115 list_for_each_entry(t, &pktgen_threads, th_list)
3116 t->control |= (T_REMDEVALL);
3118 mutex_unlock(&pktgen_thread_lock);
3120 /* Propagate thread->control */
3121 schedule_timeout_interruptible(msecs_to_jiffies(125));
3123 pktgen_wait_all_threads_run();
3126 static void show_results(struct pktgen_dev *pkt_dev, int nr_frags)
3128 __u64 bps, mbps, pps;
3129 char *p = pkt_dev->result;
3130 ktime_t elapsed = ktime_sub(pkt_dev->stopped_at,
3131 pkt_dev->started_at);
3132 ktime_t idle = ns_to_ktime(pkt_dev->idle_acc);
3134 p += sprintf(p, "OK: %llu(c%llu+d%llu) usec, %llu (%dbyte,%dfrags)\n",
3135 (unsigned long long)ktime_to_us(elapsed),
3136 (unsigned long long)ktime_to_us(ktime_sub(elapsed, idle)),
3137 (unsigned long long)ktime_to_us(idle),
3138 (unsigned long long)pkt_dev->sofar,
3139 pkt_dev->cur_pkt_size, nr_frags);
3141 pps = div64_u64(pkt_dev->sofar * NSEC_PER_SEC,
3142 ktime_to_ns(elapsed));
3144 bps = pps * 8 * pkt_dev->cur_pkt_size;
3147 do_div(mbps, 1000000);
3148 p += sprintf(p, " %llupps %lluMb/sec (%llubps) errors: %llu",
3149 (unsigned long long)pps,
3150 (unsigned long long)mbps,
3151 (unsigned long long)bps,
3152 (unsigned long long)pkt_dev->errors);
3155 /* Set stopped-at timer, remove from running list, do counters & statistics */
3156 static int pktgen_stop_device(struct pktgen_dev *pkt_dev)
3158 int nr_frags = pkt_dev->skb ? skb_shinfo(pkt_dev->skb)->nr_frags : -1;
3160 if (!pkt_dev->running) {
3161 pr_warning("interface: %s is already stopped\n",
3166 kfree_skb(pkt_dev->skb);
3167 pkt_dev->skb = NULL;
3168 pkt_dev->stopped_at = ktime_now();
3169 pkt_dev->running = 0;
3171 show_results(pkt_dev, nr_frags);
3176 static struct pktgen_dev *next_to_run(struct pktgen_thread *t)
3178 struct pktgen_dev *pkt_dev, *best = NULL;
3182 list_for_each_entry(pkt_dev, &t->if_list, list) {
3183 if (!pkt_dev->running)
3187 else if (ktime_lt(pkt_dev->next_tx, best->next_tx))
3194 static void pktgen_stop(struct pktgen_thread *t)
3196 struct pktgen_dev *pkt_dev;
3202 list_for_each_entry(pkt_dev, &t->if_list, list) {
3203 pktgen_stop_device(pkt_dev);
3210 * one of our devices needs to be removed - find it
3213 static void pktgen_rem_one_if(struct pktgen_thread *t)
3215 struct list_head *q, *n;
3216 struct pktgen_dev *cur;
3222 list_for_each_safe(q, n, &t->if_list) {
3223 cur = list_entry(q, struct pktgen_dev, list);
3225 if (!cur->removal_mark)
3228 kfree_skb(cur->skb);
3231 pktgen_remove_device(t, cur);
3239 static void pktgen_rem_all_ifs(struct pktgen_thread *t)
3241 struct list_head *q, *n;
3242 struct pktgen_dev *cur;
3246 /* Remove all devices, free mem */
3250 list_for_each_safe(q, n, &t->if_list) {
3251 cur = list_entry(q, struct pktgen_dev, list);
3253 kfree_skb(cur->skb);
3256 pktgen_remove_device(t, cur);
3262 static void pktgen_rem_thread(struct pktgen_thread *t)
3264 /* Remove from the thread list */
3266 remove_proc_entry(t->tsk->comm, pg_proc_dir);
3270 static void pktgen_resched(struct pktgen_dev *pkt_dev)
3272 ktime_t idle_start = ktime_now();
3274 pkt_dev->idle_acc += ktime_to_ns(ktime_sub(ktime_now(), idle_start));
3277 static void pktgen_wait_for_skb(struct pktgen_dev *pkt_dev)
3279 ktime_t idle_start = ktime_now();
3281 while (atomic_read(&(pkt_dev->skb->users)) != 1) {
3282 if (signal_pending(current))
3286 pktgen_resched(pkt_dev);
3290 pkt_dev->idle_acc += ktime_to_ns(ktime_sub(ktime_now(), idle_start));
3293 static void pktgen_xmit(struct pktgen_dev *pkt_dev)
3295 struct net_device *odev = pkt_dev->odev;
3296 netdev_tx_t (*xmit)(struct sk_buff *, struct net_device *)
3297 = odev->netdev_ops->ndo_start_xmit;
3298 struct netdev_queue *txq;
3302 /* If device is offline, then don't send */
3303 if (unlikely(!netif_running(odev) || !netif_carrier_ok(odev))) {
3304 pktgen_stop_device(pkt_dev);
3308 /* This is max DELAY, this has special meaning of
3311 if (unlikely(pkt_dev->delay == ULLONG_MAX)) {
3312 pkt_dev->next_tx = ktime_add_ns(ktime_now(), ULONG_MAX);
3316 /* If no skb or clone count exhausted then get new one */
3317 if (!pkt_dev->skb || (pkt_dev->last_ok &&
3318 ++pkt_dev->clone_count >= pkt_dev->clone_skb)) {
3319 /* build a new pkt */
3320 kfree_skb(pkt_dev->skb);
3322 pkt_dev->skb = fill_packet(odev, pkt_dev);
3323 if (pkt_dev->skb == NULL) {
3324 pr_err("ERROR: couldn't allocate skb in fill_packet\n");
3326 pkt_dev->clone_count--; /* back out increment, OOM */
3329 pkt_dev->last_pkt_size = pkt_dev->skb->len;
3330 pkt_dev->allocated_skbs++;
3331 pkt_dev->clone_count = 0; /* reset counter */
3334 if (pkt_dev->delay && pkt_dev->last_ok)
3335 spin(pkt_dev, pkt_dev->next_tx);
3337 queue_map = skb_get_queue_mapping(pkt_dev->skb);
3338 txq = netdev_get_tx_queue(odev, queue_map);
3340 __netif_tx_lock_bh(txq);
3342 if (unlikely(netif_tx_queue_frozen_or_stopped(txq))) {
3343 ret = NETDEV_TX_BUSY;
3344 pkt_dev->last_ok = 0;
3347 atomic_inc(&(pkt_dev->skb->users));
3348 ret = (*xmit)(pkt_dev->skb, odev);
3352 txq_trans_update(txq);
3353 pkt_dev->last_ok = 1;
3356 pkt_dev->tx_bytes += pkt_dev->last_pkt_size;
3360 case NET_XMIT_POLICED:
3361 /* skb has been consumed */
3364 default: /* Drivers are not supposed to return other values! */
3365 if (net_ratelimit())
3366 pr_info("%s xmit error: %d\n", pkt_dev->odevname, ret);
3369 case NETDEV_TX_LOCKED:
3370 case NETDEV_TX_BUSY:
3371 /* Retry it next time */
3372 atomic_dec(&(pkt_dev->skb->users));
3373 pkt_dev->last_ok = 0;
3376 __netif_tx_unlock_bh(txq);
3378 /* If pkt_dev->count is zero, then run forever */
3379 if ((pkt_dev->count != 0) && (pkt_dev->sofar >= pkt_dev->count)) {
3380 pktgen_wait_for_skb(pkt_dev);
3382 /* Done with this */
3383 pktgen_stop_device(pkt_dev);
3388 * Main loop of the thread goes here
3391 static int pktgen_thread_worker(void *arg)
3394 struct pktgen_thread *t = arg;
3395 struct pktgen_dev *pkt_dev = NULL;
3398 BUG_ON(smp_processor_id() != cpu);
3400 init_waitqueue_head(&t->queue);
3401 complete(&t->start_done);
3403 pr_debug("starting pktgen/%d: pid=%d\n", cpu, task_pid_nr(current));
3405 set_current_state(TASK_INTERRUPTIBLE);
3409 while (!kthread_should_stop()) {
3410 pkt_dev = next_to_run(t);
3412 if (unlikely(!pkt_dev && t->control == 0)) {
3415 wait_event_interruptible_timeout(t->queue,
3422 __set_current_state(TASK_RUNNING);
3424 if (likely(pkt_dev)) {
3425 pktgen_xmit(pkt_dev);
3428 pktgen_resched(pkt_dev);
3433 if (t->control & T_STOP) {
3435 t->control &= ~(T_STOP);
3438 if (t->control & T_RUN) {
3440 t->control &= ~(T_RUN);
3443 if (t->control & T_REMDEVALL) {
3444 pktgen_rem_all_ifs(t);
3445 t->control &= ~(T_REMDEVALL);
3448 if (t->control & T_REMDEV) {
3449 pktgen_rem_one_if(t);
3450 t->control &= ~(T_REMDEV);
3455 set_current_state(TASK_INTERRUPTIBLE);
3458 pr_debug("%s stopping all device\n", t->tsk->comm);
3461 pr_debug("%s removing all device\n", t->tsk->comm);
3462 pktgen_rem_all_ifs(t);
3464 pr_debug("%s removing thread\n", t->tsk->comm);
3465 pktgen_rem_thread(t);
3467 /* Wait for kthread_stop */
3468 while (!kthread_should_stop()) {
3469 set_current_state(TASK_INTERRUPTIBLE);
3472 __set_current_state(TASK_RUNNING);
3477 static struct pktgen_dev *pktgen_find_dev(struct pktgen_thread *t,
3478 const char *ifname, bool exact)
3480 struct pktgen_dev *p, *pkt_dev = NULL;
3481 size_t len = strlen(ifname);
3484 list_for_each_entry(p, &t->if_list, list)
3485 if (strncmp(p->odevname, ifname, len) == 0) {
3486 if (p->odevname[len]) {
3487 if (exact || p->odevname[len] != '@')
3495 pr_debug("find_dev(%s) returning %p\n", ifname, pkt_dev);
3500 * Adds a dev at front of if_list.
3503 static int add_dev_to_thread(struct pktgen_thread *t,
3504 struct pktgen_dev *pkt_dev)
3510 if (pkt_dev->pg_thread) {
3511 pr_err("ERROR: already assigned to a thread\n");
3516 list_add(&pkt_dev->list, &t->if_list);
3517 pkt_dev->pg_thread = t;
3518 pkt_dev->running = 0;
3525 /* Called under thread lock */
3527 static int pktgen_add_device(struct pktgen_thread *t, const char *ifname)
3529 struct pktgen_dev *pkt_dev;
3531 int node = cpu_to_node(t->cpu);
3533 /* We don't allow a device to be on several threads */
3535 pkt_dev = __pktgen_NN_threads(ifname, FIND);
3537 pr_err("ERROR: interface already used\n");
3541 pkt_dev = kzalloc_node(sizeof(struct pktgen_dev), GFP_KERNEL, node);
3545 strcpy(pkt_dev->odevname, ifname);
3546 pkt_dev->flows = vmalloc_node(MAX_CFLOWS * sizeof(struct flow_state),
3548 if (pkt_dev->flows == NULL) {
3552 memset(pkt_dev->flows, 0, MAX_CFLOWS * sizeof(struct flow_state));
3554 pkt_dev->removal_mark = 0;
3555 pkt_dev->min_pkt_size = ETH_ZLEN;
3556 pkt_dev->max_pkt_size = ETH_ZLEN;
3557 pkt_dev->nfrags = 0;
3558 pkt_dev->clone_skb = pg_clone_skb_d;
3559 pkt_dev->delay = pg_delay_d;
3560 pkt_dev->count = pg_count_d;
3562 pkt_dev->udp_src_min = 9; /* sink port */
3563 pkt_dev->udp_src_max = 9;
3564 pkt_dev->udp_dst_min = 9;
3565 pkt_dev->udp_dst_max = 9;
3567 pkt_dev->vlan_p = 0;
3568 pkt_dev->vlan_cfi = 0;
3569 pkt_dev->vlan_id = 0xffff;
3570 pkt_dev->svlan_p = 0;
3571 pkt_dev->svlan_cfi = 0;
3572 pkt_dev->svlan_id = 0xffff;
3575 err = pktgen_setup_dev(pkt_dev, ifname);
3579 pkt_dev->entry = proc_create_data(ifname, 0600, pg_proc_dir,
3580 &pktgen_if_fops, pkt_dev);
3581 if (!pkt_dev->entry) {
3582 pr_err("cannot create %s/%s procfs entry\n",
3583 PG_PROC_DIR, ifname);
3588 pkt_dev->ipsmode = XFRM_MODE_TRANSPORT;
3589 pkt_dev->ipsproto = IPPROTO_ESP;
3592 return add_dev_to_thread(t, pkt_dev);
3594 dev_put(pkt_dev->odev);
3599 vfree(pkt_dev->flows);
3604 static int __init pktgen_create_thread(int cpu)
3606 struct pktgen_thread *t;
3607 struct proc_dir_entry *pe;
3608 struct task_struct *p;
3610 t = kzalloc_node(sizeof(struct pktgen_thread), GFP_KERNEL,
3613 pr_err("ERROR: out of memory, can't create new thread\n");
3617 spin_lock_init(&t->if_lock);
3620 INIT_LIST_HEAD(&t->if_list);
3622 list_add_tail(&t->th_list, &pktgen_threads);
3623 init_completion(&t->start_done);
3625 p = kthread_create_on_node(pktgen_thread_worker,
3628 "kpktgend_%d", cpu);
3630 pr_err("kernel_thread() failed for cpu %d\n", t->cpu);
3631 list_del(&t->th_list);
3635 kthread_bind(p, cpu);
3638 pe = proc_create_data(t->tsk->comm, 0600, pg_proc_dir,
3639 &pktgen_thread_fops, t);
3641 pr_err("cannot create %s/%s procfs entry\n",
3642 PG_PROC_DIR, t->tsk->comm);
3644 list_del(&t->th_list);
3650 wait_for_completion(&t->start_done);
3656 * Removes a device from the thread if_list.
3658 static void _rem_dev_from_if_list(struct pktgen_thread *t,
3659 struct pktgen_dev *pkt_dev)
3661 struct list_head *q, *n;
3662 struct pktgen_dev *p;
3664 list_for_each_safe(q, n, &t->if_list) {
3665 p = list_entry(q, struct pktgen_dev, list);
3671 static int pktgen_remove_device(struct pktgen_thread *t,
3672 struct pktgen_dev *pkt_dev)
3675 pr_debug("remove_device pkt_dev=%p\n", pkt_dev);
3677 if (pkt_dev->running) {
3678 pr_warning("WARNING: trying to remove a running interface, stopping it now\n");
3679 pktgen_stop_device(pkt_dev);
3682 /* Dis-associate from the interface */
3684 if (pkt_dev->odev) {
3685 dev_put(pkt_dev->odev);
3686 pkt_dev->odev = NULL;
3689 /* And update the thread if_list */
3691 _rem_dev_from_if_list(t, pkt_dev);
3694 remove_proc_entry(pkt_dev->entry->name, pg_proc_dir);
3699 vfree(pkt_dev->flows);
3701 put_page(pkt_dev->page);
3706 static int __init pg_init(void)
3709 struct proc_dir_entry *pe;
3711 pr_info("%s", version);
3713 pg_proc_dir = proc_mkdir(PG_PROC_DIR, init_net.proc_net);
3717 pe = proc_create(PGCTRL, 0600, pg_proc_dir, &pktgen_fops);
3719 pr_err("ERROR: cannot create %s procfs entry\n", PGCTRL);
3720 proc_net_remove(&init_net, PG_PROC_DIR);
3724 /* Register us to receive netdevice events */
3725 register_netdevice_notifier(&pktgen_notifier_block);
3727 for_each_online_cpu(cpu) {
3730 err = pktgen_create_thread(cpu);
3732 pr_warning("WARNING: Cannot create thread for cpu %d (%d)\n",
3736 if (list_empty(&pktgen_threads)) {
3737 pr_err("ERROR: Initialization failed for all threads\n");
3738 unregister_netdevice_notifier(&pktgen_notifier_block);
3739 remove_proc_entry(PGCTRL, pg_proc_dir);
3740 proc_net_remove(&init_net, PG_PROC_DIR);
3747 static void __exit pg_cleanup(void)
3749 struct pktgen_thread *t;
3750 struct list_head *q, *n;
3752 /* Stop all interfaces & threads */
3753 pktgen_exiting = true;
3755 list_for_each_safe(q, n, &pktgen_threads) {
3756 t = list_entry(q, struct pktgen_thread, th_list);
3757 kthread_stop(t->tsk);
3761 /* Un-register us from receiving netdevice events */
3762 unregister_netdevice_notifier(&pktgen_notifier_block);
3764 /* Clean up proc file system */
3765 remove_proc_entry(PGCTRL, pg_proc_dir);
3766 proc_net_remove(&init_net, PG_PROC_DIR);
3769 module_init(pg_init);
3770 module_exit(pg_cleanup);
3772 MODULE_AUTHOR("Robert Olsson <robert.olsson@its.uu.se>");
3773 MODULE_DESCRIPTION("Packet Generator tool");
3774 MODULE_LICENSE("GPL");
3775 MODULE_VERSION(VERSION);
3776 module_param(pg_count_d, int, 0);
3777 MODULE_PARM_DESC(pg_count_d, "Default number of packets to inject");
3778 module_param(pg_delay_d, int, 0);
3779 MODULE_PARM_DESC(pg_delay_d, "Default delay between packets (nanoseconds)");
3780 module_param(pg_clone_skb_d, int, 0);
3781 MODULE_PARM_DESC(pg_clone_skb_d, "Default number of copies of the same packet");
3782 module_param(debug, int, 0);
3783 MODULE_PARM_DESC(debug, "Enable debugging of pktgen module");