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
73 * The if_list is RCU protected, and the if_lock remains to protect updating
74 * of if_list, from "add_device" as it invoked from userspace (via proc write).
76 * By design there should only be *one* "controlling" process. In practice
77 * multiple write accesses gives unpredictable result. Understood by "write"
78 * to /proc gives result code thats should be read be the "writer".
79 * For practical use this should be no problem.
81 * Note when adding devices to a specific CPU there good idea to also assign
82 * /proc/irq/XX/smp_affinity so TX-interrupts gets bound to the same CPU.
85 * Fix refcount off by one if first packet fails, potential null deref,
88 * First "ranges" functionality for ipv6 030726 --ro
90 * Included flow support. 030802 ANK.
92 * Fixed unaligned access on IA-64 Grant Grundler <grundler@parisc-linux.org>
94 * Remove if fix from added Harald Welte <laforge@netfilter.org> 040419
95 * ia64 compilation fix from Aron Griffis <aron@hp.com> 040604
97 * New xmit() return, do_div and misc clean up by Stephen Hemminger
98 * <shemminger@osdl.org> 040923
100 * Randy Dunlap fixed u64 printk compiler waring
102 * Remove FCS from BW calculation. Lennert Buytenhek <buytenh@wantstofly.org>
103 * New time handling. Lennert Buytenhek <buytenh@wantstofly.org> 041213
105 * Corrections from Nikolai Malykh (nmalykh@bilim.com)
106 * Removed unused flags F_SET_SRCMAC & F_SET_SRCIP 041230
108 * interruptible_sleep_on_timeout() replaced Nishanth Aravamudan <nacc@us.ibm.com>
111 * MPLS support by Steven Whitehouse <steve@chygwyn.com>
113 * 802.1Q/Q-in-Q support by Francesco Fondelli (FF) <francesco.fondelli@gmail.com>
115 * Fixed src_mac command to set source mac of packet to value specified in
116 * command by Adit Ranadive <adit.262@gmail.com>
120 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
122 #include <linux/sys.h>
123 #include <linux/types.h>
124 #include <linux/module.h>
125 #include <linux/moduleparam.h>
126 #include <linux/kernel.h>
127 #include <linux/mutex.h>
128 #include <linux/sched.h>
129 #include <linux/slab.h>
130 #include <linux/vmalloc.h>
131 #include <linux/unistd.h>
132 #include <linux/string.h>
133 #include <linux/ptrace.h>
134 #include <linux/errno.h>
135 #include <linux/ioport.h>
136 #include <linux/interrupt.h>
137 #include <linux/capability.h>
138 #include <linux/hrtimer.h>
139 #include <linux/freezer.h>
140 #include <linux/delay.h>
141 #include <linux/timer.h>
142 #include <linux/list.h>
143 #include <linux/init.h>
144 #include <linux/skbuff.h>
145 #include <linux/netdevice.h>
146 #include <linux/inet.h>
147 #include <linux/inetdevice.h>
148 #include <linux/rtnetlink.h>
149 #include <linux/if_arp.h>
150 #include <linux/if_vlan.h>
151 #include <linux/in.h>
152 #include <linux/ip.h>
153 #include <linux/ipv6.h>
154 #include <linux/udp.h>
155 #include <linux/proc_fs.h>
156 #include <linux/seq_file.h>
157 #include <linux/wait.h>
158 #include <linux/etherdevice.h>
159 #include <linux/kthread.h>
160 #include <linux/prefetch.h>
161 #include <net/net_namespace.h>
162 #include <net/checksum.h>
163 #include <net/ipv6.h>
165 #include <net/ip6_checksum.h>
166 #include <net/addrconf.h>
168 #include <net/xfrm.h>
170 #include <net/netns/generic.h>
171 #include <asm/byteorder.h>
172 #include <linux/rcupdate.h>
173 #include <linux/bitops.h>
174 #include <linux/io.h>
175 #include <linux/timex.h>
176 #include <linux/uaccess.h>
178 #include <asm/div64.h> /* do_div */
180 #define VERSION "2.74"
181 #define IP_NAME_SZ 32
182 #define MAX_MPLS_LABELS 16 /* This is the max label stack depth */
183 #define MPLS_STACK_BOTTOM htonl(0x00000100)
185 #define func_enter() pr_debug("entering %s\n", __func__);
187 /* Device flag bits */
188 #define F_IPSRC_RND (1<<0) /* IP-Src Random */
189 #define F_IPDST_RND (1<<1) /* IP-Dst Random */
190 #define F_UDPSRC_RND (1<<2) /* UDP-Src Random */
191 #define F_UDPDST_RND (1<<3) /* UDP-Dst Random */
192 #define F_MACSRC_RND (1<<4) /* MAC-Src Random */
193 #define F_MACDST_RND (1<<5) /* MAC-Dst Random */
194 #define F_TXSIZE_RND (1<<6) /* Transmit size is random */
195 #define F_IPV6 (1<<7) /* Interface in IPV6 Mode */
196 #define F_MPLS_RND (1<<8) /* Random MPLS labels */
197 #define F_VID_RND (1<<9) /* Random VLAN ID */
198 #define F_SVID_RND (1<<10) /* Random SVLAN ID */
199 #define F_FLOW_SEQ (1<<11) /* Sequential flows */
200 #define F_IPSEC_ON (1<<12) /* ipsec on for flows */
201 #define F_QUEUE_MAP_RND (1<<13) /* queue map Random */
202 #define F_QUEUE_MAP_CPU (1<<14) /* queue map mirrors smp_processor_id() */
203 #define F_NODE (1<<15) /* Node memory alloc*/
204 #define F_UDPCSUM (1<<16) /* Include UDP checksum */
205 #define F_NO_TIMESTAMP (1<<17) /* Don't timestamp packets (default TS) */
207 /* Thread control flag bits */
208 #define T_STOP (1<<0) /* Stop run */
209 #define T_RUN (1<<1) /* Start run */
210 #define T_REMDEVALL (1<<2) /* Remove all devs */
211 #define T_REMDEV (1<<3) /* Remove one dev */
213 /* If lock -- protects updating of if_list */
214 #define if_lock(t) spin_lock(&(t->if_lock));
215 #define if_unlock(t) spin_unlock(&(t->if_lock));
217 /* Used to help with determining the pkts on receive */
218 #define PKTGEN_MAGIC 0xbe9be955
219 #define PG_PROC_DIR "pktgen"
220 #define PGCTRL "pgctrl"
222 #define MAX_CFLOWS 65536
224 #define VLAN_TAG_SIZE(x) ((x)->vlan_id == 0xffff ? 0 : 4)
225 #define SVLAN_TAG_SIZE(x) ((x)->svlan_id == 0xffff ? 0 : 4)
231 struct xfrm_state *x;
237 #define F_INIT (1<<0) /* flow has been initialized */
241 * Try to keep frequent/infrequent used vars. separated.
243 struct proc_dir_entry *entry; /* proc file */
244 struct pktgen_thread *pg_thread;/* the owner */
245 struct list_head list; /* chaining in the thread's run-queue */
246 struct rcu_head rcu; /* freed by RCU */
248 int running; /* if false, the test will stop */
250 /* If min != max, then we will either do a linear iteration, or
251 * we will do a random selection from within the range.
254 int removal_mark; /* non-zero => the device is marked for
255 * removal by worker thread */
259 int pkt_overhead; /* overhead for MPLS, VLANs, IPSEC etc */
262 u64 delay; /* nano-seconds */
264 __u64 count; /* Default No packets to send */
265 __u64 sofar; /* How many pkts we've sent so far */
266 __u64 tx_bytes; /* How many bytes we've transmitted */
267 __u64 errors; /* Errors when trying to transmit, */
269 /* runtime counters relating to clone_skb */
271 __u64 allocated_skbs;
273 int last_ok; /* Was last skb sent?
274 * Or a failed transmit of some sort?
275 * This will keep sequence numbers in order
280 u64 idle_acc; /* nano-seconds */
285 * Use multiple SKBs during packet gen.
286 * If this number is greater than 1, then
287 * that many copies of the same packet will be
288 * sent before a new packet is allocated.
289 * If you want to send 1024 identical packets
290 * before creating a new packet,
291 * set clone_skb to 1024.
294 char dst_min[IP_NAME_SZ]; /* IP, ie 1.2.3.4 */
295 char dst_max[IP_NAME_SZ]; /* IP, ie 1.2.3.4 */
296 char src_min[IP_NAME_SZ]; /* IP, ie 1.2.3.4 */
297 char src_max[IP_NAME_SZ]; /* IP, ie 1.2.3.4 */
299 struct in6_addr in6_saddr;
300 struct in6_addr in6_daddr;
301 struct in6_addr cur_in6_daddr;
302 struct in6_addr cur_in6_saddr;
304 struct in6_addr min_in6_daddr;
305 struct in6_addr max_in6_daddr;
306 struct in6_addr min_in6_saddr;
307 struct in6_addr max_in6_saddr;
309 /* If we're doing ranges, random or incremental, then this
310 * defines the min/max for those ranges.
312 __be32 saddr_min; /* inclusive, source IP address */
313 __be32 saddr_max; /* exclusive, source IP address */
314 __be32 daddr_min; /* inclusive, dest IP address */
315 __be32 daddr_max; /* exclusive, dest IP address */
317 __u16 udp_src_min; /* inclusive, source UDP port */
318 __u16 udp_src_max; /* exclusive, source UDP port */
319 __u16 udp_dst_min; /* inclusive, dest UDP port */
320 __u16 udp_dst_max; /* exclusive, dest UDP port */
323 __u8 tos; /* six MSB of (former) IPv4 TOS
324 are for dscp codepoint */
325 __u8 traffic_class; /* ditto for the (former) Traffic Class in IPv6
326 (see RFC 3260, sec. 4) */
329 unsigned int nr_labels; /* Depth of stack, 0 = no MPLS */
330 __be32 labels[MAX_MPLS_LABELS];
332 /* VLAN/SVLAN (802.1Q/Q-in-Q) */
335 __u16 vlan_id; /* 0xffff means no vlan tag */
339 __u16 svlan_id; /* 0xffff means no svlan tag */
341 __u32 src_mac_count; /* How many MACs to iterate through */
342 __u32 dst_mac_count; /* How many MACs to iterate through */
344 unsigned char dst_mac[ETH_ALEN];
345 unsigned char src_mac[ETH_ALEN];
347 __u32 cur_dst_mac_offset;
348 __u32 cur_src_mac_offset;
360 0x00, 0x80, 0xC8, 0x79, 0xB3, 0xCB,
362 We fill in SRC address later
363 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
367 __u16 pad; /* pad out the hh struct to an even 16 bytes */
369 struct sk_buff *skb; /* skb we are to transmit next, used for when we
370 * are transmitting the same one multiple times
372 struct net_device *odev; /* The out-going device.
373 * Note that the device should have it's
374 * pg_info pointer pointing back to this
376 * Set when the user specifies the out-going
377 * device name (not when the inject is
378 * started as it used to do.)
381 struct flow_state *flows;
382 unsigned int cflows; /* Concurrent flows (config) */
383 unsigned int lflow; /* Flow length (config) */
384 unsigned int nflows; /* accumulated flows (stats) */
385 unsigned int curfl; /* current sequenced flow (state)*/
389 __u32 skb_priority; /* skb priority field */
390 int node; /* Memory node */
393 __u8 ipsmode; /* IPSEC mode (config) */
394 __u8 ipsproto; /* IPSEC type (config) */
396 struct dst_entry dst;
397 struct dst_ops dstops;
410 static int pg_net_id __read_mostly;
414 struct proc_dir_entry *proc_dir;
415 struct list_head pktgen_threads;
419 struct pktgen_thread {
420 spinlock_t if_lock; /* for list of devices */
421 struct list_head if_list; /* All device here */
422 struct list_head th_list;
423 struct task_struct *tsk;
426 /* Field for thread to receive "posted" events terminate,
432 wait_queue_head_t queue;
433 struct completion start_done;
434 struct pktgen_net *net;
440 static const char version[] =
441 "Packet Generator for packet performance testing. "
442 "Version: " VERSION "\n";
444 static int pktgen_remove_device(struct pktgen_thread *t, struct pktgen_dev *i);
445 static int pktgen_add_device(struct pktgen_thread *t, const char *ifname);
446 static struct pktgen_dev *pktgen_find_dev(struct pktgen_thread *t,
447 const char *ifname, bool exact);
448 static int pktgen_device_event(struct notifier_block *, unsigned long, void *);
449 static void pktgen_run_all_threads(struct pktgen_net *pn);
450 static void pktgen_reset_all_threads(struct pktgen_net *pn);
451 static void pktgen_stop_all_threads_ifs(struct pktgen_net *pn);
453 static void pktgen_stop(struct pktgen_thread *t);
454 static void pktgen_clear_counters(struct pktgen_dev *pkt_dev);
456 /* Module parameters, defaults. */
457 static int pg_count_d __read_mostly = 1000;
458 static int pg_delay_d __read_mostly;
459 static int pg_clone_skb_d __read_mostly;
460 static int debug __read_mostly;
462 static DEFINE_MUTEX(pktgen_thread_lock);
464 static struct notifier_block pktgen_notifier_block = {
465 .notifier_call = pktgen_device_event,
469 * /proc handling functions
473 static int pgctrl_show(struct seq_file *seq, void *v)
475 seq_puts(seq, version);
479 static ssize_t pgctrl_write(struct file *file, const char __user *buf,
480 size_t count, loff_t *ppos)
483 struct pktgen_net *pn = net_generic(current->nsproxy->net_ns, pg_net_id);
485 if (!capable(CAP_NET_ADMIN))
491 if (count > sizeof(data))
492 count = sizeof(data);
494 if (copy_from_user(data, buf, count))
497 data[count - 1] = 0; /* Strip trailing '\n' and terminate string */
499 if (!strcmp(data, "stop"))
500 pktgen_stop_all_threads_ifs(pn);
502 else if (!strcmp(data, "start"))
503 pktgen_run_all_threads(pn);
505 else if (!strcmp(data, "reset"))
506 pktgen_reset_all_threads(pn);
509 pr_warning("Unknown command: %s\n", data);
514 static int pgctrl_open(struct inode *inode, struct file *file)
516 return single_open(file, pgctrl_show, PDE_DATA(inode));
519 static const struct file_operations pktgen_fops = {
520 .owner = THIS_MODULE,
524 .write = pgctrl_write,
525 .release = single_release,
528 static int pktgen_if_show(struct seq_file *seq, void *v)
530 const struct pktgen_dev *pkt_dev = seq->private;
535 "Params: count %llu min_pkt_size: %u max_pkt_size: %u\n",
536 (unsigned long long)pkt_dev->count, pkt_dev->min_pkt_size,
537 pkt_dev->max_pkt_size);
540 " frags: %d delay: %llu clone_skb: %d ifname: %s\n",
541 pkt_dev->nfrags, (unsigned long long) pkt_dev->delay,
542 pkt_dev->clone_skb, pkt_dev->odevname);
544 seq_printf(seq, " flows: %u flowlen: %u\n", pkt_dev->cflows,
548 " queue_map_min: %u queue_map_max: %u\n",
549 pkt_dev->queue_map_min,
550 pkt_dev->queue_map_max);
552 if (pkt_dev->skb_priority)
553 seq_printf(seq, " skb_priority: %u\n",
554 pkt_dev->skb_priority);
556 if (pkt_dev->flags & F_IPV6) {
558 " saddr: %pI6c min_saddr: %pI6c max_saddr: %pI6c\n"
559 " daddr: %pI6c min_daddr: %pI6c max_daddr: %pI6c\n",
561 &pkt_dev->min_in6_saddr, &pkt_dev->max_in6_saddr,
563 &pkt_dev->min_in6_daddr, &pkt_dev->max_in6_daddr);
566 " dst_min: %s dst_max: %s\n",
567 pkt_dev->dst_min, pkt_dev->dst_max);
569 " src_min: %s src_max: %s\n",
570 pkt_dev->src_min, pkt_dev->src_max);
573 seq_puts(seq, " src_mac: ");
575 seq_printf(seq, "%pM ",
576 is_zero_ether_addr(pkt_dev->src_mac) ?
577 pkt_dev->odev->dev_addr : pkt_dev->src_mac);
579 seq_puts(seq, "dst_mac: ");
580 seq_printf(seq, "%pM\n", pkt_dev->dst_mac);
583 " udp_src_min: %d udp_src_max: %d"
584 " udp_dst_min: %d udp_dst_max: %d\n",
585 pkt_dev->udp_src_min, pkt_dev->udp_src_max,
586 pkt_dev->udp_dst_min, pkt_dev->udp_dst_max);
589 " src_mac_count: %d dst_mac_count: %d\n",
590 pkt_dev->src_mac_count, pkt_dev->dst_mac_count);
592 if (pkt_dev->nr_labels) {
594 seq_puts(seq, " mpls: ");
595 for (i = 0; i < pkt_dev->nr_labels; i++)
596 seq_printf(seq, "%08x%s", ntohl(pkt_dev->labels[i]),
597 i == pkt_dev->nr_labels-1 ? "\n" : ", ");
600 if (pkt_dev->vlan_id != 0xffff)
601 seq_printf(seq, " vlan_id: %u vlan_p: %u vlan_cfi: %u\n",
602 pkt_dev->vlan_id, pkt_dev->vlan_p,
605 if (pkt_dev->svlan_id != 0xffff)
606 seq_printf(seq, " svlan_id: %u vlan_p: %u vlan_cfi: %u\n",
607 pkt_dev->svlan_id, pkt_dev->svlan_p,
611 seq_printf(seq, " tos: 0x%02x\n", pkt_dev->tos);
613 if (pkt_dev->traffic_class)
614 seq_printf(seq, " traffic_class: 0x%02x\n", pkt_dev->traffic_class);
616 if (pkt_dev->node >= 0)
617 seq_printf(seq, " node: %d\n", pkt_dev->node);
619 seq_puts(seq, " Flags: ");
621 if (pkt_dev->flags & F_IPV6)
622 seq_puts(seq, "IPV6 ");
624 if (pkt_dev->flags & F_IPSRC_RND)
625 seq_puts(seq, "IPSRC_RND ");
627 if (pkt_dev->flags & F_IPDST_RND)
628 seq_puts(seq, "IPDST_RND ");
630 if (pkt_dev->flags & F_TXSIZE_RND)
631 seq_puts(seq, "TXSIZE_RND ");
633 if (pkt_dev->flags & F_UDPSRC_RND)
634 seq_puts(seq, "UDPSRC_RND ");
636 if (pkt_dev->flags & F_UDPDST_RND)
637 seq_puts(seq, "UDPDST_RND ");
639 if (pkt_dev->flags & F_UDPCSUM)
640 seq_puts(seq, "UDPCSUM ");
642 if (pkt_dev->flags & F_NO_TIMESTAMP)
643 seq_puts(seq, "NO_TIMESTAMP ");
645 if (pkt_dev->flags & F_MPLS_RND)
646 seq_puts(seq, "MPLS_RND ");
648 if (pkt_dev->flags & F_QUEUE_MAP_RND)
649 seq_puts(seq, "QUEUE_MAP_RND ");
651 if (pkt_dev->flags & F_QUEUE_MAP_CPU)
652 seq_puts(seq, "QUEUE_MAP_CPU ");
654 if (pkt_dev->cflows) {
655 if (pkt_dev->flags & F_FLOW_SEQ)
656 seq_puts(seq, "FLOW_SEQ "); /*in sequence flows*/
658 seq_puts(seq, "FLOW_RND ");
662 if (pkt_dev->flags & F_IPSEC_ON) {
663 seq_puts(seq, "IPSEC ");
665 seq_printf(seq, "spi:%u", pkt_dev->spi);
669 if (pkt_dev->flags & F_MACSRC_RND)
670 seq_puts(seq, "MACSRC_RND ");
672 if (pkt_dev->flags & F_MACDST_RND)
673 seq_puts(seq, "MACDST_RND ");
675 if (pkt_dev->flags & F_VID_RND)
676 seq_puts(seq, "VID_RND ");
678 if (pkt_dev->flags & F_SVID_RND)
679 seq_puts(seq, "SVID_RND ");
681 if (pkt_dev->flags & F_NODE)
682 seq_puts(seq, "NODE_ALLOC ");
686 /* not really stopped, more like last-running-at */
687 stopped = pkt_dev->running ? ktime_get() : pkt_dev->stopped_at;
688 idle = pkt_dev->idle_acc;
689 do_div(idle, NSEC_PER_USEC);
692 "Current:\n pkts-sofar: %llu errors: %llu\n",
693 (unsigned long long)pkt_dev->sofar,
694 (unsigned long long)pkt_dev->errors);
697 " started: %lluus stopped: %lluus idle: %lluus\n",
698 (unsigned long long) ktime_to_us(pkt_dev->started_at),
699 (unsigned long long) ktime_to_us(stopped),
700 (unsigned long long) idle);
703 " seq_num: %d cur_dst_mac_offset: %d cur_src_mac_offset: %d\n",
704 pkt_dev->seq_num, pkt_dev->cur_dst_mac_offset,
705 pkt_dev->cur_src_mac_offset);
707 if (pkt_dev->flags & F_IPV6) {
708 seq_printf(seq, " cur_saddr: %pI6c cur_daddr: %pI6c\n",
709 &pkt_dev->cur_in6_saddr,
710 &pkt_dev->cur_in6_daddr);
712 seq_printf(seq, " cur_saddr: %pI4 cur_daddr: %pI4\n",
713 &pkt_dev->cur_saddr, &pkt_dev->cur_daddr);
715 seq_printf(seq, " cur_udp_dst: %d cur_udp_src: %d\n",
716 pkt_dev->cur_udp_dst, pkt_dev->cur_udp_src);
718 seq_printf(seq, " cur_queue_map: %u\n", pkt_dev->cur_queue_map);
720 seq_printf(seq, " flows: %u\n", pkt_dev->nflows);
722 if (pkt_dev->result[0])
723 seq_printf(seq, "Result: %s\n", pkt_dev->result);
725 seq_puts(seq, "Result: Idle\n");
731 static int hex32_arg(const char __user *user_buffer, unsigned long maxlen,
737 for (; i < maxlen; i++) {
741 if (get_user(c, &user_buffer[i]))
743 value = hex_to_bin(c);
752 static int count_trail_chars(const char __user * user_buffer,
757 for (i = 0; i < maxlen; i++) {
759 if (get_user(c, &user_buffer[i]))
777 static long num_arg(const char __user *user_buffer, unsigned long maxlen,
783 for (i = 0; i < maxlen; i++) {
785 if (get_user(c, &user_buffer[i]))
787 if ((c >= '0') && (c <= '9')) {
796 static int strn_len(const char __user * user_buffer, unsigned int maxlen)
800 for (i = 0; i < maxlen; i++) {
802 if (get_user(c, &user_buffer[i]))
819 static ssize_t get_labels(const char __user *buffer, struct pktgen_dev *pkt_dev)
826 pkt_dev->nr_labels = 0;
829 len = hex32_arg(&buffer[i], 8, &tmp);
832 pkt_dev->labels[n] = htonl(tmp);
833 if (pkt_dev->labels[n] & MPLS_STACK_BOTTOM)
834 pkt_dev->flags |= F_MPLS_RND;
836 if (get_user(c, &buffer[i]))
840 if (n >= MAX_MPLS_LABELS)
844 pkt_dev->nr_labels = n;
848 static ssize_t pktgen_if_write(struct file *file,
849 const char __user * user_buffer, size_t count,
852 struct seq_file *seq = file->private_data;
853 struct pktgen_dev *pkt_dev = seq->private;
855 char name[16], valstr[32];
856 unsigned long value = 0;
857 char *pg_result = NULL;
861 pg_result = &(pkt_dev->result[0]);
864 pr_warning("wrong command format\n");
869 tmp = count_trail_chars(user_buffer, max);
871 pr_warning("illegal format\n");
876 /* Read variable name */
878 len = strn_len(&user_buffer[i], sizeof(name) - 1);
882 memset(name, 0, sizeof(name));
883 if (copy_from_user(name, &user_buffer[i], len))
888 len = count_trail_chars(&user_buffer[i], max);
895 size_t copy = min_t(size_t, count, 1023);
897 if (copy_from_user(tb, user_buffer, copy))
900 pr_debug("%s,%lu buffer -:%s:-\n",
901 name, (unsigned long)count, tb);
904 if (!strcmp(name, "min_pkt_size")) {
905 len = num_arg(&user_buffer[i], 10, &value);
910 if (value < 14 + 20 + 8)
912 if (value != pkt_dev->min_pkt_size) {
913 pkt_dev->min_pkt_size = value;
914 pkt_dev->cur_pkt_size = value;
916 sprintf(pg_result, "OK: min_pkt_size=%u",
917 pkt_dev->min_pkt_size);
921 if (!strcmp(name, "max_pkt_size")) {
922 len = num_arg(&user_buffer[i], 10, &value);
927 if (value < 14 + 20 + 8)
929 if (value != pkt_dev->max_pkt_size) {
930 pkt_dev->max_pkt_size = value;
931 pkt_dev->cur_pkt_size = value;
933 sprintf(pg_result, "OK: max_pkt_size=%u",
934 pkt_dev->max_pkt_size);
938 /* Shortcut for min = max */
940 if (!strcmp(name, "pkt_size")) {
941 len = num_arg(&user_buffer[i], 10, &value);
946 if (value < 14 + 20 + 8)
948 if (value != pkt_dev->min_pkt_size) {
949 pkt_dev->min_pkt_size = value;
950 pkt_dev->max_pkt_size = value;
951 pkt_dev->cur_pkt_size = value;
953 sprintf(pg_result, "OK: pkt_size=%u", pkt_dev->min_pkt_size);
957 if (!strcmp(name, "debug")) {
958 len = num_arg(&user_buffer[i], 10, &value);
964 sprintf(pg_result, "OK: debug=%u", debug);
968 if (!strcmp(name, "frags")) {
969 len = num_arg(&user_buffer[i], 10, &value);
974 pkt_dev->nfrags = value;
975 sprintf(pg_result, "OK: frags=%u", pkt_dev->nfrags);
978 if (!strcmp(name, "delay")) {
979 len = num_arg(&user_buffer[i], 10, &value);
984 if (value == 0x7FFFFFFF)
985 pkt_dev->delay = ULLONG_MAX;
987 pkt_dev->delay = (u64)value;
989 sprintf(pg_result, "OK: delay=%llu",
990 (unsigned long long) pkt_dev->delay);
993 if (!strcmp(name, "rate")) {
994 len = num_arg(&user_buffer[i], 10, &value);
1001 pkt_dev->delay = pkt_dev->min_pkt_size*8*NSEC_PER_USEC/value;
1003 pr_info("Delay set at: %llu ns\n", pkt_dev->delay);
1005 sprintf(pg_result, "OK: rate=%lu", value);
1008 if (!strcmp(name, "ratep")) {
1009 len = num_arg(&user_buffer[i], 10, &value);
1016 pkt_dev->delay = NSEC_PER_SEC/value;
1018 pr_info("Delay set at: %llu ns\n", pkt_dev->delay);
1020 sprintf(pg_result, "OK: rate=%lu", value);
1023 if (!strcmp(name, "udp_src_min")) {
1024 len = num_arg(&user_buffer[i], 10, &value);
1029 if (value != pkt_dev->udp_src_min) {
1030 pkt_dev->udp_src_min = value;
1031 pkt_dev->cur_udp_src = value;
1033 sprintf(pg_result, "OK: udp_src_min=%u", pkt_dev->udp_src_min);
1036 if (!strcmp(name, "udp_dst_min")) {
1037 len = num_arg(&user_buffer[i], 10, &value);
1042 if (value != pkt_dev->udp_dst_min) {
1043 pkt_dev->udp_dst_min = value;
1044 pkt_dev->cur_udp_dst = value;
1046 sprintf(pg_result, "OK: udp_dst_min=%u", pkt_dev->udp_dst_min);
1049 if (!strcmp(name, "udp_src_max")) {
1050 len = num_arg(&user_buffer[i], 10, &value);
1055 if (value != pkt_dev->udp_src_max) {
1056 pkt_dev->udp_src_max = value;
1057 pkt_dev->cur_udp_src = value;
1059 sprintf(pg_result, "OK: udp_src_max=%u", pkt_dev->udp_src_max);
1062 if (!strcmp(name, "udp_dst_max")) {
1063 len = num_arg(&user_buffer[i], 10, &value);
1068 if (value != pkt_dev->udp_dst_max) {
1069 pkt_dev->udp_dst_max = value;
1070 pkt_dev->cur_udp_dst = value;
1072 sprintf(pg_result, "OK: udp_dst_max=%u", pkt_dev->udp_dst_max);
1075 if (!strcmp(name, "clone_skb")) {
1076 len = num_arg(&user_buffer[i], 10, &value);
1080 (!(pkt_dev->odev->priv_flags & IFF_TX_SKB_SHARING)))
1083 pkt_dev->clone_skb = value;
1085 sprintf(pg_result, "OK: clone_skb=%d", pkt_dev->clone_skb);
1088 if (!strcmp(name, "count")) {
1089 len = num_arg(&user_buffer[i], 10, &value);
1094 pkt_dev->count = value;
1095 sprintf(pg_result, "OK: count=%llu",
1096 (unsigned long long)pkt_dev->count);
1099 if (!strcmp(name, "src_mac_count")) {
1100 len = num_arg(&user_buffer[i], 10, &value);
1105 if (pkt_dev->src_mac_count != value) {
1106 pkt_dev->src_mac_count = value;
1107 pkt_dev->cur_src_mac_offset = 0;
1109 sprintf(pg_result, "OK: src_mac_count=%d",
1110 pkt_dev->src_mac_count);
1113 if (!strcmp(name, "dst_mac_count")) {
1114 len = num_arg(&user_buffer[i], 10, &value);
1119 if (pkt_dev->dst_mac_count != value) {
1120 pkt_dev->dst_mac_count = value;
1121 pkt_dev->cur_dst_mac_offset = 0;
1123 sprintf(pg_result, "OK: dst_mac_count=%d",
1124 pkt_dev->dst_mac_count);
1127 if (!strcmp(name, "node")) {
1128 len = num_arg(&user_buffer[i], 10, &value);
1134 if (node_possible(value)) {
1135 pkt_dev->node = value;
1136 sprintf(pg_result, "OK: node=%d", pkt_dev->node);
1137 if (pkt_dev->page) {
1138 put_page(pkt_dev->page);
1139 pkt_dev->page = NULL;
1143 sprintf(pg_result, "ERROR: node not possible");
1146 if (!strcmp(name, "flag")) {
1149 len = strn_len(&user_buffer[i], sizeof(f) - 1);
1153 if (copy_from_user(f, &user_buffer[i], len))
1156 if (strcmp(f, "IPSRC_RND") == 0)
1157 pkt_dev->flags |= F_IPSRC_RND;
1159 else if (strcmp(f, "!IPSRC_RND") == 0)
1160 pkt_dev->flags &= ~F_IPSRC_RND;
1162 else if (strcmp(f, "TXSIZE_RND") == 0)
1163 pkt_dev->flags |= F_TXSIZE_RND;
1165 else if (strcmp(f, "!TXSIZE_RND") == 0)
1166 pkt_dev->flags &= ~F_TXSIZE_RND;
1168 else if (strcmp(f, "IPDST_RND") == 0)
1169 pkt_dev->flags |= F_IPDST_RND;
1171 else if (strcmp(f, "!IPDST_RND") == 0)
1172 pkt_dev->flags &= ~F_IPDST_RND;
1174 else if (strcmp(f, "UDPSRC_RND") == 0)
1175 pkt_dev->flags |= F_UDPSRC_RND;
1177 else if (strcmp(f, "!UDPSRC_RND") == 0)
1178 pkt_dev->flags &= ~F_UDPSRC_RND;
1180 else if (strcmp(f, "UDPDST_RND") == 0)
1181 pkt_dev->flags |= F_UDPDST_RND;
1183 else if (strcmp(f, "!UDPDST_RND") == 0)
1184 pkt_dev->flags &= ~F_UDPDST_RND;
1186 else if (strcmp(f, "MACSRC_RND") == 0)
1187 pkt_dev->flags |= F_MACSRC_RND;
1189 else if (strcmp(f, "!MACSRC_RND") == 0)
1190 pkt_dev->flags &= ~F_MACSRC_RND;
1192 else if (strcmp(f, "MACDST_RND") == 0)
1193 pkt_dev->flags |= F_MACDST_RND;
1195 else if (strcmp(f, "!MACDST_RND") == 0)
1196 pkt_dev->flags &= ~F_MACDST_RND;
1198 else if (strcmp(f, "MPLS_RND") == 0)
1199 pkt_dev->flags |= F_MPLS_RND;
1201 else if (strcmp(f, "!MPLS_RND") == 0)
1202 pkt_dev->flags &= ~F_MPLS_RND;
1204 else if (strcmp(f, "VID_RND") == 0)
1205 pkt_dev->flags |= F_VID_RND;
1207 else if (strcmp(f, "!VID_RND") == 0)
1208 pkt_dev->flags &= ~F_VID_RND;
1210 else if (strcmp(f, "SVID_RND") == 0)
1211 pkt_dev->flags |= F_SVID_RND;
1213 else if (strcmp(f, "!SVID_RND") == 0)
1214 pkt_dev->flags &= ~F_SVID_RND;
1216 else if (strcmp(f, "FLOW_SEQ") == 0)
1217 pkt_dev->flags |= F_FLOW_SEQ;
1219 else if (strcmp(f, "QUEUE_MAP_RND") == 0)
1220 pkt_dev->flags |= F_QUEUE_MAP_RND;
1222 else if (strcmp(f, "!QUEUE_MAP_RND") == 0)
1223 pkt_dev->flags &= ~F_QUEUE_MAP_RND;
1225 else if (strcmp(f, "QUEUE_MAP_CPU") == 0)
1226 pkt_dev->flags |= F_QUEUE_MAP_CPU;
1228 else if (strcmp(f, "!QUEUE_MAP_CPU") == 0)
1229 pkt_dev->flags &= ~F_QUEUE_MAP_CPU;
1231 else if (strcmp(f, "IPSEC") == 0)
1232 pkt_dev->flags |= F_IPSEC_ON;
1235 else if (strcmp(f, "!IPV6") == 0)
1236 pkt_dev->flags &= ~F_IPV6;
1238 else if (strcmp(f, "NODE_ALLOC") == 0)
1239 pkt_dev->flags |= F_NODE;
1241 else if (strcmp(f, "!NODE_ALLOC") == 0)
1242 pkt_dev->flags &= ~F_NODE;
1244 else if (strcmp(f, "UDPCSUM") == 0)
1245 pkt_dev->flags |= F_UDPCSUM;
1247 else if (strcmp(f, "!UDPCSUM") == 0)
1248 pkt_dev->flags &= ~F_UDPCSUM;
1250 else if (strcmp(f, "NO_TIMESTAMP") == 0)
1251 pkt_dev->flags |= F_NO_TIMESTAMP;
1255 "Flag -:%s:- unknown\nAvailable flags, (prepend ! to un-set flag):\n%s",
1257 "IPSRC_RND, IPDST_RND, UDPSRC_RND, UDPDST_RND, "
1258 "MACSRC_RND, MACDST_RND, TXSIZE_RND, IPV6, "
1259 "MPLS_RND, VID_RND, SVID_RND, FLOW_SEQ, "
1260 "QUEUE_MAP_RND, QUEUE_MAP_CPU, UDPCSUM, "
1268 sprintf(pg_result, "OK: flags=0x%x", pkt_dev->flags);
1271 if (!strcmp(name, "dst_min") || !strcmp(name, "dst")) {
1272 len = strn_len(&user_buffer[i], sizeof(pkt_dev->dst_min) - 1);
1276 if (copy_from_user(buf, &user_buffer[i], len))
1279 if (strcmp(buf, pkt_dev->dst_min) != 0) {
1280 memset(pkt_dev->dst_min, 0, sizeof(pkt_dev->dst_min));
1281 strncpy(pkt_dev->dst_min, buf, len);
1282 pkt_dev->daddr_min = in_aton(pkt_dev->dst_min);
1283 pkt_dev->cur_daddr = pkt_dev->daddr_min;
1286 pr_debug("dst_min set to: %s\n", pkt_dev->dst_min);
1288 sprintf(pg_result, "OK: dst_min=%s", pkt_dev->dst_min);
1291 if (!strcmp(name, "dst_max")) {
1292 len = strn_len(&user_buffer[i], sizeof(pkt_dev->dst_max) - 1);
1297 if (copy_from_user(buf, &user_buffer[i], len))
1301 if (strcmp(buf, pkt_dev->dst_max) != 0) {
1302 memset(pkt_dev->dst_max, 0, sizeof(pkt_dev->dst_max));
1303 strncpy(pkt_dev->dst_max, buf, len);
1304 pkt_dev->daddr_max = in_aton(pkt_dev->dst_max);
1305 pkt_dev->cur_daddr = pkt_dev->daddr_max;
1308 pr_debug("dst_max set to: %s\n", pkt_dev->dst_max);
1310 sprintf(pg_result, "OK: dst_max=%s", pkt_dev->dst_max);
1313 if (!strcmp(name, "dst6")) {
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 in6_pton(buf, -1, pkt_dev->in6_daddr.s6_addr, -1, NULL);
1325 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->in6_daddr);
1327 pkt_dev->cur_in6_daddr = pkt_dev->in6_daddr;
1330 pr_debug("dst6 set to: %s\n", buf);
1333 sprintf(pg_result, "OK: dst6=%s", buf);
1336 if (!strcmp(name, "dst6_min")) {
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 in6_pton(buf, -1, pkt_dev->min_in6_daddr.s6_addr, -1, NULL);
1348 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->min_in6_daddr);
1350 pkt_dev->cur_in6_daddr = pkt_dev->min_in6_daddr;
1352 pr_debug("dst6_min set to: %s\n", buf);
1355 sprintf(pg_result, "OK: dst6_min=%s", buf);
1358 if (!strcmp(name, "dst6_max")) {
1359 len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1363 pkt_dev->flags |= F_IPV6;
1365 if (copy_from_user(buf, &user_buffer[i], len))
1369 in6_pton(buf, -1, pkt_dev->max_in6_daddr.s6_addr, -1, NULL);
1370 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->max_in6_daddr);
1373 pr_debug("dst6_max set to: %s\n", buf);
1376 sprintf(pg_result, "OK: dst6_max=%s", buf);
1379 if (!strcmp(name, "src6")) {
1380 len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1384 pkt_dev->flags |= F_IPV6;
1386 if (copy_from_user(buf, &user_buffer[i], len))
1390 in6_pton(buf, -1, pkt_dev->in6_saddr.s6_addr, -1, NULL);
1391 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->in6_saddr);
1393 pkt_dev->cur_in6_saddr = pkt_dev->in6_saddr;
1396 pr_debug("src6 set to: %s\n", buf);
1399 sprintf(pg_result, "OK: src6=%s", buf);
1402 if (!strcmp(name, "src_min")) {
1403 len = strn_len(&user_buffer[i], sizeof(pkt_dev->src_min) - 1);
1407 if (copy_from_user(buf, &user_buffer[i], len))
1410 if (strcmp(buf, pkt_dev->src_min) != 0) {
1411 memset(pkt_dev->src_min, 0, sizeof(pkt_dev->src_min));
1412 strncpy(pkt_dev->src_min, buf, len);
1413 pkt_dev->saddr_min = in_aton(pkt_dev->src_min);
1414 pkt_dev->cur_saddr = pkt_dev->saddr_min;
1417 pr_debug("src_min set to: %s\n", pkt_dev->src_min);
1419 sprintf(pg_result, "OK: src_min=%s", pkt_dev->src_min);
1422 if (!strcmp(name, "src_max")) {
1423 len = strn_len(&user_buffer[i], sizeof(pkt_dev->src_max) - 1);
1427 if (copy_from_user(buf, &user_buffer[i], len))
1430 if (strcmp(buf, pkt_dev->src_max) != 0) {
1431 memset(pkt_dev->src_max, 0, sizeof(pkt_dev->src_max));
1432 strncpy(pkt_dev->src_max, buf, len);
1433 pkt_dev->saddr_max = in_aton(pkt_dev->src_max);
1434 pkt_dev->cur_saddr = pkt_dev->saddr_max;
1437 pr_debug("src_max set to: %s\n", pkt_dev->src_max);
1439 sprintf(pg_result, "OK: src_max=%s", pkt_dev->src_max);
1442 if (!strcmp(name, "dst_mac")) {
1443 len = strn_len(&user_buffer[i], sizeof(valstr) - 1);
1447 memset(valstr, 0, sizeof(valstr));
1448 if (copy_from_user(valstr, &user_buffer[i], len))
1451 if (!mac_pton(valstr, pkt_dev->dst_mac))
1453 /* Set up Dest MAC */
1454 ether_addr_copy(&pkt_dev->hh[0], pkt_dev->dst_mac);
1456 sprintf(pg_result, "OK: dstmac %pM", pkt_dev->dst_mac);
1459 if (!strcmp(name, "src_mac")) {
1460 len = strn_len(&user_buffer[i], sizeof(valstr) - 1);
1464 memset(valstr, 0, sizeof(valstr));
1465 if (copy_from_user(valstr, &user_buffer[i], len))
1468 if (!mac_pton(valstr, pkt_dev->src_mac))
1470 /* Set up Src MAC */
1471 ether_addr_copy(&pkt_dev->hh[6], pkt_dev->src_mac);
1473 sprintf(pg_result, "OK: srcmac %pM", pkt_dev->src_mac);
1477 if (!strcmp(name, "clear_counters")) {
1478 pktgen_clear_counters(pkt_dev);
1479 sprintf(pg_result, "OK: Clearing counters.\n");
1483 if (!strcmp(name, "flows")) {
1484 len = num_arg(&user_buffer[i], 10, &value);
1489 if (value > MAX_CFLOWS)
1492 pkt_dev->cflows = value;
1493 sprintf(pg_result, "OK: flows=%u", pkt_dev->cflows);
1497 if (!strcmp(name, "spi")) {
1498 len = num_arg(&user_buffer[i], 10, &value);
1503 pkt_dev->spi = value;
1504 sprintf(pg_result, "OK: spi=%u", pkt_dev->spi);
1508 if (!strcmp(name, "flowlen")) {
1509 len = num_arg(&user_buffer[i], 10, &value);
1514 pkt_dev->lflow = value;
1515 sprintf(pg_result, "OK: flowlen=%u", pkt_dev->lflow);
1519 if (!strcmp(name, "queue_map_min")) {
1520 len = num_arg(&user_buffer[i], 5, &value);
1525 pkt_dev->queue_map_min = value;
1526 sprintf(pg_result, "OK: queue_map_min=%u", pkt_dev->queue_map_min);
1530 if (!strcmp(name, "queue_map_max")) {
1531 len = num_arg(&user_buffer[i], 5, &value);
1536 pkt_dev->queue_map_max = value;
1537 sprintf(pg_result, "OK: queue_map_max=%u", pkt_dev->queue_map_max);
1541 if (!strcmp(name, "mpls")) {
1542 unsigned int n, cnt;
1544 len = get_labels(&user_buffer[i], pkt_dev);
1548 cnt = sprintf(pg_result, "OK: mpls=");
1549 for (n = 0; n < pkt_dev->nr_labels; n++)
1550 cnt += sprintf(pg_result + cnt,
1551 "%08x%s", ntohl(pkt_dev->labels[n]),
1552 n == pkt_dev->nr_labels-1 ? "" : ",");
1554 if (pkt_dev->nr_labels && pkt_dev->vlan_id != 0xffff) {
1555 pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1556 pkt_dev->svlan_id = 0xffff;
1559 pr_debug("VLAN/SVLAN auto turned off\n");
1564 if (!strcmp(name, "vlan_id")) {
1565 len = num_arg(&user_buffer[i], 4, &value);
1570 if (value <= 4095) {
1571 pkt_dev->vlan_id = value; /* turn on VLAN */
1574 pr_debug("VLAN turned on\n");
1576 if (debug && pkt_dev->nr_labels)
1577 pr_debug("MPLS auto turned off\n");
1579 pkt_dev->nr_labels = 0; /* turn off MPLS */
1580 sprintf(pg_result, "OK: vlan_id=%u", pkt_dev->vlan_id);
1582 pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1583 pkt_dev->svlan_id = 0xffff;
1586 pr_debug("VLAN/SVLAN turned off\n");
1591 if (!strcmp(name, "vlan_p")) {
1592 len = num_arg(&user_buffer[i], 1, &value);
1597 if ((value <= 7) && (pkt_dev->vlan_id != 0xffff)) {
1598 pkt_dev->vlan_p = value;
1599 sprintf(pg_result, "OK: vlan_p=%u", pkt_dev->vlan_p);
1601 sprintf(pg_result, "ERROR: vlan_p must be 0-7");
1606 if (!strcmp(name, "vlan_cfi")) {
1607 len = num_arg(&user_buffer[i], 1, &value);
1612 if ((value <= 1) && (pkt_dev->vlan_id != 0xffff)) {
1613 pkt_dev->vlan_cfi = value;
1614 sprintf(pg_result, "OK: vlan_cfi=%u", pkt_dev->vlan_cfi);
1616 sprintf(pg_result, "ERROR: vlan_cfi must be 0-1");
1621 if (!strcmp(name, "svlan_id")) {
1622 len = num_arg(&user_buffer[i], 4, &value);
1627 if ((value <= 4095) && ((pkt_dev->vlan_id != 0xffff))) {
1628 pkt_dev->svlan_id = value; /* turn on SVLAN */
1631 pr_debug("SVLAN turned on\n");
1633 if (debug && pkt_dev->nr_labels)
1634 pr_debug("MPLS auto turned off\n");
1636 pkt_dev->nr_labels = 0; /* turn off MPLS */
1637 sprintf(pg_result, "OK: svlan_id=%u", pkt_dev->svlan_id);
1639 pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1640 pkt_dev->svlan_id = 0xffff;
1643 pr_debug("VLAN/SVLAN turned off\n");
1648 if (!strcmp(name, "svlan_p")) {
1649 len = num_arg(&user_buffer[i], 1, &value);
1654 if ((value <= 7) && (pkt_dev->svlan_id != 0xffff)) {
1655 pkt_dev->svlan_p = value;
1656 sprintf(pg_result, "OK: svlan_p=%u", pkt_dev->svlan_p);
1658 sprintf(pg_result, "ERROR: svlan_p must be 0-7");
1663 if (!strcmp(name, "svlan_cfi")) {
1664 len = num_arg(&user_buffer[i], 1, &value);
1669 if ((value <= 1) && (pkt_dev->svlan_id != 0xffff)) {
1670 pkt_dev->svlan_cfi = value;
1671 sprintf(pg_result, "OK: svlan_cfi=%u", pkt_dev->svlan_cfi);
1673 sprintf(pg_result, "ERROR: svlan_cfi must be 0-1");
1678 if (!strcmp(name, "tos")) {
1679 __u32 tmp_value = 0;
1680 len = hex32_arg(&user_buffer[i], 2, &tmp_value);
1686 pkt_dev->tos = tmp_value;
1687 sprintf(pg_result, "OK: tos=0x%02x", pkt_dev->tos);
1689 sprintf(pg_result, "ERROR: tos must be 00-ff");
1694 if (!strcmp(name, "traffic_class")) {
1695 __u32 tmp_value = 0;
1696 len = hex32_arg(&user_buffer[i], 2, &tmp_value);
1702 pkt_dev->traffic_class = tmp_value;
1703 sprintf(pg_result, "OK: traffic_class=0x%02x", pkt_dev->traffic_class);
1705 sprintf(pg_result, "ERROR: traffic_class must be 00-ff");
1710 if (!strcmp(name, "skb_priority")) {
1711 len = num_arg(&user_buffer[i], 9, &value);
1716 pkt_dev->skb_priority = value;
1717 sprintf(pg_result, "OK: skb_priority=%i",
1718 pkt_dev->skb_priority);
1722 sprintf(pkt_dev->result, "No such parameter \"%s\"", name);
1726 static int pktgen_if_open(struct inode *inode, struct file *file)
1728 return single_open(file, pktgen_if_show, PDE_DATA(inode));
1731 static const struct file_operations pktgen_if_fops = {
1732 .owner = THIS_MODULE,
1733 .open = pktgen_if_open,
1735 .llseek = seq_lseek,
1736 .write = pktgen_if_write,
1737 .release = single_release,
1740 static int pktgen_thread_show(struct seq_file *seq, void *v)
1742 struct pktgen_thread *t = seq->private;
1743 const struct pktgen_dev *pkt_dev;
1747 seq_puts(seq, "Running: ");
1750 list_for_each_entry_rcu(pkt_dev, &t->if_list, list)
1751 if (pkt_dev->running)
1752 seq_printf(seq, "%s ", pkt_dev->odevname);
1754 seq_puts(seq, "\nStopped: ");
1756 list_for_each_entry_rcu(pkt_dev, &t->if_list, list)
1757 if (!pkt_dev->running)
1758 seq_printf(seq, "%s ", pkt_dev->odevname);
1761 seq_printf(seq, "\nResult: %s\n", t->result);
1763 seq_puts(seq, "\nResult: NA\n");
1770 static ssize_t pktgen_thread_write(struct file *file,
1771 const char __user * user_buffer,
1772 size_t count, loff_t * offset)
1774 struct seq_file *seq = file->private_data;
1775 struct pktgen_thread *t = seq->private;
1776 int i, max, len, ret;
1781 // sprintf(pg_result, "Wrong command format");
1786 len = count_trail_chars(user_buffer, max);
1792 /* Read variable name */
1794 len = strn_len(&user_buffer[i], sizeof(name) - 1);
1798 memset(name, 0, sizeof(name));
1799 if (copy_from_user(name, &user_buffer[i], len))
1804 len = count_trail_chars(&user_buffer[i], max);
1811 pr_debug("t=%s, count=%lu\n", name, (unsigned long)count);
1814 pr_err("ERROR: No thread\n");
1819 pg_result = &(t->result[0]);
1821 if (!strcmp(name, "add_device")) {
1824 len = strn_len(&user_buffer[i], sizeof(f) - 1);
1829 if (copy_from_user(f, &user_buffer[i], len))
1832 mutex_lock(&pktgen_thread_lock);
1833 ret = pktgen_add_device(t, f);
1834 mutex_unlock(&pktgen_thread_lock);
1837 sprintf(pg_result, "OK: add_device=%s", f);
1839 sprintf(pg_result, "ERROR: can not add device %s", f);
1843 if (!strcmp(name, "rem_device_all")) {
1844 mutex_lock(&pktgen_thread_lock);
1845 t->control |= T_REMDEVALL;
1846 mutex_unlock(&pktgen_thread_lock);
1847 schedule_timeout_interruptible(msecs_to_jiffies(125)); /* Propagate thread->control */
1849 sprintf(pg_result, "OK: rem_device_all");
1853 if (!strcmp(name, "max_before_softirq")) {
1854 sprintf(pg_result, "OK: Note! max_before_softirq is obsoleted -- Do not use");
1864 static int pktgen_thread_open(struct inode *inode, struct file *file)
1866 return single_open(file, pktgen_thread_show, PDE_DATA(inode));
1869 static const struct file_operations pktgen_thread_fops = {
1870 .owner = THIS_MODULE,
1871 .open = pktgen_thread_open,
1873 .llseek = seq_lseek,
1874 .write = pktgen_thread_write,
1875 .release = single_release,
1878 /* Think find or remove for NN */
1879 static struct pktgen_dev *__pktgen_NN_threads(const struct pktgen_net *pn,
1880 const char *ifname, int remove)
1882 struct pktgen_thread *t;
1883 struct pktgen_dev *pkt_dev = NULL;
1884 bool exact = (remove == FIND);
1886 list_for_each_entry(t, &pn->pktgen_threads, th_list) {
1887 pkt_dev = pktgen_find_dev(t, ifname, exact);
1890 pkt_dev->removal_mark = 1;
1891 t->control |= T_REMDEV;
1900 * mark a device for removal
1902 static void pktgen_mark_device(const struct pktgen_net *pn, const char *ifname)
1904 struct pktgen_dev *pkt_dev = NULL;
1905 const int max_tries = 10, msec_per_try = 125;
1908 mutex_lock(&pktgen_thread_lock);
1909 pr_debug("%s: marking %s for removal\n", __func__, ifname);
1913 pkt_dev = __pktgen_NN_threads(pn, ifname, REMOVE);
1914 if (pkt_dev == NULL)
1915 break; /* success */
1917 mutex_unlock(&pktgen_thread_lock);
1918 pr_debug("%s: waiting for %s to disappear....\n",
1920 schedule_timeout_interruptible(msecs_to_jiffies(msec_per_try));
1921 mutex_lock(&pktgen_thread_lock);
1923 if (++i >= max_tries) {
1924 pr_err("%s: timed out after waiting %d msec for device %s to be removed\n",
1925 __func__, msec_per_try * i, ifname);
1931 mutex_unlock(&pktgen_thread_lock);
1934 static void pktgen_change_name(const struct pktgen_net *pn, struct net_device *dev)
1936 struct pktgen_thread *t;
1938 list_for_each_entry(t, &pn->pktgen_threads, th_list) {
1939 struct pktgen_dev *pkt_dev;
1942 list_for_each_entry_rcu(pkt_dev, &t->if_list, list) {
1943 if (pkt_dev->odev != dev)
1946 proc_remove(pkt_dev->entry);
1948 pkt_dev->entry = proc_create_data(dev->name, 0600,
1952 if (!pkt_dev->entry)
1953 pr_err("can't move proc entry for '%s'\n",
1961 static int pktgen_device_event(struct notifier_block *unused,
1962 unsigned long event, void *ptr)
1964 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1965 struct pktgen_net *pn = net_generic(dev_net(dev), pg_net_id);
1967 if (pn->pktgen_exiting)
1970 /* It is OK that we do not hold the group lock right now,
1971 * as we run under the RTNL lock.
1975 case NETDEV_CHANGENAME:
1976 pktgen_change_name(pn, dev);
1979 case NETDEV_UNREGISTER:
1980 pktgen_mark_device(pn, dev->name);
1987 static struct net_device *pktgen_dev_get_by_name(const struct pktgen_net *pn,
1988 struct pktgen_dev *pkt_dev,
1994 for (i = 0; ifname[i] != '@'; i++) {
2002 return dev_get_by_name(pn->net, b);
2006 /* Associate pktgen_dev with a device. */
2008 static int pktgen_setup_dev(const struct pktgen_net *pn,
2009 struct pktgen_dev *pkt_dev, const char *ifname)
2011 struct net_device *odev;
2014 /* Clean old setups */
2015 if (pkt_dev->odev) {
2016 dev_put(pkt_dev->odev);
2017 pkt_dev->odev = NULL;
2020 odev = pktgen_dev_get_by_name(pn, pkt_dev, ifname);
2022 pr_err("no such netdevice: \"%s\"\n", ifname);
2026 if (odev->type != ARPHRD_ETHER) {
2027 pr_err("not an ethernet device: \"%s\"\n", ifname);
2029 } else if (!netif_running(odev)) {
2030 pr_err("device is down: \"%s\"\n", ifname);
2033 pkt_dev->odev = odev;
2041 /* Read pkt_dev from the interface and set up internal pktgen_dev
2042 * structure to have the right information to create/send packets
2044 static void pktgen_setup_inject(struct pktgen_dev *pkt_dev)
2048 if (!pkt_dev->odev) {
2049 pr_err("ERROR: pkt_dev->odev == NULL in setup_inject\n");
2050 sprintf(pkt_dev->result,
2051 "ERROR: pkt_dev->odev == NULL in setup_inject.\n");
2055 /* make sure that we don't pick a non-existing transmit queue */
2056 ntxq = pkt_dev->odev->real_num_tx_queues;
2058 if (ntxq <= pkt_dev->queue_map_min) {
2059 pr_warning("WARNING: Requested queue_map_min (zero-based) (%d) exceeds valid range [0 - %d] for (%d) queues on %s, resetting\n",
2060 pkt_dev->queue_map_min, (ntxq ?: 1) - 1, ntxq,
2062 pkt_dev->queue_map_min = (ntxq ?: 1) - 1;
2064 if (pkt_dev->queue_map_max >= ntxq) {
2065 pr_warning("WARNING: Requested queue_map_max (zero-based) (%d) exceeds valid range [0 - %d] for (%d) queues on %s, resetting\n",
2066 pkt_dev->queue_map_max, (ntxq ?: 1) - 1, ntxq,
2068 pkt_dev->queue_map_max = (ntxq ?: 1) - 1;
2071 /* Default to the interface's mac if not explicitly set. */
2073 if (is_zero_ether_addr(pkt_dev->src_mac))
2074 ether_addr_copy(&(pkt_dev->hh[6]), pkt_dev->odev->dev_addr);
2076 /* Set up Dest MAC */
2077 ether_addr_copy(&(pkt_dev->hh[0]), pkt_dev->dst_mac);
2079 if (pkt_dev->flags & F_IPV6) {
2080 int i, set = 0, err = 1;
2081 struct inet6_dev *idev;
2083 if (pkt_dev->min_pkt_size == 0) {
2084 pkt_dev->min_pkt_size = 14 + sizeof(struct ipv6hdr)
2085 + sizeof(struct udphdr)
2086 + sizeof(struct pktgen_hdr)
2087 + pkt_dev->pkt_overhead;
2090 for (i = 0; i < IN6_ADDR_HSIZE; i++)
2091 if (pkt_dev->cur_in6_saddr.s6_addr[i]) {
2099 * Use linklevel address if unconfigured.
2101 * use ipv6_get_lladdr if/when it's get exported
2105 idev = __in6_dev_get(pkt_dev->odev);
2107 struct inet6_ifaddr *ifp;
2109 read_lock_bh(&idev->lock);
2110 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2111 if ((ifp->scope & IFA_LINK) &&
2112 !(ifp->flags & IFA_F_TENTATIVE)) {
2113 pkt_dev->cur_in6_saddr = ifp->addr;
2118 read_unlock_bh(&idev->lock);
2122 pr_err("ERROR: IPv6 link address not available\n");
2125 if (pkt_dev->min_pkt_size == 0) {
2126 pkt_dev->min_pkt_size = 14 + sizeof(struct iphdr)
2127 + sizeof(struct udphdr)
2128 + sizeof(struct pktgen_hdr)
2129 + pkt_dev->pkt_overhead;
2132 pkt_dev->saddr_min = 0;
2133 pkt_dev->saddr_max = 0;
2134 if (strlen(pkt_dev->src_min) == 0) {
2136 struct in_device *in_dev;
2139 in_dev = __in_dev_get_rcu(pkt_dev->odev);
2141 if (in_dev->ifa_list) {
2142 pkt_dev->saddr_min =
2143 in_dev->ifa_list->ifa_address;
2144 pkt_dev->saddr_max = pkt_dev->saddr_min;
2149 pkt_dev->saddr_min = in_aton(pkt_dev->src_min);
2150 pkt_dev->saddr_max = in_aton(pkt_dev->src_max);
2153 pkt_dev->daddr_min = in_aton(pkt_dev->dst_min);
2154 pkt_dev->daddr_max = in_aton(pkt_dev->dst_max);
2156 /* Initialize current values. */
2157 pkt_dev->cur_pkt_size = pkt_dev->min_pkt_size;
2158 if (pkt_dev->min_pkt_size > pkt_dev->max_pkt_size)
2159 pkt_dev->max_pkt_size = pkt_dev->min_pkt_size;
2161 pkt_dev->cur_dst_mac_offset = 0;
2162 pkt_dev->cur_src_mac_offset = 0;
2163 pkt_dev->cur_saddr = pkt_dev->saddr_min;
2164 pkt_dev->cur_daddr = pkt_dev->daddr_min;
2165 pkt_dev->cur_udp_dst = pkt_dev->udp_dst_min;
2166 pkt_dev->cur_udp_src = pkt_dev->udp_src_min;
2167 pkt_dev->nflows = 0;
2171 static void spin(struct pktgen_dev *pkt_dev, ktime_t spin_until)
2173 ktime_t start_time, end_time;
2175 struct hrtimer_sleeper t;
2177 hrtimer_init_on_stack(&t.timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS);
2178 hrtimer_set_expires(&t.timer, spin_until);
2180 remaining = ktime_to_ns(hrtimer_expires_remaining(&t.timer));
2181 if (remaining <= 0) {
2182 pkt_dev->next_tx = ktime_add_ns(spin_until, pkt_dev->delay);
2186 start_time = ktime_get();
2187 if (remaining < 100000) {
2188 /* for small delays (<100us), just loop until limit is reached */
2190 end_time = ktime_get();
2191 } while (ktime_compare(end_time, spin_until) < 0);
2193 /* see do_nanosleep */
2194 hrtimer_init_sleeper(&t, current);
2196 set_current_state(TASK_INTERRUPTIBLE);
2197 hrtimer_start_expires(&t.timer, HRTIMER_MODE_ABS);
2198 if (!hrtimer_active(&t.timer))
2204 hrtimer_cancel(&t.timer);
2205 } while (t.task && pkt_dev->running && !signal_pending(current));
2206 __set_current_state(TASK_RUNNING);
2207 end_time = ktime_get();
2210 pkt_dev->idle_acc += ktime_to_ns(ktime_sub(end_time, start_time));
2211 pkt_dev->next_tx = ktime_add_ns(spin_until, pkt_dev->delay);
2214 static inline void set_pkt_overhead(struct pktgen_dev *pkt_dev)
2216 pkt_dev->pkt_overhead = 0;
2217 pkt_dev->pkt_overhead += pkt_dev->nr_labels*sizeof(u32);
2218 pkt_dev->pkt_overhead += VLAN_TAG_SIZE(pkt_dev);
2219 pkt_dev->pkt_overhead += SVLAN_TAG_SIZE(pkt_dev);
2222 static inline int f_seen(const struct pktgen_dev *pkt_dev, int flow)
2224 return !!(pkt_dev->flows[flow].flags & F_INIT);
2227 static inline int f_pick(struct pktgen_dev *pkt_dev)
2229 int flow = pkt_dev->curfl;
2231 if (pkt_dev->flags & F_FLOW_SEQ) {
2232 if (pkt_dev->flows[flow].count >= pkt_dev->lflow) {
2234 pkt_dev->flows[flow].count = 0;
2235 pkt_dev->flows[flow].flags = 0;
2236 pkt_dev->curfl += 1;
2237 if (pkt_dev->curfl >= pkt_dev->cflows)
2238 pkt_dev->curfl = 0; /*reset */
2241 flow = prandom_u32() % pkt_dev->cflows;
2242 pkt_dev->curfl = flow;
2244 if (pkt_dev->flows[flow].count > pkt_dev->lflow) {
2245 pkt_dev->flows[flow].count = 0;
2246 pkt_dev->flows[flow].flags = 0;
2250 return pkt_dev->curfl;
2255 /* If there was already an IPSEC SA, we keep it as is, else
2256 * we go look for it ...
2258 #define DUMMY_MARK 0
2259 static void get_ipsec_sa(struct pktgen_dev *pkt_dev, int flow)
2261 struct xfrm_state *x = pkt_dev->flows[flow].x;
2262 struct pktgen_net *pn = net_generic(dev_net(pkt_dev->odev), pg_net_id);
2266 /* We need as quick as possible to find the right SA
2267 * Searching with minimum criteria to archieve this.
2269 x = xfrm_state_lookup_byspi(pn->net, htonl(pkt_dev->spi), AF_INET);
2271 /* slow path: we dont already have xfrm_state */
2272 x = xfrm_stateonly_find(pn->net, DUMMY_MARK,
2273 (xfrm_address_t *)&pkt_dev->cur_daddr,
2274 (xfrm_address_t *)&pkt_dev->cur_saddr,
2277 pkt_dev->ipsproto, 0);
2280 pkt_dev->flows[flow].x = x;
2281 set_pkt_overhead(pkt_dev);
2282 pkt_dev->pkt_overhead += x->props.header_len;
2288 static void set_cur_queue_map(struct pktgen_dev *pkt_dev)
2291 if (pkt_dev->flags & F_QUEUE_MAP_CPU)
2292 pkt_dev->cur_queue_map = smp_processor_id();
2294 else if (pkt_dev->queue_map_min <= pkt_dev->queue_map_max) {
2296 if (pkt_dev->flags & F_QUEUE_MAP_RND) {
2298 (pkt_dev->queue_map_max -
2299 pkt_dev->queue_map_min + 1)
2300 + pkt_dev->queue_map_min;
2302 t = pkt_dev->cur_queue_map + 1;
2303 if (t > pkt_dev->queue_map_max)
2304 t = pkt_dev->queue_map_min;
2306 pkt_dev->cur_queue_map = t;
2308 pkt_dev->cur_queue_map = pkt_dev->cur_queue_map % pkt_dev->odev->real_num_tx_queues;
2311 /* Increment/randomize headers according to flags and current values
2312 * for IP src/dest, UDP src/dst port, MAC-Addr src/dst
2314 static void mod_cur_headers(struct pktgen_dev *pkt_dev)
2320 if (pkt_dev->cflows)
2321 flow = f_pick(pkt_dev);
2323 /* Deal with source MAC */
2324 if (pkt_dev->src_mac_count > 1) {
2328 if (pkt_dev->flags & F_MACSRC_RND)
2329 mc = prandom_u32() % pkt_dev->src_mac_count;
2331 mc = pkt_dev->cur_src_mac_offset++;
2332 if (pkt_dev->cur_src_mac_offset >=
2333 pkt_dev->src_mac_count)
2334 pkt_dev->cur_src_mac_offset = 0;
2337 tmp = pkt_dev->src_mac[5] + (mc & 0xFF);
2338 pkt_dev->hh[11] = tmp;
2339 tmp = (pkt_dev->src_mac[4] + ((mc >> 8) & 0xFF) + (tmp >> 8));
2340 pkt_dev->hh[10] = tmp;
2341 tmp = (pkt_dev->src_mac[3] + ((mc >> 16) & 0xFF) + (tmp >> 8));
2342 pkt_dev->hh[9] = tmp;
2343 tmp = (pkt_dev->src_mac[2] + ((mc >> 24) & 0xFF) + (tmp >> 8));
2344 pkt_dev->hh[8] = tmp;
2345 tmp = (pkt_dev->src_mac[1] + (tmp >> 8));
2346 pkt_dev->hh[7] = tmp;
2349 /* Deal with Destination MAC */
2350 if (pkt_dev->dst_mac_count > 1) {
2354 if (pkt_dev->flags & F_MACDST_RND)
2355 mc = prandom_u32() % pkt_dev->dst_mac_count;
2358 mc = pkt_dev->cur_dst_mac_offset++;
2359 if (pkt_dev->cur_dst_mac_offset >=
2360 pkt_dev->dst_mac_count) {
2361 pkt_dev->cur_dst_mac_offset = 0;
2365 tmp = pkt_dev->dst_mac[5] + (mc & 0xFF);
2366 pkt_dev->hh[5] = tmp;
2367 tmp = (pkt_dev->dst_mac[4] + ((mc >> 8) & 0xFF) + (tmp >> 8));
2368 pkt_dev->hh[4] = tmp;
2369 tmp = (pkt_dev->dst_mac[3] + ((mc >> 16) & 0xFF) + (tmp >> 8));
2370 pkt_dev->hh[3] = tmp;
2371 tmp = (pkt_dev->dst_mac[2] + ((mc >> 24) & 0xFF) + (tmp >> 8));
2372 pkt_dev->hh[2] = tmp;
2373 tmp = (pkt_dev->dst_mac[1] + (tmp >> 8));
2374 pkt_dev->hh[1] = tmp;
2377 if (pkt_dev->flags & F_MPLS_RND) {
2379 for (i = 0; i < pkt_dev->nr_labels; i++)
2380 if (pkt_dev->labels[i] & MPLS_STACK_BOTTOM)
2381 pkt_dev->labels[i] = MPLS_STACK_BOTTOM |
2382 ((__force __be32)prandom_u32() &
2386 if ((pkt_dev->flags & F_VID_RND) && (pkt_dev->vlan_id != 0xffff)) {
2387 pkt_dev->vlan_id = prandom_u32() & (4096 - 1);
2390 if ((pkt_dev->flags & F_SVID_RND) && (pkt_dev->svlan_id != 0xffff)) {
2391 pkt_dev->svlan_id = prandom_u32() & (4096 - 1);
2394 if (pkt_dev->udp_src_min < pkt_dev->udp_src_max) {
2395 if (pkt_dev->flags & F_UDPSRC_RND)
2396 pkt_dev->cur_udp_src = prandom_u32() %
2397 (pkt_dev->udp_src_max - pkt_dev->udp_src_min)
2398 + pkt_dev->udp_src_min;
2401 pkt_dev->cur_udp_src++;
2402 if (pkt_dev->cur_udp_src >= pkt_dev->udp_src_max)
2403 pkt_dev->cur_udp_src = pkt_dev->udp_src_min;
2407 if (pkt_dev->udp_dst_min < pkt_dev->udp_dst_max) {
2408 if (pkt_dev->flags & F_UDPDST_RND) {
2409 pkt_dev->cur_udp_dst = prandom_u32() %
2410 (pkt_dev->udp_dst_max - pkt_dev->udp_dst_min)
2411 + pkt_dev->udp_dst_min;
2413 pkt_dev->cur_udp_dst++;
2414 if (pkt_dev->cur_udp_dst >= pkt_dev->udp_dst_max)
2415 pkt_dev->cur_udp_dst = pkt_dev->udp_dst_min;
2419 if (!(pkt_dev->flags & F_IPV6)) {
2421 imn = ntohl(pkt_dev->saddr_min);
2422 imx = ntohl(pkt_dev->saddr_max);
2425 if (pkt_dev->flags & F_IPSRC_RND)
2426 t = prandom_u32() % (imx - imn) + imn;
2428 t = ntohl(pkt_dev->cur_saddr);
2434 pkt_dev->cur_saddr = htonl(t);
2437 if (pkt_dev->cflows && f_seen(pkt_dev, flow)) {
2438 pkt_dev->cur_daddr = pkt_dev->flows[flow].cur_daddr;
2440 imn = ntohl(pkt_dev->daddr_min);
2441 imx = ntohl(pkt_dev->daddr_max);
2445 if (pkt_dev->flags & F_IPDST_RND) {
2451 } while (ipv4_is_loopback(s) ||
2452 ipv4_is_multicast(s) ||
2453 ipv4_is_lbcast(s) ||
2454 ipv4_is_zeronet(s) ||
2455 ipv4_is_local_multicast(s));
2456 pkt_dev->cur_daddr = s;
2458 t = ntohl(pkt_dev->cur_daddr);
2463 pkt_dev->cur_daddr = htonl(t);
2466 if (pkt_dev->cflows) {
2467 pkt_dev->flows[flow].flags |= F_INIT;
2468 pkt_dev->flows[flow].cur_daddr =
2471 if (pkt_dev->flags & F_IPSEC_ON)
2472 get_ipsec_sa(pkt_dev, flow);
2477 } else { /* IPV6 * */
2479 if (!ipv6_addr_any(&pkt_dev->min_in6_daddr)) {
2482 /* Only random destinations yet */
2484 for (i = 0; i < 4; i++) {
2485 pkt_dev->cur_in6_daddr.s6_addr32[i] =
2486 (((__force __be32)prandom_u32() |
2487 pkt_dev->min_in6_daddr.s6_addr32[i]) &
2488 pkt_dev->max_in6_daddr.s6_addr32[i]);
2493 if (pkt_dev->min_pkt_size < pkt_dev->max_pkt_size) {
2495 if (pkt_dev->flags & F_TXSIZE_RND) {
2497 (pkt_dev->max_pkt_size - pkt_dev->min_pkt_size)
2498 + pkt_dev->min_pkt_size;
2500 t = pkt_dev->cur_pkt_size + 1;
2501 if (t > pkt_dev->max_pkt_size)
2502 t = pkt_dev->min_pkt_size;
2504 pkt_dev->cur_pkt_size = t;
2507 set_cur_queue_map(pkt_dev);
2509 pkt_dev->flows[flow].count++;
2514 static u32 pktgen_dst_metrics[RTAX_MAX + 1] = {
2516 [RTAX_HOPLIMIT] = 0x5, /* Set a static hoplimit */
2519 static int pktgen_output_ipsec(struct sk_buff *skb, struct pktgen_dev *pkt_dev)
2521 struct xfrm_state *x = pkt_dev->flows[pkt_dev->curfl].x;
2523 struct net *net = dev_net(pkt_dev->odev);
2527 /* XXX: we dont support tunnel mode for now until
2528 * we resolve the dst issue */
2529 if ((x->props.mode != XFRM_MODE_TRANSPORT) && (pkt_dev->spi == 0))
2532 /* But when user specify an valid SPI, transformation
2533 * supports both transport/tunnel mode + ESP/AH type.
2535 if ((x->props.mode == XFRM_MODE_TUNNEL) && (pkt_dev->spi != 0))
2536 skb->_skb_refdst = (unsigned long)&pkt_dev->dst | SKB_DST_NOREF;
2539 err = x->outer_mode->output(x, skb);
2540 rcu_read_unlock_bh();
2542 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATEMODEERROR);
2545 err = x->type->output(x, skb);
2547 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATEPROTOERROR);
2550 spin_lock_bh(&x->lock);
2551 x->curlft.bytes += skb->len;
2552 x->curlft.packets++;
2553 spin_unlock_bh(&x->lock);
2558 static void free_SAs(struct pktgen_dev *pkt_dev)
2560 if (pkt_dev->cflows) {
2561 /* let go of the SAs if we have them */
2563 for (i = 0; i < pkt_dev->cflows; i++) {
2564 struct xfrm_state *x = pkt_dev->flows[i].x;
2567 pkt_dev->flows[i].x = NULL;
2573 static int process_ipsec(struct pktgen_dev *pkt_dev,
2574 struct sk_buff *skb, __be16 protocol)
2576 if (pkt_dev->flags & F_IPSEC_ON) {
2577 struct xfrm_state *x = pkt_dev->flows[pkt_dev->curfl].x;
2584 nhead = x->props.header_len - skb_headroom(skb);
2586 ret = pskb_expand_head(skb, nhead, 0, GFP_ATOMIC);
2588 pr_err("Error expanding ipsec packet %d\n",
2594 /* ipsec is not expecting ll header */
2595 skb_pull(skb, ETH_HLEN);
2596 ret = pktgen_output_ipsec(skb, pkt_dev);
2598 pr_err("Error creating ipsec packet %d\n", ret);
2602 eth = (__u8 *) skb_push(skb, ETH_HLEN);
2603 memcpy(eth, pkt_dev->hh, 12);
2604 *(u16 *) ð[12] = protocol;
2606 /* Update IPv4 header len as well as checksum value */
2608 iph->tot_len = htons(skb->len - ETH_HLEN);
2619 static void mpls_push(__be32 *mpls, struct pktgen_dev *pkt_dev)
2622 for (i = 0; i < pkt_dev->nr_labels; i++)
2623 *mpls++ = pkt_dev->labels[i] & ~MPLS_STACK_BOTTOM;
2626 *mpls |= MPLS_STACK_BOTTOM;
2629 static inline __be16 build_tci(unsigned int id, unsigned int cfi,
2632 return htons(id | (cfi << 12) | (prio << 13));
2635 static void pktgen_finalize_skb(struct pktgen_dev *pkt_dev, struct sk_buff *skb,
2638 struct timeval timestamp;
2639 struct pktgen_hdr *pgh;
2641 pgh = (struct pktgen_hdr *)skb_put(skb, sizeof(*pgh));
2642 datalen -= sizeof(*pgh);
2644 if (pkt_dev->nfrags <= 0) {
2645 memset(skb_put(skb, datalen), 0, datalen);
2647 int frags = pkt_dev->nfrags;
2652 if (frags > MAX_SKB_FRAGS)
2653 frags = MAX_SKB_FRAGS;
2654 len = datalen - frags * PAGE_SIZE;
2656 memset(skb_put(skb, len), 0, len);
2657 datalen = frags * PAGE_SIZE;
2661 frag_len = (datalen/frags) < PAGE_SIZE ?
2662 (datalen/frags) : PAGE_SIZE;
2663 while (datalen > 0) {
2664 if (unlikely(!pkt_dev->page)) {
2665 int node = numa_node_id();
2667 if (pkt_dev->node >= 0 && (pkt_dev->flags & F_NODE))
2668 node = pkt_dev->node;
2669 pkt_dev->page = alloc_pages_node(node, GFP_KERNEL | __GFP_ZERO, 0);
2673 get_page(pkt_dev->page);
2674 skb_frag_set_page(skb, i, pkt_dev->page);
2675 skb_shinfo(skb)->frags[i].page_offset = 0;
2676 /*last fragment, fill rest of data*/
2677 if (i == (frags - 1))
2678 skb_frag_size_set(&skb_shinfo(skb)->frags[i],
2679 (datalen < PAGE_SIZE ? datalen : PAGE_SIZE));
2681 skb_frag_size_set(&skb_shinfo(skb)->frags[i], frag_len);
2682 datalen -= skb_frag_size(&skb_shinfo(skb)->frags[i]);
2683 skb->len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2684 skb->data_len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2686 skb_shinfo(skb)->nr_frags = i;
2690 /* Stamp the time, and sequence number,
2691 * convert them to network byte order
2693 pgh->pgh_magic = htonl(PKTGEN_MAGIC);
2694 pgh->seq_num = htonl(pkt_dev->seq_num);
2696 if (pkt_dev->flags & F_NO_TIMESTAMP) {
2700 do_gettimeofday(×tamp);
2701 pgh->tv_sec = htonl(timestamp.tv_sec);
2702 pgh->tv_usec = htonl(timestamp.tv_usec);
2706 static struct sk_buff *pktgen_alloc_skb(struct net_device *dev,
2707 struct pktgen_dev *pkt_dev,
2708 unsigned int extralen)
2710 struct sk_buff *skb = NULL;
2711 unsigned int size = pkt_dev->cur_pkt_size + 64 + extralen +
2712 pkt_dev->pkt_overhead;
2714 if (pkt_dev->flags & F_NODE) {
2715 int node = pkt_dev->node >= 0 ? pkt_dev->node : numa_node_id();
2717 skb = __alloc_skb(NET_SKB_PAD + size, GFP_NOWAIT, 0, node);
2719 skb_reserve(skb, NET_SKB_PAD);
2723 skb = __netdev_alloc_skb(dev, size, GFP_NOWAIT);
2729 static struct sk_buff *fill_packet_ipv4(struct net_device *odev,
2730 struct pktgen_dev *pkt_dev)
2732 struct sk_buff *skb = NULL;
2734 struct udphdr *udph;
2737 __be16 protocol = htons(ETH_P_IP);
2739 __be16 *vlan_tci = NULL; /* Encapsulates priority and VLAN ID */
2740 __be16 *vlan_encapsulated_proto = NULL; /* packet type ID field (or len) for VLAN tag */
2741 __be16 *svlan_tci = NULL; /* Encapsulates priority and SVLAN ID */
2742 __be16 *svlan_encapsulated_proto = NULL; /* packet type ID field (or len) for SVLAN tag */
2745 if (pkt_dev->nr_labels)
2746 protocol = htons(ETH_P_MPLS_UC);
2748 if (pkt_dev->vlan_id != 0xffff)
2749 protocol = htons(ETH_P_8021Q);
2751 /* Update any of the values, used when we're incrementing various
2754 mod_cur_headers(pkt_dev);
2755 queue_map = pkt_dev->cur_queue_map;
2757 datalen = (odev->hard_header_len + 16) & ~0xf;
2759 skb = pktgen_alloc_skb(odev, pkt_dev, datalen);
2761 sprintf(pkt_dev->result, "No memory");
2765 prefetchw(skb->data);
2766 skb_reserve(skb, datalen);
2768 /* Reserve for ethernet and IP header */
2769 eth = (__u8 *) skb_push(skb, 14);
2770 mpls = (__be32 *)skb_put(skb, pkt_dev->nr_labels*sizeof(__u32));
2771 if (pkt_dev->nr_labels)
2772 mpls_push(mpls, pkt_dev);
2774 if (pkt_dev->vlan_id != 0xffff) {
2775 if (pkt_dev->svlan_id != 0xffff) {
2776 svlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2777 *svlan_tci = build_tci(pkt_dev->svlan_id,
2780 svlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2781 *svlan_encapsulated_proto = htons(ETH_P_8021Q);
2783 vlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2784 *vlan_tci = build_tci(pkt_dev->vlan_id,
2787 vlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2788 *vlan_encapsulated_proto = htons(ETH_P_IP);
2791 skb_set_mac_header(skb, 0);
2792 skb_set_network_header(skb, skb->len);
2793 iph = (struct iphdr *) skb_put(skb, sizeof(struct iphdr));
2795 skb_set_transport_header(skb, skb->len);
2796 udph = (struct udphdr *) skb_put(skb, sizeof(struct udphdr));
2797 skb_set_queue_mapping(skb, queue_map);
2798 skb->priority = pkt_dev->skb_priority;
2800 memcpy(eth, pkt_dev->hh, 12);
2801 *(__be16 *) & eth[12] = protocol;
2803 /* Eth + IPh + UDPh + mpls */
2804 datalen = pkt_dev->cur_pkt_size - 14 - 20 - 8 -
2805 pkt_dev->pkt_overhead;
2806 if (datalen < 0 || datalen < sizeof(struct pktgen_hdr))
2807 datalen = sizeof(struct pktgen_hdr);
2809 udph->source = htons(pkt_dev->cur_udp_src);
2810 udph->dest = htons(pkt_dev->cur_udp_dst);
2811 udph->len = htons(datalen + 8); /* DATA + udphdr */
2817 iph->tos = pkt_dev->tos;
2818 iph->protocol = IPPROTO_UDP; /* UDP */
2819 iph->saddr = pkt_dev->cur_saddr;
2820 iph->daddr = pkt_dev->cur_daddr;
2821 iph->id = htons(pkt_dev->ip_id);
2824 iplen = 20 + 8 + datalen;
2825 iph->tot_len = htons(iplen);
2827 skb->protocol = protocol;
2829 skb->pkt_type = PACKET_HOST;
2831 if (!(pkt_dev->flags & F_UDPCSUM)) {
2832 skb->ip_summed = CHECKSUM_NONE;
2833 } else if (odev->features & NETIF_F_V4_CSUM) {
2834 skb->ip_summed = CHECKSUM_PARTIAL;
2836 udp4_hwcsum(skb, udph->source, udph->dest);
2838 __wsum csum = udp_csum(skb);
2840 /* add protocol-dependent pseudo-header */
2841 udph->check = csum_tcpudp_magic(udph->source, udph->dest,
2842 datalen + 8, IPPROTO_UDP, csum);
2844 if (udph->check == 0)
2845 udph->check = CSUM_MANGLED_0;
2848 pktgen_finalize_skb(pkt_dev, skb, datalen);
2851 if (!process_ipsec(pkt_dev, skb, protocol))
2858 static struct sk_buff *fill_packet_ipv6(struct net_device *odev,
2859 struct pktgen_dev *pkt_dev)
2861 struct sk_buff *skb = NULL;
2863 struct udphdr *udph;
2864 int datalen, udplen;
2865 struct ipv6hdr *iph;
2866 __be16 protocol = htons(ETH_P_IPV6);
2868 __be16 *vlan_tci = NULL; /* Encapsulates priority and VLAN ID */
2869 __be16 *vlan_encapsulated_proto = NULL; /* packet type ID field (or len) for VLAN tag */
2870 __be16 *svlan_tci = NULL; /* Encapsulates priority and SVLAN ID */
2871 __be16 *svlan_encapsulated_proto = NULL; /* packet type ID field (or len) for SVLAN tag */
2874 if (pkt_dev->nr_labels)
2875 protocol = htons(ETH_P_MPLS_UC);
2877 if (pkt_dev->vlan_id != 0xffff)
2878 protocol = htons(ETH_P_8021Q);
2880 /* Update any of the values, used when we're incrementing various
2883 mod_cur_headers(pkt_dev);
2884 queue_map = pkt_dev->cur_queue_map;
2886 skb = pktgen_alloc_skb(odev, pkt_dev, 16);
2888 sprintf(pkt_dev->result, "No memory");
2892 prefetchw(skb->data);
2893 skb_reserve(skb, 16);
2895 /* Reserve for ethernet and IP header */
2896 eth = (__u8 *) skb_push(skb, 14);
2897 mpls = (__be32 *)skb_put(skb, pkt_dev->nr_labels*sizeof(__u32));
2898 if (pkt_dev->nr_labels)
2899 mpls_push(mpls, pkt_dev);
2901 if (pkt_dev->vlan_id != 0xffff) {
2902 if (pkt_dev->svlan_id != 0xffff) {
2903 svlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2904 *svlan_tci = build_tci(pkt_dev->svlan_id,
2907 svlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2908 *svlan_encapsulated_proto = htons(ETH_P_8021Q);
2910 vlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2911 *vlan_tci = build_tci(pkt_dev->vlan_id,
2914 vlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2915 *vlan_encapsulated_proto = htons(ETH_P_IPV6);
2918 skb_set_mac_header(skb, 0);
2919 skb_set_network_header(skb, skb->len);
2920 iph = (struct ipv6hdr *) skb_put(skb, sizeof(struct ipv6hdr));
2922 skb_set_transport_header(skb, skb->len);
2923 udph = (struct udphdr *) skb_put(skb, sizeof(struct udphdr));
2924 skb_set_queue_mapping(skb, queue_map);
2925 skb->priority = pkt_dev->skb_priority;
2927 memcpy(eth, pkt_dev->hh, 12);
2928 *(__be16 *) ð[12] = protocol;
2930 /* Eth + IPh + UDPh + mpls */
2931 datalen = pkt_dev->cur_pkt_size - 14 -
2932 sizeof(struct ipv6hdr) - sizeof(struct udphdr) -
2933 pkt_dev->pkt_overhead;
2935 if (datalen < 0 || datalen < sizeof(struct pktgen_hdr)) {
2936 datalen = sizeof(struct pktgen_hdr);
2937 net_info_ratelimited("increased datalen to %d\n", datalen);
2940 udplen = datalen + sizeof(struct udphdr);
2941 udph->source = htons(pkt_dev->cur_udp_src);
2942 udph->dest = htons(pkt_dev->cur_udp_dst);
2943 udph->len = htons(udplen);
2946 *(__be32 *) iph = htonl(0x60000000); /* Version + flow */
2948 if (pkt_dev->traffic_class) {
2949 /* Version + traffic class + flow (0) */
2950 *(__be32 *)iph |= htonl(0x60000000 | (pkt_dev->traffic_class << 20));
2953 iph->hop_limit = 32;
2955 iph->payload_len = htons(udplen);
2956 iph->nexthdr = IPPROTO_UDP;
2958 iph->daddr = pkt_dev->cur_in6_daddr;
2959 iph->saddr = pkt_dev->cur_in6_saddr;
2961 skb->protocol = protocol;
2963 skb->pkt_type = PACKET_HOST;
2965 if (!(pkt_dev->flags & F_UDPCSUM)) {
2966 skb->ip_summed = CHECKSUM_NONE;
2967 } else if (odev->features & NETIF_F_V6_CSUM) {
2968 skb->ip_summed = CHECKSUM_PARTIAL;
2969 skb->csum_start = skb_transport_header(skb) - skb->head;
2970 skb->csum_offset = offsetof(struct udphdr, check);
2971 udph->check = ~csum_ipv6_magic(&iph->saddr, &iph->daddr, udplen, IPPROTO_UDP, 0);
2973 __wsum csum = udp_csum(skb);
2975 /* add protocol-dependent pseudo-header */
2976 udph->check = csum_ipv6_magic(&iph->saddr, &iph->daddr, udplen, IPPROTO_UDP, csum);
2978 if (udph->check == 0)
2979 udph->check = CSUM_MANGLED_0;
2982 pktgen_finalize_skb(pkt_dev, skb, datalen);
2987 static struct sk_buff *fill_packet(struct net_device *odev,
2988 struct pktgen_dev *pkt_dev)
2990 if (pkt_dev->flags & F_IPV6)
2991 return fill_packet_ipv6(odev, pkt_dev);
2993 return fill_packet_ipv4(odev, pkt_dev);
2996 static void pktgen_clear_counters(struct pktgen_dev *pkt_dev)
2998 pkt_dev->seq_num = 1;
2999 pkt_dev->idle_acc = 0;
3001 pkt_dev->tx_bytes = 0;
3002 pkt_dev->errors = 0;
3005 /* Set up structure for sending pkts, clear counters */
3007 static void pktgen_run(struct pktgen_thread *t)
3009 struct pktgen_dev *pkt_dev;
3015 list_for_each_entry_rcu(pkt_dev, &t->if_list, list) {
3018 * setup odev and create initial packet.
3020 pktgen_setup_inject(pkt_dev);
3022 if (pkt_dev->odev) {
3023 pktgen_clear_counters(pkt_dev);
3024 pkt_dev->skb = NULL;
3025 pkt_dev->started_at = pkt_dev->next_tx = ktime_get();
3027 set_pkt_overhead(pkt_dev);
3029 strcpy(pkt_dev->result, "Starting");
3030 pkt_dev->running = 1; /* Cranke yeself! */
3033 strcpy(pkt_dev->result, "Error starting");
3037 t->control &= ~(T_STOP);
3040 static void pktgen_stop_all_threads_ifs(struct pktgen_net *pn)
3042 struct pktgen_thread *t;
3046 mutex_lock(&pktgen_thread_lock);
3048 list_for_each_entry(t, &pn->pktgen_threads, th_list)
3049 t->control |= T_STOP;
3051 mutex_unlock(&pktgen_thread_lock);
3054 static int thread_is_running(const struct pktgen_thread *t)
3056 const struct pktgen_dev *pkt_dev;
3059 list_for_each_entry_rcu(pkt_dev, &t->if_list, list)
3060 if (pkt_dev->running) {
3068 static int pktgen_wait_thread_run(struct pktgen_thread *t)
3070 while (thread_is_running(t)) {
3072 msleep_interruptible(100);
3074 if (signal_pending(current))
3082 static int pktgen_wait_all_threads_run(struct pktgen_net *pn)
3084 struct pktgen_thread *t;
3087 mutex_lock(&pktgen_thread_lock);
3089 list_for_each_entry(t, &pn->pktgen_threads, th_list) {
3090 sig = pktgen_wait_thread_run(t);
3096 list_for_each_entry(t, &pn->pktgen_threads, th_list)
3097 t->control |= (T_STOP);
3099 mutex_unlock(&pktgen_thread_lock);
3103 static void pktgen_run_all_threads(struct pktgen_net *pn)
3105 struct pktgen_thread *t;
3109 mutex_lock(&pktgen_thread_lock);
3111 list_for_each_entry(t, &pn->pktgen_threads, th_list)
3112 t->control |= (T_RUN);
3114 mutex_unlock(&pktgen_thread_lock);
3116 /* Propagate thread->control */
3117 schedule_timeout_interruptible(msecs_to_jiffies(125));
3119 pktgen_wait_all_threads_run(pn);
3122 static void pktgen_reset_all_threads(struct pktgen_net *pn)
3124 struct pktgen_thread *t;
3128 mutex_lock(&pktgen_thread_lock);
3130 list_for_each_entry(t, &pn->pktgen_threads, th_list)
3131 t->control |= (T_REMDEVALL);
3133 mutex_unlock(&pktgen_thread_lock);
3135 /* Propagate thread->control */
3136 schedule_timeout_interruptible(msecs_to_jiffies(125));
3138 pktgen_wait_all_threads_run(pn);
3141 static void show_results(struct pktgen_dev *pkt_dev, int nr_frags)
3143 __u64 bps, mbps, pps;
3144 char *p = pkt_dev->result;
3145 ktime_t elapsed = ktime_sub(pkt_dev->stopped_at,
3146 pkt_dev->started_at);
3147 ktime_t idle = ns_to_ktime(pkt_dev->idle_acc);
3149 p += sprintf(p, "OK: %llu(c%llu+d%llu) usec, %llu (%dbyte,%dfrags)\n",
3150 (unsigned long long)ktime_to_us(elapsed),
3151 (unsigned long long)ktime_to_us(ktime_sub(elapsed, idle)),
3152 (unsigned long long)ktime_to_us(idle),
3153 (unsigned long long)pkt_dev->sofar,
3154 pkt_dev->cur_pkt_size, nr_frags);
3156 pps = div64_u64(pkt_dev->sofar * NSEC_PER_SEC,
3157 ktime_to_ns(elapsed));
3159 bps = pps * 8 * pkt_dev->cur_pkt_size;
3162 do_div(mbps, 1000000);
3163 p += sprintf(p, " %llupps %lluMb/sec (%llubps) errors: %llu",
3164 (unsigned long long)pps,
3165 (unsigned long long)mbps,
3166 (unsigned long long)bps,
3167 (unsigned long long)pkt_dev->errors);
3170 /* Set stopped-at timer, remove from running list, do counters & statistics */
3171 static int pktgen_stop_device(struct pktgen_dev *pkt_dev)
3173 int nr_frags = pkt_dev->skb ? skb_shinfo(pkt_dev->skb)->nr_frags : -1;
3175 if (!pkt_dev->running) {
3176 pr_warning("interface: %s is already stopped\n",
3181 pkt_dev->running = 0;
3182 kfree_skb(pkt_dev->skb);
3183 pkt_dev->skb = NULL;
3184 pkt_dev->stopped_at = ktime_get();
3186 show_results(pkt_dev, nr_frags);
3191 static struct pktgen_dev *next_to_run(struct pktgen_thread *t)
3193 struct pktgen_dev *pkt_dev, *best = NULL;
3196 list_for_each_entry_rcu(pkt_dev, &t->if_list, list) {
3197 if (!pkt_dev->running)
3201 else if (ktime_compare(pkt_dev->next_tx, best->next_tx) < 0)
3209 static void pktgen_stop(struct pktgen_thread *t)
3211 struct pktgen_dev *pkt_dev;
3217 list_for_each_entry_rcu(pkt_dev, &t->if_list, list) {
3218 pktgen_stop_device(pkt_dev);
3225 * one of our devices needs to be removed - find it
3228 static void pktgen_rem_one_if(struct pktgen_thread *t)
3230 struct list_head *q, *n;
3231 struct pktgen_dev *cur;
3235 list_for_each_safe(q, n, &t->if_list) {
3236 cur = list_entry(q, struct pktgen_dev, list);
3238 if (!cur->removal_mark)
3241 kfree_skb(cur->skb);
3244 pktgen_remove_device(t, cur);
3250 static void pktgen_rem_all_ifs(struct pktgen_thread *t)
3252 struct list_head *q, *n;
3253 struct pktgen_dev *cur;
3257 /* Remove all devices, free mem */
3259 list_for_each_safe(q, n, &t->if_list) {
3260 cur = list_entry(q, struct pktgen_dev, list);
3262 kfree_skb(cur->skb);
3265 pktgen_remove_device(t, cur);
3269 static void pktgen_rem_thread(struct pktgen_thread *t)
3271 /* Remove from the thread list */
3272 remove_proc_entry(t->tsk->comm, t->net->proc_dir);
3275 static void pktgen_resched(struct pktgen_dev *pkt_dev)
3277 ktime_t idle_start = ktime_get();
3279 pkt_dev->idle_acc += ktime_to_ns(ktime_sub(ktime_get(), idle_start));
3282 static void pktgen_wait_for_skb(struct pktgen_dev *pkt_dev)
3284 ktime_t idle_start = ktime_get();
3286 while (atomic_read(&(pkt_dev->skb->users)) != 1) {
3287 if (signal_pending(current))
3291 pktgen_resched(pkt_dev);
3295 pkt_dev->idle_acc += ktime_to_ns(ktime_sub(ktime_get(), idle_start));
3298 static void pktgen_xmit(struct pktgen_dev *pkt_dev)
3300 struct net_device *odev = pkt_dev->odev;
3301 struct netdev_queue *txq;
3304 /* If device is offline, then don't send */
3305 if (unlikely(!netif_running(odev) || !netif_carrier_ok(odev))) {
3306 pktgen_stop_device(pkt_dev);
3310 /* This is max DELAY, this has special meaning of
3313 if (unlikely(pkt_dev->delay == ULLONG_MAX)) {
3314 pkt_dev->next_tx = ktime_add_ns(ktime_get(), ULONG_MAX);
3318 /* If no skb or clone count exhausted then get new one */
3319 if (!pkt_dev->skb || (pkt_dev->last_ok &&
3320 ++pkt_dev->clone_count >= pkt_dev->clone_skb)) {
3321 /* build a new pkt */
3322 kfree_skb(pkt_dev->skb);
3324 pkt_dev->skb = fill_packet(odev, pkt_dev);
3325 if (pkt_dev->skb == NULL) {
3326 pr_err("ERROR: couldn't allocate skb in fill_packet\n");
3328 pkt_dev->clone_count--; /* back out increment, OOM */
3331 pkt_dev->last_pkt_size = pkt_dev->skb->len;
3332 pkt_dev->allocated_skbs++;
3333 pkt_dev->clone_count = 0; /* reset counter */
3336 if (pkt_dev->delay && pkt_dev->last_ok)
3337 spin(pkt_dev, pkt_dev->next_tx);
3339 txq = skb_get_tx_queue(odev, pkt_dev->skb);
3343 HARD_TX_LOCK(odev, txq, smp_processor_id());
3345 if (unlikely(netif_xmit_frozen_or_drv_stopped(txq))) {
3346 ret = NETDEV_TX_BUSY;
3347 pkt_dev->last_ok = 0;
3350 atomic_inc(&(pkt_dev->skb->users));
3351 ret = netdev_start_xmit(pkt_dev->skb, odev, txq, false);
3355 pkt_dev->last_ok = 1;
3358 pkt_dev->tx_bytes += pkt_dev->last_pkt_size;
3362 case NET_XMIT_POLICED:
3363 /* skb has been consumed */
3366 default: /* Drivers are not supposed to return other values! */
3367 net_info_ratelimited("%s xmit error: %d\n",
3368 pkt_dev->odevname, ret);
3371 case NETDEV_TX_LOCKED:
3372 case NETDEV_TX_BUSY:
3373 /* Retry it next time */
3374 atomic_dec(&(pkt_dev->skb->users));
3375 pkt_dev->last_ok = 0;
3378 HARD_TX_UNLOCK(odev, txq);
3382 /* If pkt_dev->count is zero, then run forever */
3383 if ((pkt_dev->count != 0) && (pkt_dev->sofar >= pkt_dev->count)) {
3384 pktgen_wait_for_skb(pkt_dev);
3386 /* Done with this */
3387 pktgen_stop_device(pkt_dev);
3392 * Main loop of the thread goes here
3395 static int pktgen_thread_worker(void *arg)
3398 struct pktgen_thread *t = arg;
3399 struct pktgen_dev *pkt_dev = NULL;
3402 BUG_ON(smp_processor_id() != cpu);
3404 init_waitqueue_head(&t->queue);
3405 complete(&t->start_done);
3407 pr_debug("starting pktgen/%d: pid=%d\n", cpu, task_pid_nr(current));
3411 __set_current_state(TASK_RUNNING);
3413 while (!kthread_should_stop()) {
3414 pkt_dev = next_to_run(t);
3416 if (unlikely(!pkt_dev && t->control == 0)) {
3417 if (t->net->pktgen_exiting)
3419 wait_event_interruptible_timeout(t->queue,
3426 if (likely(pkt_dev)) {
3427 pktgen_xmit(pkt_dev);
3430 pktgen_resched(pkt_dev);
3435 if (t->control & T_STOP) {
3437 t->control &= ~(T_STOP);
3440 if (t->control & T_RUN) {
3442 t->control &= ~(T_RUN);
3445 if (t->control & T_REMDEVALL) {
3446 pktgen_rem_all_ifs(t);
3447 t->control &= ~(T_REMDEVALL);
3450 if (t->control & T_REMDEV) {
3451 pktgen_rem_one_if(t);
3452 t->control &= ~(T_REMDEV);
3457 set_current_state(TASK_INTERRUPTIBLE);
3459 pr_debug("%s stopping all device\n", t->tsk->comm);
3462 pr_debug("%s removing all device\n", t->tsk->comm);
3463 pktgen_rem_all_ifs(t);
3465 pr_debug("%s removing thread\n", t->tsk->comm);
3466 pktgen_rem_thread(t);
3468 /* Wait for kthread_stop */
3469 while (!kthread_should_stop()) {
3470 set_current_state(TASK_INTERRUPTIBLE);
3473 __set_current_state(TASK_RUNNING);
3478 static struct pktgen_dev *pktgen_find_dev(struct pktgen_thread *t,
3479 const char *ifname, bool exact)
3481 struct pktgen_dev *p, *pkt_dev = NULL;
3482 size_t len = strlen(ifname);
3485 list_for_each_entry_rcu(p, &t->if_list, list)
3486 if (strncmp(p->odevname, ifname, len) == 0) {
3487 if (p->odevname[len]) {
3488 if (exact || p->odevname[len] != '@')
3496 pr_debug("find_dev(%s) returning %p\n", ifname, pkt_dev);
3501 * Adds a dev at front of if_list.
3504 static int add_dev_to_thread(struct pktgen_thread *t,
3505 struct pktgen_dev *pkt_dev)
3509 /* This function cannot be called concurrently, as its called
3510 * under pktgen_thread_lock mutex, but it can run from
3511 * userspace on another CPU than the kthread. The if_lock()
3512 * is used here to sync with concurrent instances of
3513 * _rem_dev_from_if_list() invoked via kthread, which is also
3514 * updating the if_list */
3517 if (pkt_dev->pg_thread) {
3518 pr_err("ERROR: already assigned to a thread\n");
3523 pkt_dev->running = 0;
3524 pkt_dev->pg_thread = t;
3525 list_add_rcu(&pkt_dev->list, &t->if_list);
3532 /* Called under thread lock */
3534 static int pktgen_add_device(struct pktgen_thread *t, const char *ifname)
3536 struct pktgen_dev *pkt_dev;
3538 int node = cpu_to_node(t->cpu);
3540 /* We don't allow a device to be on several threads */
3542 pkt_dev = __pktgen_NN_threads(t->net, ifname, FIND);
3544 pr_err("ERROR: interface already used\n");
3548 pkt_dev = kzalloc_node(sizeof(struct pktgen_dev), GFP_KERNEL, node);
3552 strcpy(pkt_dev->odevname, ifname);
3553 pkt_dev->flows = vzalloc_node(MAX_CFLOWS * sizeof(struct flow_state),
3555 if (pkt_dev->flows == NULL) {
3560 pkt_dev->removal_mark = 0;
3561 pkt_dev->nfrags = 0;
3562 pkt_dev->delay = pg_delay_d;
3563 pkt_dev->count = pg_count_d;
3565 pkt_dev->udp_src_min = 9; /* sink port */
3566 pkt_dev->udp_src_max = 9;
3567 pkt_dev->udp_dst_min = 9;
3568 pkt_dev->udp_dst_max = 9;
3569 pkt_dev->vlan_p = 0;
3570 pkt_dev->vlan_cfi = 0;
3571 pkt_dev->vlan_id = 0xffff;
3572 pkt_dev->svlan_p = 0;
3573 pkt_dev->svlan_cfi = 0;
3574 pkt_dev->svlan_id = 0xffff;
3577 err = pktgen_setup_dev(t->net, pkt_dev, ifname);
3580 if (pkt_dev->odev->priv_flags & IFF_TX_SKB_SHARING)
3581 pkt_dev->clone_skb = pg_clone_skb_d;
3583 pkt_dev->entry = proc_create_data(ifname, 0600, t->net->proc_dir,
3584 &pktgen_if_fops, pkt_dev);
3585 if (!pkt_dev->entry) {
3586 pr_err("cannot create %s/%s procfs entry\n",
3587 PG_PROC_DIR, ifname);
3592 pkt_dev->ipsmode = XFRM_MODE_TRANSPORT;
3593 pkt_dev->ipsproto = IPPROTO_ESP;
3595 /* xfrm tunnel mode needs additional dst to extract outter
3596 * ip header protocol/ttl/id field, here creat a phony one.
3597 * instead of looking for a valid rt, which definitely hurting
3598 * performance under such circumstance.
3600 pkt_dev->dstops.family = AF_INET;
3601 pkt_dev->dst.dev = pkt_dev->odev;
3602 dst_init_metrics(&pkt_dev->dst, pktgen_dst_metrics, false);
3603 pkt_dev->dst.child = &pkt_dev->dst;
3604 pkt_dev->dst.ops = &pkt_dev->dstops;
3607 return add_dev_to_thread(t, pkt_dev);
3609 dev_put(pkt_dev->odev);
3614 vfree(pkt_dev->flows);
3619 static int __net_init pktgen_create_thread(int cpu, struct pktgen_net *pn)
3621 struct pktgen_thread *t;
3622 struct proc_dir_entry *pe;
3623 struct task_struct *p;
3625 t = kzalloc_node(sizeof(struct pktgen_thread), GFP_KERNEL,
3628 pr_err("ERROR: out of memory, can't create new thread\n");
3632 spin_lock_init(&t->if_lock);
3635 INIT_LIST_HEAD(&t->if_list);
3637 list_add_tail(&t->th_list, &pn->pktgen_threads);
3638 init_completion(&t->start_done);
3640 p = kthread_create_on_node(pktgen_thread_worker,
3643 "kpktgend_%d", cpu);
3645 pr_err("kernel_thread() failed for cpu %d\n", t->cpu);
3646 list_del(&t->th_list);
3650 kthread_bind(p, cpu);
3653 pe = proc_create_data(t->tsk->comm, 0600, pn->proc_dir,
3654 &pktgen_thread_fops, t);
3656 pr_err("cannot create %s/%s procfs entry\n",
3657 PG_PROC_DIR, t->tsk->comm);
3659 list_del(&t->th_list);
3666 wait_for_completion(&t->start_done);
3672 * Removes a device from the thread if_list.
3674 static void _rem_dev_from_if_list(struct pktgen_thread *t,
3675 struct pktgen_dev *pkt_dev)
3677 struct list_head *q, *n;
3678 struct pktgen_dev *p;
3681 list_for_each_safe(q, n, &t->if_list) {
3682 p = list_entry(q, struct pktgen_dev, list);
3684 list_del_rcu(&p->list);
3689 static int pktgen_remove_device(struct pktgen_thread *t,
3690 struct pktgen_dev *pkt_dev)
3692 pr_debug("remove_device pkt_dev=%p\n", pkt_dev);
3694 if (pkt_dev->running) {
3695 pr_warning("WARNING: trying to remove a running interface, stopping it now\n");
3696 pktgen_stop_device(pkt_dev);
3699 /* Dis-associate from the interface */
3701 if (pkt_dev->odev) {
3702 dev_put(pkt_dev->odev);
3703 pkt_dev->odev = NULL;
3706 /* Remove proc before if_list entry, because add_device uses
3707 * list to determine if interface already exist, avoid race
3708 * with proc_create_data() */
3710 proc_remove(pkt_dev->entry);
3712 /* And update the thread if_list */
3713 _rem_dev_from_if_list(t, pkt_dev);
3718 vfree(pkt_dev->flows);
3720 put_page(pkt_dev->page);
3721 kfree_rcu(pkt_dev, rcu);
3725 static int __net_init pg_net_init(struct net *net)
3727 struct pktgen_net *pn = net_generic(net, pg_net_id);
3728 struct proc_dir_entry *pe;
3732 INIT_LIST_HEAD(&pn->pktgen_threads);
3733 pn->pktgen_exiting = false;
3734 pn->proc_dir = proc_mkdir(PG_PROC_DIR, pn->net->proc_net);
3735 if (!pn->proc_dir) {
3736 pr_warn("cannot create /proc/net/%s\n", PG_PROC_DIR);
3739 pe = proc_create(PGCTRL, 0600, pn->proc_dir, &pktgen_fops);
3741 pr_err("cannot create %s procfs entry\n", PGCTRL);
3746 for_each_online_cpu(cpu) {
3749 err = pktgen_create_thread(cpu, pn);
3751 pr_warn("Cannot create thread for cpu %d (%d)\n",
3755 if (list_empty(&pn->pktgen_threads)) {
3756 pr_err("Initialization failed for all threads\n");
3764 remove_proc_entry(PGCTRL, pn->proc_dir);
3766 remove_proc_entry(PG_PROC_DIR, pn->net->proc_net);
3770 static void __net_exit pg_net_exit(struct net *net)
3772 struct pktgen_net *pn = net_generic(net, pg_net_id);
3773 struct pktgen_thread *t;
3774 struct list_head *q, *n;
3777 /* Stop all interfaces & threads */
3778 pn->pktgen_exiting = true;
3780 mutex_lock(&pktgen_thread_lock);
3781 list_splice_init(&pn->pktgen_threads, &list);
3782 mutex_unlock(&pktgen_thread_lock);
3784 list_for_each_safe(q, n, &list) {
3785 t = list_entry(q, struct pktgen_thread, th_list);
3786 list_del(&t->th_list);
3787 kthread_stop(t->tsk);
3791 remove_proc_entry(PGCTRL, pn->proc_dir);
3792 remove_proc_entry(PG_PROC_DIR, pn->net->proc_net);
3795 static struct pernet_operations pg_net_ops = {
3796 .init = pg_net_init,
3797 .exit = pg_net_exit,
3799 .size = sizeof(struct pktgen_net),
3802 static int __init pg_init(void)
3806 pr_info("%s", version);
3807 ret = register_pernet_subsys(&pg_net_ops);
3810 ret = register_netdevice_notifier(&pktgen_notifier_block);
3812 unregister_pernet_subsys(&pg_net_ops);
3817 static void __exit pg_cleanup(void)
3819 unregister_netdevice_notifier(&pktgen_notifier_block);
3820 unregister_pernet_subsys(&pg_net_ops);
3821 /* Don't need rcu_barrier() due to use of kfree_rcu() */
3824 module_init(pg_init);
3825 module_exit(pg_cleanup);
3827 MODULE_AUTHOR("Robert Olsson <robert.olsson@its.uu.se>");
3828 MODULE_DESCRIPTION("Packet Generator tool");
3829 MODULE_LICENSE("GPL");
3830 MODULE_VERSION(VERSION);
3831 module_param(pg_count_d, int, 0);
3832 MODULE_PARM_DESC(pg_count_d, "Default number of packets to inject");
3833 module_param(pg_delay_d, int, 0);
3834 MODULE_PARM_DESC(pg_delay_d, "Default delay between packets (nanoseconds)");
3835 module_param(pg_clone_skb_d, int, 0);
3836 MODULE_PARM_DESC(pg_clone_skb_d, "Default number of copies of the same packet");
3837 module_param(debug, int, 0);
3838 MODULE_PARM_DESC(debug, "Enable debugging of pktgen module");