dm thin: wake worker when discard is prepared
[linux-2.6.git] / net / core / pktgen.c
blobd1dc14c2aac4ee00fa6dab706131bbe41d3790f7
1 /*
2 * Authors:
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
52 * clones.
54 * Also moved to /proc/net/pktgen/
55 * --ro
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.
67 * The new operation:
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
73 * into this too.
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.
82 * --ro
84 * Fix refcount off by one if first packet fails, potential null deref,
85 * memleak 030710- KJP
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>
108 * 050103
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 <linux/prefetch.h>
160 #include <net/net_namespace.h>
161 #include <net/checksum.h>
162 #include <net/ipv6.h>
163 #include <net/addrconf.h>
164 #ifdef CONFIG_XFRM
165 #include <net/xfrm.h>
166 #endif
167 #include <asm/byteorder.h>
168 #include <linux/rcupdate.h>
169 #include <linux/bitops.h>
170 #include <linux/io.h>
171 #include <linux/timex.h>
172 #include <linux/uaccess.h>
173 #include <asm/dma.h>
174 #include <asm/div64.h> /* do_div */
176 #define VERSION "2.74"
177 #define IP_NAME_SZ 32
178 #define MAX_MPLS_LABELS 16 /* This is the max label stack depth */
179 #define MPLS_STACK_BOTTOM htonl(0x00000100)
181 #define func_enter() pr_debug("entering %s\n", __func__);
183 /* Device flag bits */
184 #define F_IPSRC_RND (1<<0) /* IP-Src Random */
185 #define F_IPDST_RND (1<<1) /* IP-Dst Random */
186 #define F_UDPSRC_RND (1<<2) /* UDP-Src Random */
187 #define F_UDPDST_RND (1<<3) /* UDP-Dst Random */
188 #define F_MACSRC_RND (1<<4) /* MAC-Src Random */
189 #define F_MACDST_RND (1<<5) /* MAC-Dst Random */
190 #define F_TXSIZE_RND (1<<6) /* Transmit size is random */
191 #define F_IPV6 (1<<7) /* Interface in IPV6 Mode */
192 #define F_MPLS_RND (1<<8) /* Random MPLS labels */
193 #define F_VID_RND (1<<9) /* Random VLAN ID */
194 #define F_SVID_RND (1<<10) /* Random SVLAN ID */
195 #define F_FLOW_SEQ (1<<11) /* Sequential flows */
196 #define F_IPSEC_ON (1<<12) /* ipsec on for flows */
197 #define F_QUEUE_MAP_RND (1<<13) /* queue map Random */
198 #define F_QUEUE_MAP_CPU (1<<14) /* queue map mirrors smp_processor_id() */
199 #define F_NODE (1<<15) /* Node memory alloc*/
201 /* Thread control flag bits */
202 #define T_STOP (1<<0) /* Stop run */
203 #define T_RUN (1<<1) /* Start run */
204 #define T_REMDEVALL (1<<2) /* Remove all devs */
205 #define T_REMDEV (1<<3) /* Remove one dev */
207 /* If lock -- can be removed after some work */
208 #define if_lock(t) spin_lock(&(t->if_lock));
209 #define if_unlock(t) spin_unlock(&(t->if_lock));
211 /* Used to help with determining the pkts on receive */
212 #define PKTGEN_MAGIC 0xbe9be955
213 #define PG_PROC_DIR "pktgen"
214 #define PGCTRL "pgctrl"
215 static struct proc_dir_entry *pg_proc_dir;
217 #define MAX_CFLOWS 65536
219 #define VLAN_TAG_SIZE(x) ((x)->vlan_id == 0xffff ? 0 : 4)
220 #define SVLAN_TAG_SIZE(x) ((x)->svlan_id == 0xffff ? 0 : 4)
222 struct flow_state {
223 __be32 cur_daddr;
224 int count;
225 #ifdef CONFIG_XFRM
226 struct xfrm_state *x;
227 #endif
228 __u32 flags;
231 /* flow flag bits */
232 #define F_INIT (1<<0) /* flow has been initialized */
234 struct pktgen_dev {
236 * Try to keep frequent/infrequent used vars. separated.
238 struct proc_dir_entry *entry; /* proc file */
239 struct pktgen_thread *pg_thread;/* the owner */
240 struct list_head list; /* chaining in the thread's run-queue */
242 int running; /* if false, the test will stop */
244 /* If min != max, then we will either do a linear iteration, or
245 * we will do a random selection from within the range.
247 __u32 flags;
248 int removal_mark; /* non-zero => the device is marked for
249 * removal by worker thread */
251 int min_pkt_size;
252 int max_pkt_size;
253 int pkt_overhead; /* overhead for MPLS, VLANs, IPSEC etc */
254 int nfrags;
255 struct page *page;
256 u64 delay; /* nano-seconds */
258 __u64 count; /* Default No packets to send */
259 __u64 sofar; /* How many pkts we've sent so far */
260 __u64 tx_bytes; /* How many bytes we've transmitted */
261 __u64 errors; /* Errors when trying to transmit, */
263 /* runtime counters relating to clone_skb */
265 __u64 allocated_skbs;
266 __u32 clone_count;
267 int last_ok; /* Was last skb sent?
268 * Or a failed transmit of some sort?
269 * This will keep sequence numbers in order
271 ktime_t next_tx;
272 ktime_t started_at;
273 ktime_t stopped_at;
274 u64 idle_acc; /* nano-seconds */
276 __u32 seq_num;
278 int clone_skb; /*
279 * Use multiple SKBs during packet gen.
280 * If this number is greater than 1, then
281 * that many copies of the same packet will be
282 * sent before a new packet is allocated.
283 * If you want to send 1024 identical packets
284 * before creating a new packet,
285 * set clone_skb to 1024.
288 char dst_min[IP_NAME_SZ]; /* IP, ie 1.2.3.4 */
289 char dst_max[IP_NAME_SZ]; /* IP, ie 1.2.3.4 */
290 char src_min[IP_NAME_SZ]; /* IP, ie 1.2.3.4 */
291 char src_max[IP_NAME_SZ]; /* IP, ie 1.2.3.4 */
293 struct in6_addr in6_saddr;
294 struct in6_addr in6_daddr;
295 struct in6_addr cur_in6_daddr;
296 struct in6_addr cur_in6_saddr;
297 /* For ranges */
298 struct in6_addr min_in6_daddr;
299 struct in6_addr max_in6_daddr;
300 struct in6_addr min_in6_saddr;
301 struct in6_addr max_in6_saddr;
303 /* If we're doing ranges, random or incremental, then this
304 * defines the min/max for those ranges.
306 __be32 saddr_min; /* inclusive, source IP address */
307 __be32 saddr_max; /* exclusive, source IP address */
308 __be32 daddr_min; /* inclusive, dest IP address */
309 __be32 daddr_max; /* exclusive, dest IP address */
311 __u16 udp_src_min; /* inclusive, source UDP port */
312 __u16 udp_src_max; /* exclusive, source UDP port */
313 __u16 udp_dst_min; /* inclusive, dest UDP port */
314 __u16 udp_dst_max; /* exclusive, dest UDP port */
316 /* DSCP + ECN */
317 __u8 tos; /* six MSB of (former) IPv4 TOS
318 are for dscp codepoint */
319 __u8 traffic_class; /* ditto for the (former) Traffic Class in IPv6
320 (see RFC 3260, sec. 4) */
322 /* MPLS */
323 unsigned int nr_labels; /* Depth of stack, 0 = no MPLS */
324 __be32 labels[MAX_MPLS_LABELS];
326 /* VLAN/SVLAN (802.1Q/Q-in-Q) */
327 __u8 vlan_p;
328 __u8 vlan_cfi;
329 __u16 vlan_id; /* 0xffff means no vlan tag */
331 __u8 svlan_p;
332 __u8 svlan_cfi;
333 __u16 svlan_id; /* 0xffff means no svlan tag */
335 __u32 src_mac_count; /* How many MACs to iterate through */
336 __u32 dst_mac_count; /* How many MACs to iterate through */
338 unsigned char dst_mac[ETH_ALEN];
339 unsigned char src_mac[ETH_ALEN];
341 __u32 cur_dst_mac_offset;
342 __u32 cur_src_mac_offset;
343 __be32 cur_saddr;
344 __be32 cur_daddr;
345 __u16 ip_id;
346 __u16 cur_udp_dst;
347 __u16 cur_udp_src;
348 __u16 cur_queue_map;
349 __u32 cur_pkt_size;
350 __u32 last_pkt_size;
352 __u8 hh[14];
353 /* = {
354 0x00, 0x80, 0xC8, 0x79, 0xB3, 0xCB,
356 We fill in SRC address later
357 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
358 0x08, 0x00
361 __u16 pad; /* pad out the hh struct to an even 16 bytes */
363 struct sk_buff *skb; /* skb we are to transmit next, used for when we
364 * are transmitting the same one multiple times
366 struct net_device *odev; /* The out-going device.
367 * Note that the device should have it's
368 * pg_info pointer pointing back to this
369 * device.
370 * Set when the user specifies the out-going
371 * device name (not when the inject is
372 * started as it used to do.)
374 char odevname[32];
375 struct flow_state *flows;
376 unsigned int cflows; /* Concurrent flows (config) */
377 unsigned int lflow; /* Flow length (config) */
378 unsigned int nflows; /* accumulated flows (stats) */
379 unsigned int curfl; /* current sequenced flow (state)*/
381 u16 queue_map_min;
382 u16 queue_map_max;
383 __u32 skb_priority; /* skb priority field */
384 int node; /* Memory node */
386 #ifdef CONFIG_XFRM
387 __u8 ipsmode; /* IPSEC mode (config) */
388 __u8 ipsproto; /* IPSEC type (config) */
389 #endif
390 char result[512];
393 struct pktgen_hdr {
394 __be32 pgh_magic;
395 __be32 seq_num;
396 __be32 tv_sec;
397 __be32 tv_usec;
400 static bool pktgen_exiting __read_mostly;
402 struct pktgen_thread {
403 spinlock_t if_lock; /* for list of devices */
404 struct list_head if_list; /* All device here */
405 struct list_head th_list;
406 struct task_struct *tsk;
407 char result[512];
409 /* Field for thread to receive "posted" events terminate,
410 stop ifs etc. */
412 u32 control;
413 int cpu;
415 wait_queue_head_t queue;
416 struct completion start_done;
419 #define REMOVE 1
420 #define FIND 0
422 static inline ktime_t ktime_now(void)
424 struct timespec ts;
425 ktime_get_ts(&ts);
427 return timespec_to_ktime(ts);
430 /* This works even if 32 bit because of careful byte order choice */
431 static inline int ktime_lt(const ktime_t cmp1, const ktime_t cmp2)
433 return cmp1.tv64 < cmp2.tv64;
436 static const char version[] =
437 "Packet Generator for packet performance testing. "
438 "Version: " VERSION "\n";
440 static int pktgen_remove_device(struct pktgen_thread *t, struct pktgen_dev *i);
441 static int pktgen_add_device(struct pktgen_thread *t, const char *ifname);
442 static struct pktgen_dev *pktgen_find_dev(struct pktgen_thread *t,
443 const char *ifname, bool exact);
444 static int pktgen_device_event(struct notifier_block *, unsigned long, void *);
445 static void pktgen_run_all_threads(void);
446 static void pktgen_reset_all_threads(void);
447 static void pktgen_stop_all_threads_ifs(void);
449 static void pktgen_stop(struct pktgen_thread *t);
450 static void pktgen_clear_counters(struct pktgen_dev *pkt_dev);
452 /* Module parameters, defaults. */
453 static int pg_count_d __read_mostly = 1000;
454 static int pg_delay_d __read_mostly;
455 static int pg_clone_skb_d __read_mostly;
456 static int debug __read_mostly;
458 static DEFINE_MUTEX(pktgen_thread_lock);
459 static LIST_HEAD(pktgen_threads);
461 static struct notifier_block pktgen_notifier_block = {
462 .notifier_call = pktgen_device_event,
466 * /proc handling functions
470 static int pgctrl_show(struct seq_file *seq, void *v)
472 seq_puts(seq, version);
473 return 0;
476 static ssize_t pgctrl_write(struct file *file, const char __user *buf,
477 size_t count, loff_t *ppos)
479 int err = 0;
480 char data[128];
482 if (!capable(CAP_NET_ADMIN)) {
483 err = -EPERM;
484 goto out;
487 if (count > sizeof(data))
488 count = sizeof(data);
490 if (copy_from_user(data, buf, count)) {
491 err = -EFAULT;
492 goto out;
494 data[count - 1] = 0; /* Make string */
496 if (!strcmp(data, "stop"))
497 pktgen_stop_all_threads_ifs();
499 else if (!strcmp(data, "start"))
500 pktgen_run_all_threads();
502 else if (!strcmp(data, "reset"))
503 pktgen_reset_all_threads();
505 else
506 pr_warning("Unknown command: %s\n", data);
508 err = count;
510 out:
511 return err;
514 static int pgctrl_open(struct inode *inode, struct file *file)
516 return single_open(file, pgctrl_show, PDE(inode)->data);
519 static const struct file_operations pktgen_fops = {
520 .owner = THIS_MODULE,
521 .open = pgctrl_open,
522 .read = seq_read,
523 .llseek = seq_lseek,
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;
531 ktime_t stopped;
532 u64 idle;
534 seq_printf(seq,
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);
539 seq_printf(seq,
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,
545 pkt_dev->lflow);
547 seq_printf(seq,
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) {
557 seq_printf(seq,
558 " saddr: %pI6c min_saddr: %pI6c max_saddr: %pI6c\n"
559 " daddr: %pI6c min_daddr: %pI6c max_daddr: %pI6c\n",
560 &pkt_dev->in6_saddr,
561 &pkt_dev->min_in6_saddr, &pkt_dev->max_in6_saddr,
562 &pkt_dev->in6_daddr,
563 &pkt_dev->min_in6_daddr, &pkt_dev->max_in6_daddr);
564 } else {
565 seq_printf(seq,
566 " dst_min: %s dst_max: %s\n",
567 pkt_dev->dst_min, pkt_dev->dst_max);
568 seq_printf(seq,
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_printf(seq, "dst_mac: ");
580 seq_printf(seq, "%pM\n", pkt_dev->dst_mac);
582 seq_printf(seq,
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);
588 seq_printf(seq,
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) {
593 unsigned int i;
594 seq_printf(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,
603 pkt_dev->vlan_cfi);
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,
608 pkt_dev->svlan_cfi);
610 if (pkt_dev->tos)
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_printf(seq, " Flags: ");
621 if (pkt_dev->flags & F_IPV6)
622 seq_printf(seq, "IPV6 ");
624 if (pkt_dev->flags & F_IPSRC_RND)
625 seq_printf(seq, "IPSRC_RND ");
627 if (pkt_dev->flags & F_IPDST_RND)
628 seq_printf(seq, "IPDST_RND ");
630 if (pkt_dev->flags & F_TXSIZE_RND)
631 seq_printf(seq, "TXSIZE_RND ");
633 if (pkt_dev->flags & F_UDPSRC_RND)
634 seq_printf(seq, "UDPSRC_RND ");
636 if (pkt_dev->flags & F_UDPDST_RND)
637 seq_printf(seq, "UDPDST_RND ");
639 if (pkt_dev->flags & F_MPLS_RND)
640 seq_printf(seq, "MPLS_RND ");
642 if (pkt_dev->flags & F_QUEUE_MAP_RND)
643 seq_printf(seq, "QUEUE_MAP_RND ");
645 if (pkt_dev->flags & F_QUEUE_MAP_CPU)
646 seq_printf(seq, "QUEUE_MAP_CPU ");
648 if (pkt_dev->cflows) {
649 if (pkt_dev->flags & F_FLOW_SEQ)
650 seq_printf(seq, "FLOW_SEQ "); /*in sequence flows*/
651 else
652 seq_printf(seq, "FLOW_RND ");
655 #ifdef CONFIG_XFRM
656 if (pkt_dev->flags & F_IPSEC_ON)
657 seq_printf(seq, "IPSEC ");
658 #endif
660 if (pkt_dev->flags & F_MACSRC_RND)
661 seq_printf(seq, "MACSRC_RND ");
663 if (pkt_dev->flags & F_MACDST_RND)
664 seq_printf(seq, "MACDST_RND ");
666 if (pkt_dev->flags & F_VID_RND)
667 seq_printf(seq, "VID_RND ");
669 if (pkt_dev->flags & F_SVID_RND)
670 seq_printf(seq, "SVID_RND ");
672 if (pkt_dev->flags & F_NODE)
673 seq_printf(seq, "NODE_ALLOC ");
675 seq_puts(seq, "\n");
677 /* not really stopped, more like last-running-at */
678 stopped = pkt_dev->running ? ktime_now() : pkt_dev->stopped_at;
679 idle = pkt_dev->idle_acc;
680 do_div(idle, NSEC_PER_USEC);
682 seq_printf(seq,
683 "Current:\n pkts-sofar: %llu errors: %llu\n",
684 (unsigned long long)pkt_dev->sofar,
685 (unsigned long long)pkt_dev->errors);
687 seq_printf(seq,
688 " started: %lluus stopped: %lluus idle: %lluus\n",
689 (unsigned long long) ktime_to_us(pkt_dev->started_at),
690 (unsigned long long) ktime_to_us(stopped),
691 (unsigned long long) idle);
693 seq_printf(seq,
694 " seq_num: %d cur_dst_mac_offset: %d cur_src_mac_offset: %d\n",
695 pkt_dev->seq_num, pkt_dev->cur_dst_mac_offset,
696 pkt_dev->cur_src_mac_offset);
698 if (pkt_dev->flags & F_IPV6) {
699 seq_printf(seq, " cur_saddr: %pI6c cur_daddr: %pI6c\n",
700 &pkt_dev->cur_in6_saddr,
701 &pkt_dev->cur_in6_daddr);
702 } else
703 seq_printf(seq, " cur_saddr: %pI4 cur_daddr: %pI4\n",
704 &pkt_dev->cur_saddr, &pkt_dev->cur_daddr);
706 seq_printf(seq, " cur_udp_dst: %d cur_udp_src: %d\n",
707 pkt_dev->cur_udp_dst, pkt_dev->cur_udp_src);
709 seq_printf(seq, " cur_queue_map: %u\n", pkt_dev->cur_queue_map);
711 seq_printf(seq, " flows: %u\n", pkt_dev->nflows);
713 if (pkt_dev->result[0])
714 seq_printf(seq, "Result: %s\n", pkt_dev->result);
715 else
716 seq_printf(seq, "Result: Idle\n");
718 return 0;
722 static int hex32_arg(const char __user *user_buffer, unsigned long maxlen,
723 __u32 *num)
725 int i = 0;
726 *num = 0;
728 for (; i < maxlen; i++) {
729 int value;
730 char c;
731 *num <<= 4;
732 if (get_user(c, &user_buffer[i]))
733 return -EFAULT;
734 value = hex_to_bin(c);
735 if (value >= 0)
736 *num |= value;
737 else
738 break;
740 return i;
743 static int count_trail_chars(const char __user * user_buffer,
744 unsigned int maxlen)
746 int i;
748 for (i = 0; i < maxlen; i++) {
749 char c;
750 if (get_user(c, &user_buffer[i]))
751 return -EFAULT;
752 switch (c) {
753 case '\"':
754 case '\n':
755 case '\r':
756 case '\t':
757 case ' ':
758 case '=':
759 break;
760 default:
761 goto done;
764 done:
765 return i;
768 static long num_arg(const char __user *user_buffer, unsigned long maxlen,
769 unsigned long *num)
771 int i;
772 *num = 0;
774 for (i = 0; i < maxlen; i++) {
775 char c;
776 if (get_user(c, &user_buffer[i]))
777 return -EFAULT;
778 if ((c >= '0') && (c <= '9')) {
779 *num *= 10;
780 *num += c - '0';
781 } else
782 break;
784 return i;
787 static int strn_len(const char __user * user_buffer, unsigned int maxlen)
789 int i;
791 for (i = 0; i < maxlen; i++) {
792 char c;
793 if (get_user(c, &user_buffer[i]))
794 return -EFAULT;
795 switch (c) {
796 case '\"':
797 case '\n':
798 case '\r':
799 case '\t':
800 case ' ':
801 goto done_str;
802 break;
803 default:
804 break;
807 done_str:
808 return i;
811 static ssize_t get_labels(const char __user *buffer, struct pktgen_dev *pkt_dev)
813 unsigned int n = 0;
814 char c;
815 ssize_t i = 0;
816 int len;
818 pkt_dev->nr_labels = 0;
819 do {
820 __u32 tmp;
821 len = hex32_arg(&buffer[i], 8, &tmp);
822 if (len <= 0)
823 return len;
824 pkt_dev->labels[n] = htonl(tmp);
825 if (pkt_dev->labels[n] & MPLS_STACK_BOTTOM)
826 pkt_dev->flags |= F_MPLS_RND;
827 i += len;
828 if (get_user(c, &buffer[i]))
829 return -EFAULT;
830 i++;
831 n++;
832 if (n >= MAX_MPLS_LABELS)
833 return -E2BIG;
834 } while (c == ',');
836 pkt_dev->nr_labels = n;
837 return i;
840 static ssize_t pktgen_if_write(struct file *file,
841 const char __user * user_buffer, size_t count,
842 loff_t * offset)
844 struct seq_file *seq = file->private_data;
845 struct pktgen_dev *pkt_dev = seq->private;
846 int i, max, len;
847 char name[16], valstr[32];
848 unsigned long value = 0;
849 char *pg_result = NULL;
850 int tmp = 0;
851 char buf[128];
853 pg_result = &(pkt_dev->result[0]);
855 if (count < 1) {
856 pr_warning("wrong command format\n");
857 return -EINVAL;
860 max = count;
861 tmp = count_trail_chars(user_buffer, max);
862 if (tmp < 0) {
863 pr_warning("illegal format\n");
864 return tmp;
866 i = tmp;
868 /* Read variable name */
870 len = strn_len(&user_buffer[i], sizeof(name) - 1);
871 if (len < 0)
872 return len;
874 memset(name, 0, sizeof(name));
875 if (copy_from_user(name, &user_buffer[i], len))
876 return -EFAULT;
877 i += len;
879 max = count - i;
880 len = count_trail_chars(&user_buffer[i], max);
881 if (len < 0)
882 return len;
884 i += len;
886 if (debug) {
887 size_t copy = min_t(size_t, count, 1023);
888 char tb[copy + 1];
889 if (copy_from_user(tb, user_buffer, copy))
890 return -EFAULT;
891 tb[copy] = 0;
892 pr_debug("%s,%lu buffer -:%s:-\n",
893 name, (unsigned long)count, tb);
896 if (!strcmp(name, "min_pkt_size")) {
897 len = num_arg(&user_buffer[i], 10, &value);
898 if (len < 0)
899 return len;
901 i += len;
902 if (value < 14 + 20 + 8)
903 value = 14 + 20 + 8;
904 if (value != pkt_dev->min_pkt_size) {
905 pkt_dev->min_pkt_size = value;
906 pkt_dev->cur_pkt_size = value;
908 sprintf(pg_result, "OK: min_pkt_size=%u",
909 pkt_dev->min_pkt_size);
910 return count;
913 if (!strcmp(name, "max_pkt_size")) {
914 len = num_arg(&user_buffer[i], 10, &value);
915 if (len < 0)
916 return len;
918 i += len;
919 if (value < 14 + 20 + 8)
920 value = 14 + 20 + 8;
921 if (value != pkt_dev->max_pkt_size) {
922 pkt_dev->max_pkt_size = value;
923 pkt_dev->cur_pkt_size = value;
925 sprintf(pg_result, "OK: max_pkt_size=%u",
926 pkt_dev->max_pkt_size);
927 return count;
930 /* Shortcut for min = max */
932 if (!strcmp(name, "pkt_size")) {
933 len = num_arg(&user_buffer[i], 10, &value);
934 if (len < 0)
935 return len;
937 i += len;
938 if (value < 14 + 20 + 8)
939 value = 14 + 20 + 8;
940 if (value != pkt_dev->min_pkt_size) {
941 pkt_dev->min_pkt_size = value;
942 pkt_dev->max_pkt_size = value;
943 pkt_dev->cur_pkt_size = value;
945 sprintf(pg_result, "OK: pkt_size=%u", pkt_dev->min_pkt_size);
946 return count;
949 if (!strcmp(name, "debug")) {
950 len = num_arg(&user_buffer[i], 10, &value);
951 if (len < 0)
952 return len;
954 i += len;
955 debug = value;
956 sprintf(pg_result, "OK: debug=%u", debug);
957 return count;
960 if (!strcmp(name, "frags")) {
961 len = num_arg(&user_buffer[i], 10, &value);
962 if (len < 0)
963 return len;
965 i += len;
966 pkt_dev->nfrags = value;
967 sprintf(pg_result, "OK: frags=%u", pkt_dev->nfrags);
968 return count;
970 if (!strcmp(name, "delay")) {
971 len = num_arg(&user_buffer[i], 10, &value);
972 if (len < 0)
973 return len;
975 i += len;
976 if (value == 0x7FFFFFFF)
977 pkt_dev->delay = ULLONG_MAX;
978 else
979 pkt_dev->delay = (u64)value;
981 sprintf(pg_result, "OK: delay=%llu",
982 (unsigned long long) pkt_dev->delay);
983 return count;
985 if (!strcmp(name, "rate")) {
986 len = num_arg(&user_buffer[i], 10, &value);
987 if (len < 0)
988 return len;
990 i += len;
991 if (!value)
992 return len;
993 pkt_dev->delay = pkt_dev->min_pkt_size*8*NSEC_PER_USEC/value;
994 if (debug)
995 pr_info("Delay set at: %llu ns\n", pkt_dev->delay);
997 sprintf(pg_result, "OK: rate=%lu", value);
998 return count;
1000 if (!strcmp(name, "ratep")) {
1001 len = num_arg(&user_buffer[i], 10, &value);
1002 if (len < 0)
1003 return len;
1005 i += len;
1006 if (!value)
1007 return len;
1008 pkt_dev->delay = NSEC_PER_SEC/value;
1009 if (debug)
1010 pr_info("Delay set at: %llu ns\n", pkt_dev->delay);
1012 sprintf(pg_result, "OK: rate=%lu", value);
1013 return count;
1015 if (!strcmp(name, "udp_src_min")) {
1016 len = num_arg(&user_buffer[i], 10, &value);
1017 if (len < 0)
1018 return len;
1020 i += len;
1021 if (value != pkt_dev->udp_src_min) {
1022 pkt_dev->udp_src_min = value;
1023 pkt_dev->cur_udp_src = value;
1025 sprintf(pg_result, "OK: udp_src_min=%u", pkt_dev->udp_src_min);
1026 return count;
1028 if (!strcmp(name, "udp_dst_min")) {
1029 len = num_arg(&user_buffer[i], 10, &value);
1030 if (len < 0)
1031 return len;
1033 i += len;
1034 if (value != pkt_dev->udp_dst_min) {
1035 pkt_dev->udp_dst_min = value;
1036 pkt_dev->cur_udp_dst = value;
1038 sprintf(pg_result, "OK: udp_dst_min=%u", pkt_dev->udp_dst_min);
1039 return count;
1041 if (!strcmp(name, "udp_src_max")) {
1042 len = num_arg(&user_buffer[i], 10, &value);
1043 if (len < 0)
1044 return len;
1046 i += len;
1047 if (value != pkt_dev->udp_src_max) {
1048 pkt_dev->udp_src_max = value;
1049 pkt_dev->cur_udp_src = value;
1051 sprintf(pg_result, "OK: udp_src_max=%u", pkt_dev->udp_src_max);
1052 return count;
1054 if (!strcmp(name, "udp_dst_max")) {
1055 len = num_arg(&user_buffer[i], 10, &value);
1056 if (len < 0)
1057 return len;
1059 i += len;
1060 if (value != pkt_dev->udp_dst_max) {
1061 pkt_dev->udp_dst_max = value;
1062 pkt_dev->cur_udp_dst = value;
1064 sprintf(pg_result, "OK: udp_dst_max=%u", pkt_dev->udp_dst_max);
1065 return count;
1067 if (!strcmp(name, "clone_skb")) {
1068 len = num_arg(&user_buffer[i], 10, &value);
1069 if (len < 0)
1070 return len;
1071 if ((value > 0) &&
1072 (!(pkt_dev->odev->priv_flags & IFF_TX_SKB_SHARING)))
1073 return -ENOTSUPP;
1074 i += len;
1075 pkt_dev->clone_skb = value;
1077 sprintf(pg_result, "OK: clone_skb=%d", pkt_dev->clone_skb);
1078 return count;
1080 if (!strcmp(name, "count")) {
1081 len = num_arg(&user_buffer[i], 10, &value);
1082 if (len < 0)
1083 return len;
1085 i += len;
1086 pkt_dev->count = value;
1087 sprintf(pg_result, "OK: count=%llu",
1088 (unsigned long long)pkt_dev->count);
1089 return count;
1091 if (!strcmp(name, "src_mac_count")) {
1092 len = num_arg(&user_buffer[i], 10, &value);
1093 if (len < 0)
1094 return len;
1096 i += len;
1097 if (pkt_dev->src_mac_count != value) {
1098 pkt_dev->src_mac_count = value;
1099 pkt_dev->cur_src_mac_offset = 0;
1101 sprintf(pg_result, "OK: src_mac_count=%d",
1102 pkt_dev->src_mac_count);
1103 return count;
1105 if (!strcmp(name, "dst_mac_count")) {
1106 len = num_arg(&user_buffer[i], 10, &value);
1107 if (len < 0)
1108 return len;
1110 i += len;
1111 if (pkt_dev->dst_mac_count != value) {
1112 pkt_dev->dst_mac_count = value;
1113 pkt_dev->cur_dst_mac_offset = 0;
1115 sprintf(pg_result, "OK: dst_mac_count=%d",
1116 pkt_dev->dst_mac_count);
1117 return count;
1119 if (!strcmp(name, "node")) {
1120 len = num_arg(&user_buffer[i], 10, &value);
1121 if (len < 0)
1122 return len;
1124 i += len;
1126 if (node_possible(value)) {
1127 pkt_dev->node = value;
1128 sprintf(pg_result, "OK: node=%d", pkt_dev->node);
1129 if (pkt_dev->page) {
1130 put_page(pkt_dev->page);
1131 pkt_dev->page = NULL;
1134 else
1135 sprintf(pg_result, "ERROR: node not possible");
1136 return count;
1138 if (!strcmp(name, "flag")) {
1139 char f[32];
1140 memset(f, 0, 32);
1141 len = strn_len(&user_buffer[i], sizeof(f) - 1);
1142 if (len < 0)
1143 return len;
1145 if (copy_from_user(f, &user_buffer[i], len))
1146 return -EFAULT;
1147 i += len;
1148 if (strcmp(f, "IPSRC_RND") == 0)
1149 pkt_dev->flags |= F_IPSRC_RND;
1151 else if (strcmp(f, "!IPSRC_RND") == 0)
1152 pkt_dev->flags &= ~F_IPSRC_RND;
1154 else if (strcmp(f, "TXSIZE_RND") == 0)
1155 pkt_dev->flags |= F_TXSIZE_RND;
1157 else if (strcmp(f, "!TXSIZE_RND") == 0)
1158 pkt_dev->flags &= ~F_TXSIZE_RND;
1160 else if (strcmp(f, "IPDST_RND") == 0)
1161 pkt_dev->flags |= F_IPDST_RND;
1163 else if (strcmp(f, "!IPDST_RND") == 0)
1164 pkt_dev->flags &= ~F_IPDST_RND;
1166 else if (strcmp(f, "UDPSRC_RND") == 0)
1167 pkt_dev->flags |= F_UDPSRC_RND;
1169 else if (strcmp(f, "!UDPSRC_RND") == 0)
1170 pkt_dev->flags &= ~F_UDPSRC_RND;
1172 else if (strcmp(f, "UDPDST_RND") == 0)
1173 pkt_dev->flags |= F_UDPDST_RND;
1175 else if (strcmp(f, "!UDPDST_RND") == 0)
1176 pkt_dev->flags &= ~F_UDPDST_RND;
1178 else if (strcmp(f, "MACSRC_RND") == 0)
1179 pkt_dev->flags |= F_MACSRC_RND;
1181 else if (strcmp(f, "!MACSRC_RND") == 0)
1182 pkt_dev->flags &= ~F_MACSRC_RND;
1184 else if (strcmp(f, "MACDST_RND") == 0)
1185 pkt_dev->flags |= F_MACDST_RND;
1187 else if (strcmp(f, "!MACDST_RND") == 0)
1188 pkt_dev->flags &= ~F_MACDST_RND;
1190 else if (strcmp(f, "MPLS_RND") == 0)
1191 pkt_dev->flags |= F_MPLS_RND;
1193 else if (strcmp(f, "!MPLS_RND") == 0)
1194 pkt_dev->flags &= ~F_MPLS_RND;
1196 else if (strcmp(f, "VID_RND") == 0)
1197 pkt_dev->flags |= F_VID_RND;
1199 else if (strcmp(f, "!VID_RND") == 0)
1200 pkt_dev->flags &= ~F_VID_RND;
1202 else if (strcmp(f, "SVID_RND") == 0)
1203 pkt_dev->flags |= F_SVID_RND;
1205 else if (strcmp(f, "!SVID_RND") == 0)
1206 pkt_dev->flags &= ~F_SVID_RND;
1208 else if (strcmp(f, "FLOW_SEQ") == 0)
1209 pkt_dev->flags |= F_FLOW_SEQ;
1211 else if (strcmp(f, "QUEUE_MAP_RND") == 0)
1212 pkt_dev->flags |= F_QUEUE_MAP_RND;
1214 else if (strcmp(f, "!QUEUE_MAP_RND") == 0)
1215 pkt_dev->flags &= ~F_QUEUE_MAP_RND;
1217 else if (strcmp(f, "QUEUE_MAP_CPU") == 0)
1218 pkt_dev->flags |= F_QUEUE_MAP_CPU;
1220 else if (strcmp(f, "!QUEUE_MAP_CPU") == 0)
1221 pkt_dev->flags &= ~F_QUEUE_MAP_CPU;
1222 #ifdef CONFIG_XFRM
1223 else if (strcmp(f, "IPSEC") == 0)
1224 pkt_dev->flags |= F_IPSEC_ON;
1225 #endif
1227 else if (strcmp(f, "!IPV6") == 0)
1228 pkt_dev->flags &= ~F_IPV6;
1230 else if (strcmp(f, "NODE_ALLOC") == 0)
1231 pkt_dev->flags |= F_NODE;
1233 else if (strcmp(f, "!NODE_ALLOC") == 0)
1234 pkt_dev->flags &= ~F_NODE;
1236 else {
1237 sprintf(pg_result,
1238 "Flag -:%s:- unknown\nAvailable flags, (prepend ! to un-set flag):\n%s",
1240 "IPSRC_RND, IPDST_RND, UDPSRC_RND, UDPDST_RND, "
1241 "MACSRC_RND, MACDST_RND, TXSIZE_RND, IPV6, MPLS_RND, VID_RND, SVID_RND, FLOW_SEQ, IPSEC, NODE_ALLOC\n");
1242 return count;
1244 sprintf(pg_result, "OK: flags=0x%x", pkt_dev->flags);
1245 return count;
1247 if (!strcmp(name, "dst_min") || !strcmp(name, "dst")) {
1248 len = strn_len(&user_buffer[i], sizeof(pkt_dev->dst_min) - 1);
1249 if (len < 0)
1250 return len;
1252 if (copy_from_user(buf, &user_buffer[i], len))
1253 return -EFAULT;
1254 buf[len] = 0;
1255 if (strcmp(buf, pkt_dev->dst_min) != 0) {
1256 memset(pkt_dev->dst_min, 0, sizeof(pkt_dev->dst_min));
1257 strncpy(pkt_dev->dst_min, buf, len);
1258 pkt_dev->daddr_min = in_aton(pkt_dev->dst_min);
1259 pkt_dev->cur_daddr = pkt_dev->daddr_min;
1261 if (debug)
1262 pr_debug("dst_min set to: %s\n", pkt_dev->dst_min);
1263 i += len;
1264 sprintf(pg_result, "OK: dst_min=%s", pkt_dev->dst_min);
1265 return count;
1267 if (!strcmp(name, "dst_max")) {
1268 len = strn_len(&user_buffer[i], sizeof(pkt_dev->dst_max) - 1);
1269 if (len < 0)
1270 return len;
1273 if (copy_from_user(buf, &user_buffer[i], len))
1274 return -EFAULT;
1276 buf[len] = 0;
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;
1283 if (debug)
1284 pr_debug("dst_max set to: %s\n", pkt_dev->dst_max);
1285 i += len;
1286 sprintf(pg_result, "OK: dst_max=%s", pkt_dev->dst_max);
1287 return count;
1289 if (!strcmp(name, "dst6")) {
1290 len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1291 if (len < 0)
1292 return len;
1294 pkt_dev->flags |= F_IPV6;
1296 if (copy_from_user(buf, &user_buffer[i], len))
1297 return -EFAULT;
1298 buf[len] = 0;
1300 in6_pton(buf, -1, pkt_dev->in6_daddr.s6_addr, -1, NULL);
1301 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->in6_daddr);
1303 pkt_dev->cur_in6_daddr = pkt_dev->in6_daddr;
1305 if (debug)
1306 pr_debug("dst6 set to: %s\n", buf);
1308 i += len;
1309 sprintf(pg_result, "OK: dst6=%s", buf);
1310 return count;
1312 if (!strcmp(name, "dst6_min")) {
1313 len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1314 if (len < 0)
1315 return len;
1317 pkt_dev->flags |= F_IPV6;
1319 if (copy_from_user(buf, &user_buffer[i], len))
1320 return -EFAULT;
1321 buf[len] = 0;
1323 in6_pton(buf, -1, pkt_dev->min_in6_daddr.s6_addr, -1, NULL);
1324 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->min_in6_daddr);
1326 pkt_dev->cur_in6_daddr = pkt_dev->min_in6_daddr;
1327 if (debug)
1328 pr_debug("dst6_min set to: %s\n", buf);
1330 i += len;
1331 sprintf(pg_result, "OK: dst6_min=%s", buf);
1332 return count;
1334 if (!strcmp(name, "dst6_max")) {
1335 len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1336 if (len < 0)
1337 return len;
1339 pkt_dev->flags |= F_IPV6;
1341 if (copy_from_user(buf, &user_buffer[i], len))
1342 return -EFAULT;
1343 buf[len] = 0;
1345 in6_pton(buf, -1, pkt_dev->max_in6_daddr.s6_addr, -1, NULL);
1346 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->max_in6_daddr);
1348 if (debug)
1349 pr_debug("dst6_max set to: %s\n", buf);
1351 i += len;
1352 sprintf(pg_result, "OK: dst6_max=%s", buf);
1353 return count;
1355 if (!strcmp(name, "src6")) {
1356 len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1357 if (len < 0)
1358 return len;
1360 pkt_dev->flags |= F_IPV6;
1362 if (copy_from_user(buf, &user_buffer[i], len))
1363 return -EFAULT;
1364 buf[len] = 0;
1366 in6_pton(buf, -1, pkt_dev->in6_saddr.s6_addr, -1, NULL);
1367 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->in6_saddr);
1369 pkt_dev->cur_in6_saddr = pkt_dev->in6_saddr;
1371 if (debug)
1372 pr_debug("src6 set to: %s\n", buf);
1374 i += len;
1375 sprintf(pg_result, "OK: src6=%s", buf);
1376 return count;
1378 if (!strcmp(name, "src_min")) {
1379 len = strn_len(&user_buffer[i], sizeof(pkt_dev->src_min) - 1);
1380 if (len < 0)
1381 return len;
1383 if (copy_from_user(buf, &user_buffer[i], len))
1384 return -EFAULT;
1385 buf[len] = 0;
1386 if (strcmp(buf, pkt_dev->src_min) != 0) {
1387 memset(pkt_dev->src_min, 0, sizeof(pkt_dev->src_min));
1388 strncpy(pkt_dev->src_min, buf, len);
1389 pkt_dev->saddr_min = in_aton(pkt_dev->src_min);
1390 pkt_dev->cur_saddr = pkt_dev->saddr_min;
1392 if (debug)
1393 pr_debug("src_min set to: %s\n", pkt_dev->src_min);
1394 i += len;
1395 sprintf(pg_result, "OK: src_min=%s", pkt_dev->src_min);
1396 return count;
1398 if (!strcmp(name, "src_max")) {
1399 len = strn_len(&user_buffer[i], sizeof(pkt_dev->src_max) - 1);
1400 if (len < 0)
1401 return len;
1403 if (copy_from_user(buf, &user_buffer[i], len))
1404 return -EFAULT;
1405 buf[len] = 0;
1406 if (strcmp(buf, pkt_dev->src_max) != 0) {
1407 memset(pkt_dev->src_max, 0, sizeof(pkt_dev->src_max));
1408 strncpy(pkt_dev->src_max, buf, len);
1409 pkt_dev->saddr_max = in_aton(pkt_dev->src_max);
1410 pkt_dev->cur_saddr = pkt_dev->saddr_max;
1412 if (debug)
1413 pr_debug("src_max set to: %s\n", pkt_dev->src_max);
1414 i += len;
1415 sprintf(pg_result, "OK: src_max=%s", pkt_dev->src_max);
1416 return count;
1418 if (!strcmp(name, "dst_mac")) {
1419 len = strn_len(&user_buffer[i], sizeof(valstr) - 1);
1420 if (len < 0)
1421 return len;
1423 memset(valstr, 0, sizeof(valstr));
1424 if (copy_from_user(valstr, &user_buffer[i], len))
1425 return -EFAULT;
1427 if (!mac_pton(valstr, pkt_dev->dst_mac))
1428 return -EINVAL;
1429 /* Set up Dest MAC */
1430 memcpy(&pkt_dev->hh[0], pkt_dev->dst_mac, ETH_ALEN);
1432 sprintf(pg_result, "OK: dstmac %pM", pkt_dev->dst_mac);
1433 return count;
1435 if (!strcmp(name, "src_mac")) {
1436 len = strn_len(&user_buffer[i], sizeof(valstr) - 1);
1437 if (len < 0)
1438 return len;
1440 memset(valstr, 0, sizeof(valstr));
1441 if (copy_from_user(valstr, &user_buffer[i], len))
1442 return -EFAULT;
1444 if (!mac_pton(valstr, pkt_dev->src_mac))
1445 return -EINVAL;
1446 /* Set up Src MAC */
1447 memcpy(&pkt_dev->hh[6], pkt_dev->src_mac, ETH_ALEN);
1449 sprintf(pg_result, "OK: srcmac %pM", pkt_dev->src_mac);
1450 return count;
1453 if (!strcmp(name, "clear_counters")) {
1454 pktgen_clear_counters(pkt_dev);
1455 sprintf(pg_result, "OK: Clearing counters.\n");
1456 return count;
1459 if (!strcmp(name, "flows")) {
1460 len = num_arg(&user_buffer[i], 10, &value);
1461 if (len < 0)
1462 return len;
1464 i += len;
1465 if (value > MAX_CFLOWS)
1466 value = MAX_CFLOWS;
1468 pkt_dev->cflows = value;
1469 sprintf(pg_result, "OK: flows=%u", pkt_dev->cflows);
1470 return count;
1473 if (!strcmp(name, "flowlen")) {
1474 len = num_arg(&user_buffer[i], 10, &value);
1475 if (len < 0)
1476 return len;
1478 i += len;
1479 pkt_dev->lflow = value;
1480 sprintf(pg_result, "OK: flowlen=%u", pkt_dev->lflow);
1481 return count;
1484 if (!strcmp(name, "queue_map_min")) {
1485 len = num_arg(&user_buffer[i], 5, &value);
1486 if (len < 0)
1487 return len;
1489 i += len;
1490 pkt_dev->queue_map_min = value;
1491 sprintf(pg_result, "OK: queue_map_min=%u", pkt_dev->queue_map_min);
1492 return count;
1495 if (!strcmp(name, "queue_map_max")) {
1496 len = num_arg(&user_buffer[i], 5, &value);
1497 if (len < 0)
1498 return len;
1500 i += len;
1501 pkt_dev->queue_map_max = value;
1502 sprintf(pg_result, "OK: queue_map_max=%u", pkt_dev->queue_map_max);
1503 return count;
1506 if (!strcmp(name, "mpls")) {
1507 unsigned int n, cnt;
1509 len = get_labels(&user_buffer[i], pkt_dev);
1510 if (len < 0)
1511 return len;
1512 i += len;
1513 cnt = sprintf(pg_result, "OK: mpls=");
1514 for (n = 0; n < pkt_dev->nr_labels; n++)
1515 cnt += sprintf(pg_result + cnt,
1516 "%08x%s", ntohl(pkt_dev->labels[n]),
1517 n == pkt_dev->nr_labels-1 ? "" : ",");
1519 if (pkt_dev->nr_labels && pkt_dev->vlan_id != 0xffff) {
1520 pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1521 pkt_dev->svlan_id = 0xffff;
1523 if (debug)
1524 pr_debug("VLAN/SVLAN auto turned off\n");
1526 return count;
1529 if (!strcmp(name, "vlan_id")) {
1530 len = num_arg(&user_buffer[i], 4, &value);
1531 if (len < 0)
1532 return len;
1534 i += len;
1535 if (value <= 4095) {
1536 pkt_dev->vlan_id = value; /* turn on VLAN */
1538 if (debug)
1539 pr_debug("VLAN turned on\n");
1541 if (debug && pkt_dev->nr_labels)
1542 pr_debug("MPLS auto turned off\n");
1544 pkt_dev->nr_labels = 0; /* turn off MPLS */
1545 sprintf(pg_result, "OK: vlan_id=%u", pkt_dev->vlan_id);
1546 } else {
1547 pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1548 pkt_dev->svlan_id = 0xffff;
1550 if (debug)
1551 pr_debug("VLAN/SVLAN turned off\n");
1553 return count;
1556 if (!strcmp(name, "vlan_p")) {
1557 len = num_arg(&user_buffer[i], 1, &value);
1558 if (len < 0)
1559 return len;
1561 i += len;
1562 if ((value <= 7) && (pkt_dev->vlan_id != 0xffff)) {
1563 pkt_dev->vlan_p = value;
1564 sprintf(pg_result, "OK: vlan_p=%u", pkt_dev->vlan_p);
1565 } else {
1566 sprintf(pg_result, "ERROR: vlan_p must be 0-7");
1568 return count;
1571 if (!strcmp(name, "vlan_cfi")) {
1572 len = num_arg(&user_buffer[i], 1, &value);
1573 if (len < 0)
1574 return len;
1576 i += len;
1577 if ((value <= 1) && (pkt_dev->vlan_id != 0xffff)) {
1578 pkt_dev->vlan_cfi = value;
1579 sprintf(pg_result, "OK: vlan_cfi=%u", pkt_dev->vlan_cfi);
1580 } else {
1581 sprintf(pg_result, "ERROR: vlan_cfi must be 0-1");
1583 return count;
1586 if (!strcmp(name, "svlan_id")) {
1587 len = num_arg(&user_buffer[i], 4, &value);
1588 if (len < 0)
1589 return len;
1591 i += len;
1592 if ((value <= 4095) && ((pkt_dev->vlan_id != 0xffff))) {
1593 pkt_dev->svlan_id = value; /* turn on SVLAN */
1595 if (debug)
1596 pr_debug("SVLAN turned on\n");
1598 if (debug && pkt_dev->nr_labels)
1599 pr_debug("MPLS auto turned off\n");
1601 pkt_dev->nr_labels = 0; /* turn off MPLS */
1602 sprintf(pg_result, "OK: svlan_id=%u", pkt_dev->svlan_id);
1603 } else {
1604 pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1605 pkt_dev->svlan_id = 0xffff;
1607 if (debug)
1608 pr_debug("VLAN/SVLAN turned off\n");
1610 return count;
1613 if (!strcmp(name, "svlan_p")) {
1614 len = num_arg(&user_buffer[i], 1, &value);
1615 if (len < 0)
1616 return len;
1618 i += len;
1619 if ((value <= 7) && (pkt_dev->svlan_id != 0xffff)) {
1620 pkt_dev->svlan_p = value;
1621 sprintf(pg_result, "OK: svlan_p=%u", pkt_dev->svlan_p);
1622 } else {
1623 sprintf(pg_result, "ERROR: svlan_p must be 0-7");
1625 return count;
1628 if (!strcmp(name, "svlan_cfi")) {
1629 len = num_arg(&user_buffer[i], 1, &value);
1630 if (len < 0)
1631 return len;
1633 i += len;
1634 if ((value <= 1) && (pkt_dev->svlan_id != 0xffff)) {
1635 pkt_dev->svlan_cfi = value;
1636 sprintf(pg_result, "OK: svlan_cfi=%u", pkt_dev->svlan_cfi);
1637 } else {
1638 sprintf(pg_result, "ERROR: svlan_cfi must be 0-1");
1640 return count;
1643 if (!strcmp(name, "tos")) {
1644 __u32 tmp_value = 0;
1645 len = hex32_arg(&user_buffer[i], 2, &tmp_value);
1646 if (len < 0)
1647 return len;
1649 i += len;
1650 if (len == 2) {
1651 pkt_dev->tos = tmp_value;
1652 sprintf(pg_result, "OK: tos=0x%02x", pkt_dev->tos);
1653 } else {
1654 sprintf(pg_result, "ERROR: tos must be 00-ff");
1656 return count;
1659 if (!strcmp(name, "traffic_class")) {
1660 __u32 tmp_value = 0;
1661 len = hex32_arg(&user_buffer[i], 2, &tmp_value);
1662 if (len < 0)
1663 return len;
1665 i += len;
1666 if (len == 2) {
1667 pkt_dev->traffic_class = tmp_value;
1668 sprintf(pg_result, "OK: traffic_class=0x%02x", pkt_dev->traffic_class);
1669 } else {
1670 sprintf(pg_result, "ERROR: traffic_class must be 00-ff");
1672 return count;
1675 if (!strcmp(name, "skb_priority")) {
1676 len = num_arg(&user_buffer[i], 9, &value);
1677 if (len < 0)
1678 return len;
1680 i += len;
1681 pkt_dev->skb_priority = value;
1682 sprintf(pg_result, "OK: skb_priority=%i",
1683 pkt_dev->skb_priority);
1684 return count;
1687 sprintf(pkt_dev->result, "No such parameter \"%s\"", name);
1688 return -EINVAL;
1691 static int pktgen_if_open(struct inode *inode, struct file *file)
1693 return single_open(file, pktgen_if_show, PDE(inode)->data);
1696 static const struct file_operations pktgen_if_fops = {
1697 .owner = THIS_MODULE,
1698 .open = pktgen_if_open,
1699 .read = seq_read,
1700 .llseek = seq_lseek,
1701 .write = pktgen_if_write,
1702 .release = single_release,
1705 static int pktgen_thread_show(struct seq_file *seq, void *v)
1707 struct pktgen_thread *t = seq->private;
1708 const struct pktgen_dev *pkt_dev;
1710 BUG_ON(!t);
1712 seq_printf(seq, "Running: ");
1714 if_lock(t);
1715 list_for_each_entry(pkt_dev, &t->if_list, list)
1716 if (pkt_dev->running)
1717 seq_printf(seq, "%s ", pkt_dev->odevname);
1719 seq_printf(seq, "\nStopped: ");
1721 list_for_each_entry(pkt_dev, &t->if_list, list)
1722 if (!pkt_dev->running)
1723 seq_printf(seq, "%s ", pkt_dev->odevname);
1725 if (t->result[0])
1726 seq_printf(seq, "\nResult: %s\n", t->result);
1727 else
1728 seq_printf(seq, "\nResult: NA\n");
1730 if_unlock(t);
1732 return 0;
1735 static ssize_t pktgen_thread_write(struct file *file,
1736 const char __user * user_buffer,
1737 size_t count, loff_t * offset)
1739 struct seq_file *seq = file->private_data;
1740 struct pktgen_thread *t = seq->private;
1741 int i, max, len, ret;
1742 char name[40];
1743 char *pg_result;
1745 if (count < 1) {
1746 // sprintf(pg_result, "Wrong command format");
1747 return -EINVAL;
1750 max = count;
1751 len = count_trail_chars(user_buffer, max);
1752 if (len < 0)
1753 return len;
1755 i = len;
1757 /* Read variable name */
1759 len = strn_len(&user_buffer[i], sizeof(name) - 1);
1760 if (len < 0)
1761 return len;
1763 memset(name, 0, sizeof(name));
1764 if (copy_from_user(name, &user_buffer[i], len))
1765 return -EFAULT;
1766 i += len;
1768 max = count - i;
1769 len = count_trail_chars(&user_buffer[i], max);
1770 if (len < 0)
1771 return len;
1773 i += len;
1775 if (debug)
1776 pr_debug("t=%s, count=%lu\n", name, (unsigned long)count);
1778 if (!t) {
1779 pr_err("ERROR: No thread\n");
1780 ret = -EINVAL;
1781 goto out;
1784 pg_result = &(t->result[0]);
1786 if (!strcmp(name, "add_device")) {
1787 char f[32];
1788 memset(f, 0, 32);
1789 len = strn_len(&user_buffer[i], sizeof(f) - 1);
1790 if (len < 0) {
1791 ret = len;
1792 goto out;
1794 if (copy_from_user(f, &user_buffer[i], len))
1795 return -EFAULT;
1796 i += len;
1797 mutex_lock(&pktgen_thread_lock);
1798 pktgen_add_device(t, f);
1799 mutex_unlock(&pktgen_thread_lock);
1800 ret = count;
1801 sprintf(pg_result, "OK: add_device=%s", f);
1802 goto out;
1805 if (!strcmp(name, "rem_device_all")) {
1806 mutex_lock(&pktgen_thread_lock);
1807 t->control |= T_REMDEVALL;
1808 mutex_unlock(&pktgen_thread_lock);
1809 schedule_timeout_interruptible(msecs_to_jiffies(125)); /* Propagate thread->control */
1810 ret = count;
1811 sprintf(pg_result, "OK: rem_device_all");
1812 goto out;
1815 if (!strcmp(name, "max_before_softirq")) {
1816 sprintf(pg_result, "OK: Note! max_before_softirq is obsoleted -- Do not use");
1817 ret = count;
1818 goto out;
1821 ret = -EINVAL;
1822 out:
1823 return ret;
1826 static int pktgen_thread_open(struct inode *inode, struct file *file)
1828 return single_open(file, pktgen_thread_show, PDE(inode)->data);
1831 static const struct file_operations pktgen_thread_fops = {
1832 .owner = THIS_MODULE,
1833 .open = pktgen_thread_open,
1834 .read = seq_read,
1835 .llseek = seq_lseek,
1836 .write = pktgen_thread_write,
1837 .release = single_release,
1840 /* Think find or remove for NN */
1841 static struct pktgen_dev *__pktgen_NN_threads(const char *ifname, int remove)
1843 struct pktgen_thread *t;
1844 struct pktgen_dev *pkt_dev = NULL;
1845 bool exact = (remove == FIND);
1847 list_for_each_entry(t, &pktgen_threads, th_list) {
1848 pkt_dev = pktgen_find_dev(t, ifname, exact);
1849 if (pkt_dev) {
1850 if (remove) {
1851 if_lock(t);
1852 pkt_dev->removal_mark = 1;
1853 t->control |= T_REMDEV;
1854 if_unlock(t);
1856 break;
1859 return pkt_dev;
1863 * mark a device for removal
1865 static void pktgen_mark_device(const char *ifname)
1867 struct pktgen_dev *pkt_dev = NULL;
1868 const int max_tries = 10, msec_per_try = 125;
1869 int i = 0;
1871 mutex_lock(&pktgen_thread_lock);
1872 pr_debug("%s: marking %s for removal\n", __func__, ifname);
1874 while (1) {
1876 pkt_dev = __pktgen_NN_threads(ifname, REMOVE);
1877 if (pkt_dev == NULL)
1878 break; /* success */
1880 mutex_unlock(&pktgen_thread_lock);
1881 pr_debug("%s: waiting for %s to disappear....\n",
1882 __func__, ifname);
1883 schedule_timeout_interruptible(msecs_to_jiffies(msec_per_try));
1884 mutex_lock(&pktgen_thread_lock);
1886 if (++i >= max_tries) {
1887 pr_err("%s: timed out after waiting %d msec for device %s to be removed\n",
1888 __func__, msec_per_try * i, ifname);
1889 break;
1894 mutex_unlock(&pktgen_thread_lock);
1897 static void pktgen_change_name(struct net_device *dev)
1899 struct pktgen_thread *t;
1901 list_for_each_entry(t, &pktgen_threads, th_list) {
1902 struct pktgen_dev *pkt_dev;
1904 list_for_each_entry(pkt_dev, &t->if_list, list) {
1905 if (pkt_dev->odev != dev)
1906 continue;
1908 remove_proc_entry(pkt_dev->entry->name, pg_proc_dir);
1910 pkt_dev->entry = proc_create_data(dev->name, 0600,
1911 pg_proc_dir,
1912 &pktgen_if_fops,
1913 pkt_dev);
1914 if (!pkt_dev->entry)
1915 pr_err("can't move proc entry for '%s'\n",
1916 dev->name);
1917 break;
1922 static int pktgen_device_event(struct notifier_block *unused,
1923 unsigned long event, void *ptr)
1925 struct net_device *dev = ptr;
1927 if (!net_eq(dev_net(dev), &init_net) || pktgen_exiting)
1928 return NOTIFY_DONE;
1930 /* It is OK that we do not hold the group lock right now,
1931 * as we run under the RTNL lock.
1934 switch (event) {
1935 case NETDEV_CHANGENAME:
1936 pktgen_change_name(dev);
1937 break;
1939 case NETDEV_UNREGISTER:
1940 pktgen_mark_device(dev->name);
1941 break;
1944 return NOTIFY_DONE;
1947 static struct net_device *pktgen_dev_get_by_name(struct pktgen_dev *pkt_dev,
1948 const char *ifname)
1950 char b[IFNAMSIZ+5];
1951 int i;
1953 for (i = 0; ifname[i] != '@'; i++) {
1954 if (i == IFNAMSIZ)
1955 break;
1957 b[i] = ifname[i];
1959 b[i] = 0;
1961 return dev_get_by_name(&init_net, b);
1965 /* Associate pktgen_dev with a device. */
1967 static int pktgen_setup_dev(struct pktgen_dev *pkt_dev, const char *ifname)
1969 struct net_device *odev;
1970 int err;
1972 /* Clean old setups */
1973 if (pkt_dev->odev) {
1974 dev_put(pkt_dev->odev);
1975 pkt_dev->odev = NULL;
1978 odev = pktgen_dev_get_by_name(pkt_dev, ifname);
1979 if (!odev) {
1980 pr_err("no such netdevice: \"%s\"\n", ifname);
1981 return -ENODEV;
1984 if (odev->type != ARPHRD_ETHER) {
1985 pr_err("not an ethernet device: \"%s\"\n", ifname);
1986 err = -EINVAL;
1987 } else if (!netif_running(odev)) {
1988 pr_err("device is down: \"%s\"\n", ifname);
1989 err = -ENETDOWN;
1990 } else {
1991 pkt_dev->odev = odev;
1992 return 0;
1995 dev_put(odev);
1996 return err;
1999 /* Read pkt_dev from the interface and set up internal pktgen_dev
2000 * structure to have the right information to create/send packets
2002 static void pktgen_setup_inject(struct pktgen_dev *pkt_dev)
2004 int ntxq;
2006 if (!pkt_dev->odev) {
2007 pr_err("ERROR: pkt_dev->odev == NULL in setup_inject\n");
2008 sprintf(pkt_dev->result,
2009 "ERROR: pkt_dev->odev == NULL in setup_inject.\n");
2010 return;
2013 /* make sure that we don't pick a non-existing transmit queue */
2014 ntxq = pkt_dev->odev->real_num_tx_queues;
2016 if (ntxq <= pkt_dev->queue_map_min) {
2017 pr_warning("WARNING: Requested queue_map_min (zero-based) (%d) exceeds valid range [0 - %d] for (%d) queues on %s, resetting\n",
2018 pkt_dev->queue_map_min, (ntxq ?: 1) - 1, ntxq,
2019 pkt_dev->odevname);
2020 pkt_dev->queue_map_min = (ntxq ?: 1) - 1;
2022 if (pkt_dev->queue_map_max >= ntxq) {
2023 pr_warning("WARNING: Requested queue_map_max (zero-based) (%d) exceeds valid range [0 - %d] for (%d) queues on %s, resetting\n",
2024 pkt_dev->queue_map_max, (ntxq ?: 1) - 1, ntxq,
2025 pkt_dev->odevname);
2026 pkt_dev->queue_map_max = (ntxq ?: 1) - 1;
2029 /* Default to the interface's mac if not explicitly set. */
2031 if (is_zero_ether_addr(pkt_dev->src_mac))
2032 memcpy(&(pkt_dev->hh[6]), pkt_dev->odev->dev_addr, ETH_ALEN);
2034 /* Set up Dest MAC */
2035 memcpy(&(pkt_dev->hh[0]), pkt_dev->dst_mac, ETH_ALEN);
2037 if (pkt_dev->flags & F_IPV6) {
2038 int i, set = 0, err = 1;
2039 struct inet6_dev *idev;
2041 if (pkt_dev->min_pkt_size == 0) {
2042 pkt_dev->min_pkt_size = 14 + sizeof(struct ipv6hdr)
2043 + sizeof(struct udphdr)
2044 + sizeof(struct pktgen_hdr)
2045 + pkt_dev->pkt_overhead;
2048 for (i = 0; i < IN6_ADDR_HSIZE; i++)
2049 if (pkt_dev->cur_in6_saddr.s6_addr[i]) {
2050 set = 1;
2051 break;
2054 if (!set) {
2057 * Use linklevel address if unconfigured.
2059 * use ipv6_get_lladdr if/when it's get exported
2062 rcu_read_lock();
2063 idev = __in6_dev_get(pkt_dev->odev);
2064 if (idev) {
2065 struct inet6_ifaddr *ifp;
2067 read_lock_bh(&idev->lock);
2068 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2069 if ((ifp->scope & IFA_LINK) &&
2070 !(ifp->flags & IFA_F_TENTATIVE)) {
2071 pkt_dev->cur_in6_saddr = ifp->addr;
2072 err = 0;
2073 break;
2076 read_unlock_bh(&idev->lock);
2078 rcu_read_unlock();
2079 if (err)
2080 pr_err("ERROR: IPv6 link address not available\n");
2082 } else {
2083 if (pkt_dev->min_pkt_size == 0) {
2084 pkt_dev->min_pkt_size = 14 + sizeof(struct iphdr)
2085 + sizeof(struct udphdr)
2086 + sizeof(struct pktgen_hdr)
2087 + pkt_dev->pkt_overhead;
2090 pkt_dev->saddr_min = 0;
2091 pkt_dev->saddr_max = 0;
2092 if (strlen(pkt_dev->src_min) == 0) {
2094 struct in_device *in_dev;
2096 rcu_read_lock();
2097 in_dev = __in_dev_get_rcu(pkt_dev->odev);
2098 if (in_dev) {
2099 if (in_dev->ifa_list) {
2100 pkt_dev->saddr_min =
2101 in_dev->ifa_list->ifa_address;
2102 pkt_dev->saddr_max = pkt_dev->saddr_min;
2105 rcu_read_unlock();
2106 } else {
2107 pkt_dev->saddr_min = in_aton(pkt_dev->src_min);
2108 pkt_dev->saddr_max = in_aton(pkt_dev->src_max);
2111 pkt_dev->daddr_min = in_aton(pkt_dev->dst_min);
2112 pkt_dev->daddr_max = in_aton(pkt_dev->dst_max);
2114 /* Initialize current values. */
2115 pkt_dev->cur_pkt_size = pkt_dev->min_pkt_size;
2116 if (pkt_dev->min_pkt_size > pkt_dev->max_pkt_size)
2117 pkt_dev->max_pkt_size = pkt_dev->min_pkt_size;
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;
2132 s64 remaining;
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);
2141 return;
2144 start_time = ktime_now();
2145 if (remaining < 100000) {
2146 /* for small delays (<100us), just loop until limit is reached */
2147 do {
2148 end_time = ktime_now();
2149 } while (ktime_lt(end_time, spin_until));
2150 } else {
2151 /* see do_nanosleep */
2152 hrtimer_init_sleeper(&t, current);
2153 do {
2154 set_current_state(TASK_INTERRUPTIBLE);
2155 hrtimer_start_expires(&t.timer, HRTIMER_MODE_ABS);
2156 if (!hrtimer_active(&t.timer))
2157 t.task = NULL;
2159 if (likely(t.task))
2160 schedule();
2162 hrtimer_cancel(&t.timer);
2163 } while (t.task && pkt_dev->running && !signal_pending(current));
2164 __set_current_state(TASK_RUNNING);
2165 end_time = ktime_now();
2168 pkt_dev->idle_acc += ktime_to_ns(ktime_sub(end_time, start_time));
2169 pkt_dev->next_tx = ktime_add_ns(spin_until, pkt_dev->delay);
2172 static inline void set_pkt_overhead(struct pktgen_dev *pkt_dev)
2174 pkt_dev->pkt_overhead = 0;
2175 pkt_dev->pkt_overhead += pkt_dev->nr_labels*sizeof(u32);
2176 pkt_dev->pkt_overhead += VLAN_TAG_SIZE(pkt_dev);
2177 pkt_dev->pkt_overhead += SVLAN_TAG_SIZE(pkt_dev);
2180 static inline int f_seen(const struct pktgen_dev *pkt_dev, int flow)
2182 return !!(pkt_dev->flows[flow].flags & F_INIT);
2185 static inline int f_pick(struct pktgen_dev *pkt_dev)
2187 int flow = pkt_dev->curfl;
2189 if (pkt_dev->flags & F_FLOW_SEQ) {
2190 if (pkt_dev->flows[flow].count >= pkt_dev->lflow) {
2191 /* reset time */
2192 pkt_dev->flows[flow].count = 0;
2193 pkt_dev->flows[flow].flags = 0;
2194 pkt_dev->curfl += 1;
2195 if (pkt_dev->curfl >= pkt_dev->cflows)
2196 pkt_dev->curfl = 0; /*reset */
2198 } else {
2199 flow = random32() % pkt_dev->cflows;
2200 pkt_dev->curfl = flow;
2202 if (pkt_dev->flows[flow].count > pkt_dev->lflow) {
2203 pkt_dev->flows[flow].count = 0;
2204 pkt_dev->flows[flow].flags = 0;
2208 return pkt_dev->curfl;
2212 #ifdef CONFIG_XFRM
2213 /* If there was already an IPSEC SA, we keep it as is, else
2214 * we go look for it ...
2216 #define DUMMY_MARK 0
2217 static void get_ipsec_sa(struct pktgen_dev *pkt_dev, int flow)
2219 struct xfrm_state *x = pkt_dev->flows[flow].x;
2220 if (!x) {
2221 /*slow path: we dont already have xfrm_state*/
2222 x = xfrm_stateonly_find(&init_net, DUMMY_MARK,
2223 (xfrm_address_t *)&pkt_dev->cur_daddr,
2224 (xfrm_address_t *)&pkt_dev->cur_saddr,
2225 AF_INET,
2226 pkt_dev->ipsmode,
2227 pkt_dev->ipsproto, 0);
2228 if (x) {
2229 pkt_dev->flows[flow].x = x;
2230 set_pkt_overhead(pkt_dev);
2231 pkt_dev->pkt_overhead += x->props.header_len;
2236 #endif
2237 static void set_cur_queue_map(struct pktgen_dev *pkt_dev)
2240 if (pkt_dev->flags & F_QUEUE_MAP_CPU)
2241 pkt_dev->cur_queue_map = smp_processor_id();
2243 else if (pkt_dev->queue_map_min <= pkt_dev->queue_map_max) {
2244 __u16 t;
2245 if (pkt_dev->flags & F_QUEUE_MAP_RND) {
2246 t = random32() %
2247 (pkt_dev->queue_map_max -
2248 pkt_dev->queue_map_min + 1)
2249 + pkt_dev->queue_map_min;
2250 } else {
2251 t = pkt_dev->cur_queue_map + 1;
2252 if (t > pkt_dev->queue_map_max)
2253 t = pkt_dev->queue_map_min;
2255 pkt_dev->cur_queue_map = t;
2257 pkt_dev->cur_queue_map = pkt_dev->cur_queue_map % pkt_dev->odev->real_num_tx_queues;
2260 /* Increment/randomize headers according to flags and current values
2261 * for IP src/dest, UDP src/dst port, MAC-Addr src/dst
2263 static void mod_cur_headers(struct pktgen_dev *pkt_dev)
2265 __u32 imn;
2266 __u32 imx;
2267 int flow = 0;
2269 if (pkt_dev->cflows)
2270 flow = f_pick(pkt_dev);
2272 /* Deal with source MAC */
2273 if (pkt_dev->src_mac_count > 1) {
2274 __u32 mc;
2275 __u32 tmp;
2277 if (pkt_dev->flags & F_MACSRC_RND)
2278 mc = random32() % pkt_dev->src_mac_count;
2279 else {
2280 mc = pkt_dev->cur_src_mac_offset++;
2281 if (pkt_dev->cur_src_mac_offset >=
2282 pkt_dev->src_mac_count)
2283 pkt_dev->cur_src_mac_offset = 0;
2286 tmp = pkt_dev->src_mac[5] + (mc & 0xFF);
2287 pkt_dev->hh[11] = tmp;
2288 tmp = (pkt_dev->src_mac[4] + ((mc >> 8) & 0xFF) + (tmp >> 8));
2289 pkt_dev->hh[10] = tmp;
2290 tmp = (pkt_dev->src_mac[3] + ((mc >> 16) & 0xFF) + (tmp >> 8));
2291 pkt_dev->hh[9] = tmp;
2292 tmp = (pkt_dev->src_mac[2] + ((mc >> 24) & 0xFF) + (tmp >> 8));
2293 pkt_dev->hh[8] = tmp;
2294 tmp = (pkt_dev->src_mac[1] + (tmp >> 8));
2295 pkt_dev->hh[7] = tmp;
2298 /* Deal with Destination MAC */
2299 if (pkt_dev->dst_mac_count > 1) {
2300 __u32 mc;
2301 __u32 tmp;
2303 if (pkt_dev->flags & F_MACDST_RND)
2304 mc = random32() % pkt_dev->dst_mac_count;
2306 else {
2307 mc = pkt_dev->cur_dst_mac_offset++;
2308 if (pkt_dev->cur_dst_mac_offset >=
2309 pkt_dev->dst_mac_count) {
2310 pkt_dev->cur_dst_mac_offset = 0;
2314 tmp = pkt_dev->dst_mac[5] + (mc & 0xFF);
2315 pkt_dev->hh[5] = tmp;
2316 tmp = (pkt_dev->dst_mac[4] + ((mc >> 8) & 0xFF) + (tmp >> 8));
2317 pkt_dev->hh[4] = tmp;
2318 tmp = (pkt_dev->dst_mac[3] + ((mc >> 16) & 0xFF) + (tmp >> 8));
2319 pkt_dev->hh[3] = tmp;
2320 tmp = (pkt_dev->dst_mac[2] + ((mc >> 24) & 0xFF) + (tmp >> 8));
2321 pkt_dev->hh[2] = tmp;
2322 tmp = (pkt_dev->dst_mac[1] + (tmp >> 8));
2323 pkt_dev->hh[1] = tmp;
2326 if (pkt_dev->flags & F_MPLS_RND) {
2327 unsigned int i;
2328 for (i = 0; i < pkt_dev->nr_labels; i++)
2329 if (pkt_dev->labels[i] & MPLS_STACK_BOTTOM)
2330 pkt_dev->labels[i] = MPLS_STACK_BOTTOM |
2331 ((__force __be32)random32() &
2332 htonl(0x000fffff));
2335 if ((pkt_dev->flags & F_VID_RND) && (pkt_dev->vlan_id != 0xffff)) {
2336 pkt_dev->vlan_id = random32() & (4096-1);
2339 if ((pkt_dev->flags & F_SVID_RND) && (pkt_dev->svlan_id != 0xffff)) {
2340 pkt_dev->svlan_id = random32() & (4096 - 1);
2343 if (pkt_dev->udp_src_min < pkt_dev->udp_src_max) {
2344 if (pkt_dev->flags & F_UDPSRC_RND)
2345 pkt_dev->cur_udp_src = random32() %
2346 (pkt_dev->udp_src_max - pkt_dev->udp_src_min)
2347 + pkt_dev->udp_src_min;
2349 else {
2350 pkt_dev->cur_udp_src++;
2351 if (pkt_dev->cur_udp_src >= pkt_dev->udp_src_max)
2352 pkt_dev->cur_udp_src = pkt_dev->udp_src_min;
2356 if (pkt_dev->udp_dst_min < pkt_dev->udp_dst_max) {
2357 if (pkt_dev->flags & F_UDPDST_RND) {
2358 pkt_dev->cur_udp_dst = random32() %
2359 (pkt_dev->udp_dst_max - pkt_dev->udp_dst_min)
2360 + pkt_dev->udp_dst_min;
2361 } else {
2362 pkt_dev->cur_udp_dst++;
2363 if (pkt_dev->cur_udp_dst >= pkt_dev->udp_dst_max)
2364 pkt_dev->cur_udp_dst = pkt_dev->udp_dst_min;
2368 if (!(pkt_dev->flags & F_IPV6)) {
2370 imn = ntohl(pkt_dev->saddr_min);
2371 imx = ntohl(pkt_dev->saddr_max);
2372 if (imn < imx) {
2373 __u32 t;
2374 if (pkt_dev->flags & F_IPSRC_RND)
2375 t = random32() % (imx - imn) + imn;
2376 else {
2377 t = ntohl(pkt_dev->cur_saddr);
2378 t++;
2379 if (t > imx)
2380 t = imn;
2383 pkt_dev->cur_saddr = htonl(t);
2386 if (pkt_dev->cflows && f_seen(pkt_dev, flow)) {
2387 pkt_dev->cur_daddr = pkt_dev->flows[flow].cur_daddr;
2388 } else {
2389 imn = ntohl(pkt_dev->daddr_min);
2390 imx = ntohl(pkt_dev->daddr_max);
2391 if (imn < imx) {
2392 __u32 t;
2393 __be32 s;
2394 if (pkt_dev->flags & F_IPDST_RND) {
2396 t = random32() % (imx - imn) + imn;
2397 s = htonl(t);
2399 while (ipv4_is_loopback(s) ||
2400 ipv4_is_multicast(s) ||
2401 ipv4_is_lbcast(s) ||
2402 ipv4_is_zeronet(s) ||
2403 ipv4_is_local_multicast(s)) {
2404 t = random32() % (imx - imn) + imn;
2405 s = htonl(t);
2407 pkt_dev->cur_daddr = s;
2408 } else {
2409 t = ntohl(pkt_dev->cur_daddr);
2410 t++;
2411 if (t > imx) {
2412 t = imn;
2414 pkt_dev->cur_daddr = htonl(t);
2417 if (pkt_dev->cflows) {
2418 pkt_dev->flows[flow].flags |= F_INIT;
2419 pkt_dev->flows[flow].cur_daddr =
2420 pkt_dev->cur_daddr;
2421 #ifdef CONFIG_XFRM
2422 if (pkt_dev->flags & F_IPSEC_ON)
2423 get_ipsec_sa(pkt_dev, flow);
2424 #endif
2425 pkt_dev->nflows++;
2428 } else { /* IPV6 * */
2430 if (pkt_dev->min_in6_daddr.s6_addr32[0] == 0 &&
2431 pkt_dev->min_in6_daddr.s6_addr32[1] == 0 &&
2432 pkt_dev->min_in6_daddr.s6_addr32[2] == 0 &&
2433 pkt_dev->min_in6_daddr.s6_addr32[3] == 0) ;
2434 else {
2435 int i;
2437 /* Only random destinations yet */
2439 for (i = 0; i < 4; i++) {
2440 pkt_dev->cur_in6_daddr.s6_addr32[i] =
2441 (((__force __be32)random32() |
2442 pkt_dev->min_in6_daddr.s6_addr32[i]) &
2443 pkt_dev->max_in6_daddr.s6_addr32[i]);
2448 if (pkt_dev->min_pkt_size < pkt_dev->max_pkt_size) {
2449 __u32 t;
2450 if (pkt_dev->flags & F_TXSIZE_RND) {
2451 t = random32() %
2452 (pkt_dev->max_pkt_size - pkt_dev->min_pkt_size)
2453 + pkt_dev->min_pkt_size;
2454 } else {
2455 t = pkt_dev->cur_pkt_size + 1;
2456 if (t > pkt_dev->max_pkt_size)
2457 t = pkt_dev->min_pkt_size;
2459 pkt_dev->cur_pkt_size = t;
2462 set_cur_queue_map(pkt_dev);
2464 pkt_dev->flows[flow].count++;
2468 #ifdef CONFIG_XFRM
2469 static int pktgen_output_ipsec(struct sk_buff *skb, struct pktgen_dev *pkt_dev)
2471 struct xfrm_state *x = pkt_dev->flows[pkt_dev->curfl].x;
2472 int err = 0;
2474 if (!x)
2475 return 0;
2476 /* XXX: we dont support tunnel mode for now until
2477 * we resolve the dst issue */
2478 if (x->props.mode != XFRM_MODE_TRANSPORT)
2479 return 0;
2481 spin_lock(&x->lock);
2483 err = x->outer_mode->output(x, skb);
2484 if (err)
2485 goto error;
2486 err = x->type->output(x, skb);
2487 if (err)
2488 goto error;
2490 x->curlft.bytes += skb->len;
2491 x->curlft.packets++;
2492 error:
2493 spin_unlock(&x->lock);
2494 return err;
2497 static void free_SAs(struct pktgen_dev *pkt_dev)
2499 if (pkt_dev->cflows) {
2500 /* let go of the SAs if we have them */
2501 int i;
2502 for (i = 0; i < pkt_dev->cflows; i++) {
2503 struct xfrm_state *x = pkt_dev->flows[i].x;
2504 if (x) {
2505 xfrm_state_put(x);
2506 pkt_dev->flows[i].x = NULL;
2512 static int process_ipsec(struct pktgen_dev *pkt_dev,
2513 struct sk_buff *skb, __be16 protocol)
2515 if (pkt_dev->flags & F_IPSEC_ON) {
2516 struct xfrm_state *x = pkt_dev->flows[pkt_dev->curfl].x;
2517 int nhead = 0;
2518 if (x) {
2519 int ret;
2520 __u8 *eth;
2521 nhead = x->props.header_len - skb_headroom(skb);
2522 if (nhead > 0) {
2523 ret = pskb_expand_head(skb, nhead, 0, GFP_ATOMIC);
2524 if (ret < 0) {
2525 pr_err("Error expanding ipsec packet %d\n",
2526 ret);
2527 goto err;
2531 /* ipsec is not expecting ll header */
2532 skb_pull(skb, ETH_HLEN);
2533 ret = pktgen_output_ipsec(skb, pkt_dev);
2534 if (ret) {
2535 pr_err("Error creating ipsec packet %d\n", ret);
2536 goto err;
2538 /* restore ll */
2539 eth = (__u8 *) skb_push(skb, ETH_HLEN);
2540 memcpy(eth, pkt_dev->hh, 12);
2541 *(u16 *) &eth[12] = protocol;
2544 return 1;
2545 err:
2546 kfree_skb(skb);
2547 return 0;
2549 #endif
2551 static void mpls_push(__be32 *mpls, struct pktgen_dev *pkt_dev)
2553 unsigned int i;
2554 for (i = 0; i < pkt_dev->nr_labels; i++)
2555 *mpls++ = pkt_dev->labels[i] & ~MPLS_STACK_BOTTOM;
2557 mpls--;
2558 *mpls |= MPLS_STACK_BOTTOM;
2561 static inline __be16 build_tci(unsigned int id, unsigned int cfi,
2562 unsigned int prio)
2564 return htons(id | (cfi << 12) | (prio << 13));
2567 static void pktgen_finalize_skb(struct pktgen_dev *pkt_dev, struct sk_buff *skb,
2568 int datalen)
2570 struct timeval timestamp;
2571 struct pktgen_hdr *pgh;
2573 pgh = (struct pktgen_hdr *)skb_put(skb, sizeof(*pgh));
2574 datalen -= sizeof(*pgh);
2576 if (pkt_dev->nfrags <= 0) {
2577 memset(skb_put(skb, datalen), 0, datalen);
2578 } else {
2579 int frags = pkt_dev->nfrags;
2580 int i, len;
2581 int frag_len;
2584 if (frags > MAX_SKB_FRAGS)
2585 frags = MAX_SKB_FRAGS;
2586 len = datalen - frags * PAGE_SIZE;
2587 if (len > 0) {
2588 memset(skb_put(skb, len), 0, len);
2589 datalen = frags * PAGE_SIZE;
2592 i = 0;
2593 frag_len = (datalen/frags) < PAGE_SIZE ?
2594 (datalen/frags) : PAGE_SIZE;
2595 while (datalen > 0) {
2596 if (unlikely(!pkt_dev->page)) {
2597 int node = numa_node_id();
2599 if (pkt_dev->node >= 0 && (pkt_dev->flags & F_NODE))
2600 node = pkt_dev->node;
2601 pkt_dev->page = alloc_pages_node(node, GFP_KERNEL | __GFP_ZERO, 0);
2602 if (!pkt_dev->page)
2603 break;
2605 get_page(pkt_dev->page);
2606 skb_frag_set_page(skb, i, pkt_dev->page);
2607 skb_shinfo(skb)->frags[i].page_offset = 0;
2608 /*last fragment, fill rest of data*/
2609 if (i == (frags - 1))
2610 skb_frag_size_set(&skb_shinfo(skb)->frags[i],
2611 (datalen < PAGE_SIZE ? datalen : PAGE_SIZE));
2612 else
2613 skb_frag_size_set(&skb_shinfo(skb)->frags[i], frag_len);
2614 datalen -= skb_frag_size(&skb_shinfo(skb)->frags[i]);
2615 skb->len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2616 skb->data_len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2617 i++;
2618 skb_shinfo(skb)->nr_frags = i;
2622 /* Stamp the time, and sequence number,
2623 * convert them to network byte order
2625 pgh->pgh_magic = htonl(PKTGEN_MAGIC);
2626 pgh->seq_num = htonl(pkt_dev->seq_num);
2628 do_gettimeofday(&timestamp);
2629 pgh->tv_sec = htonl(timestamp.tv_sec);
2630 pgh->tv_usec = htonl(timestamp.tv_usec);
2633 static struct sk_buff *fill_packet_ipv4(struct net_device *odev,
2634 struct pktgen_dev *pkt_dev)
2636 struct sk_buff *skb = NULL;
2637 __u8 *eth;
2638 struct udphdr *udph;
2639 int datalen, iplen;
2640 struct iphdr *iph;
2641 __be16 protocol = htons(ETH_P_IP);
2642 __be32 *mpls;
2643 __be16 *vlan_tci = NULL; /* Encapsulates priority and VLAN ID */
2644 __be16 *vlan_encapsulated_proto = NULL; /* packet type ID field (or len) for VLAN tag */
2645 __be16 *svlan_tci = NULL; /* Encapsulates priority and SVLAN ID */
2646 __be16 *svlan_encapsulated_proto = NULL; /* packet type ID field (or len) for SVLAN tag */
2647 u16 queue_map;
2649 if (pkt_dev->nr_labels)
2650 protocol = htons(ETH_P_MPLS_UC);
2652 if (pkt_dev->vlan_id != 0xffff)
2653 protocol = htons(ETH_P_8021Q);
2655 /* Update any of the values, used when we're incrementing various
2656 * fields.
2658 mod_cur_headers(pkt_dev);
2659 queue_map = pkt_dev->cur_queue_map;
2661 datalen = (odev->hard_header_len + 16) & ~0xf;
2663 if (pkt_dev->flags & F_NODE) {
2664 int node;
2666 if (pkt_dev->node >= 0)
2667 node = pkt_dev->node;
2668 else
2669 node = numa_node_id();
2671 skb = __alloc_skb(NET_SKB_PAD + pkt_dev->cur_pkt_size + 64
2672 + datalen + pkt_dev->pkt_overhead, GFP_NOWAIT, 0, node);
2673 if (likely(skb)) {
2674 skb_reserve(skb, NET_SKB_PAD);
2675 skb->dev = odev;
2678 else
2679 skb = __netdev_alloc_skb(odev,
2680 pkt_dev->cur_pkt_size + 64
2681 + datalen + pkt_dev->pkt_overhead, GFP_NOWAIT);
2683 if (!skb) {
2684 sprintf(pkt_dev->result, "No memory");
2685 return NULL;
2687 prefetchw(skb->data);
2689 skb_reserve(skb, datalen);
2691 /* Reserve for ethernet and IP header */
2692 eth = (__u8 *) skb_push(skb, 14);
2693 mpls = (__be32 *)skb_put(skb, pkt_dev->nr_labels*sizeof(__u32));
2694 if (pkt_dev->nr_labels)
2695 mpls_push(mpls, pkt_dev);
2697 if (pkt_dev->vlan_id != 0xffff) {
2698 if (pkt_dev->svlan_id != 0xffff) {
2699 svlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2700 *svlan_tci = build_tci(pkt_dev->svlan_id,
2701 pkt_dev->svlan_cfi,
2702 pkt_dev->svlan_p);
2703 svlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2704 *svlan_encapsulated_proto = htons(ETH_P_8021Q);
2706 vlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2707 *vlan_tci = build_tci(pkt_dev->vlan_id,
2708 pkt_dev->vlan_cfi,
2709 pkt_dev->vlan_p);
2710 vlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2711 *vlan_encapsulated_proto = htons(ETH_P_IP);
2714 skb->network_header = skb->tail;
2715 skb->transport_header = skb->network_header + sizeof(struct iphdr);
2716 skb_put(skb, sizeof(struct iphdr) + sizeof(struct udphdr));
2717 skb_set_queue_mapping(skb, queue_map);
2718 skb->priority = pkt_dev->skb_priority;
2720 iph = ip_hdr(skb);
2721 udph = udp_hdr(skb);
2723 memcpy(eth, pkt_dev->hh, 12);
2724 *(__be16 *) & eth[12] = protocol;
2726 /* Eth + IPh + UDPh + mpls */
2727 datalen = pkt_dev->cur_pkt_size - 14 - 20 - 8 -
2728 pkt_dev->pkt_overhead;
2729 if (datalen < 0 || datalen < sizeof(struct pktgen_hdr))
2730 datalen = sizeof(struct pktgen_hdr);
2732 udph->source = htons(pkt_dev->cur_udp_src);
2733 udph->dest = htons(pkt_dev->cur_udp_dst);
2734 udph->len = htons(datalen + 8); /* DATA + udphdr */
2735 udph->check = 0; /* No checksum */
2737 iph->ihl = 5;
2738 iph->version = 4;
2739 iph->ttl = 32;
2740 iph->tos = pkt_dev->tos;
2741 iph->protocol = IPPROTO_UDP; /* UDP */
2742 iph->saddr = pkt_dev->cur_saddr;
2743 iph->daddr = pkt_dev->cur_daddr;
2744 iph->id = htons(pkt_dev->ip_id);
2745 pkt_dev->ip_id++;
2746 iph->frag_off = 0;
2747 iplen = 20 + 8 + datalen;
2748 iph->tot_len = htons(iplen);
2749 iph->check = 0;
2750 iph->check = ip_fast_csum((void *)iph, iph->ihl);
2751 skb->protocol = protocol;
2752 skb->mac_header = (skb->network_header - ETH_HLEN -
2753 pkt_dev->pkt_overhead);
2754 skb->dev = odev;
2755 skb->pkt_type = PACKET_HOST;
2756 pktgen_finalize_skb(pkt_dev, skb, datalen);
2758 #ifdef CONFIG_XFRM
2759 if (!process_ipsec(pkt_dev, skb, protocol))
2760 return NULL;
2761 #endif
2763 return skb;
2766 static struct sk_buff *fill_packet_ipv6(struct net_device *odev,
2767 struct pktgen_dev *pkt_dev)
2769 struct sk_buff *skb = NULL;
2770 __u8 *eth;
2771 struct udphdr *udph;
2772 int datalen;
2773 struct ipv6hdr *iph;
2774 __be16 protocol = htons(ETH_P_IPV6);
2775 __be32 *mpls;
2776 __be16 *vlan_tci = NULL; /* Encapsulates priority and VLAN ID */
2777 __be16 *vlan_encapsulated_proto = NULL; /* packet type ID field (or len) for VLAN tag */
2778 __be16 *svlan_tci = NULL; /* Encapsulates priority and SVLAN ID */
2779 __be16 *svlan_encapsulated_proto = NULL; /* packet type ID field (or len) for SVLAN tag */
2780 u16 queue_map;
2782 if (pkt_dev->nr_labels)
2783 protocol = htons(ETH_P_MPLS_UC);
2785 if (pkt_dev->vlan_id != 0xffff)
2786 protocol = htons(ETH_P_8021Q);
2788 /* Update any of the values, used when we're incrementing various
2789 * fields.
2791 mod_cur_headers(pkt_dev);
2792 queue_map = pkt_dev->cur_queue_map;
2794 skb = __netdev_alloc_skb(odev,
2795 pkt_dev->cur_pkt_size + 64
2796 + 16 + pkt_dev->pkt_overhead, GFP_NOWAIT);
2797 if (!skb) {
2798 sprintf(pkt_dev->result, "No memory");
2799 return NULL;
2801 prefetchw(skb->data);
2803 skb_reserve(skb, 16);
2805 /* Reserve for ethernet and IP header */
2806 eth = (__u8 *) skb_push(skb, 14);
2807 mpls = (__be32 *)skb_put(skb, pkt_dev->nr_labels*sizeof(__u32));
2808 if (pkt_dev->nr_labels)
2809 mpls_push(mpls, pkt_dev);
2811 if (pkt_dev->vlan_id != 0xffff) {
2812 if (pkt_dev->svlan_id != 0xffff) {
2813 svlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2814 *svlan_tci = build_tci(pkt_dev->svlan_id,
2815 pkt_dev->svlan_cfi,
2816 pkt_dev->svlan_p);
2817 svlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2818 *svlan_encapsulated_proto = htons(ETH_P_8021Q);
2820 vlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2821 *vlan_tci = build_tci(pkt_dev->vlan_id,
2822 pkt_dev->vlan_cfi,
2823 pkt_dev->vlan_p);
2824 vlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2825 *vlan_encapsulated_proto = htons(ETH_P_IPV6);
2828 skb->network_header = skb->tail;
2829 skb->transport_header = skb->network_header + sizeof(struct ipv6hdr);
2830 skb_put(skb, sizeof(struct ipv6hdr) + sizeof(struct udphdr));
2831 skb_set_queue_mapping(skb, queue_map);
2832 skb->priority = pkt_dev->skb_priority;
2833 iph = ipv6_hdr(skb);
2834 udph = udp_hdr(skb);
2836 memcpy(eth, pkt_dev->hh, 12);
2837 *(__be16 *) &eth[12] = protocol;
2839 /* Eth + IPh + UDPh + mpls */
2840 datalen = pkt_dev->cur_pkt_size - 14 -
2841 sizeof(struct ipv6hdr) - sizeof(struct udphdr) -
2842 pkt_dev->pkt_overhead;
2844 if (datalen < 0 || datalen < sizeof(struct pktgen_hdr)) {
2845 datalen = sizeof(struct pktgen_hdr);
2846 net_info_ratelimited("increased datalen to %d\n", datalen);
2849 udph->source = htons(pkt_dev->cur_udp_src);
2850 udph->dest = htons(pkt_dev->cur_udp_dst);
2851 udph->len = htons(datalen + sizeof(struct udphdr));
2852 udph->check = 0; /* No checksum */
2854 *(__be32 *) iph = htonl(0x60000000); /* Version + flow */
2856 if (pkt_dev->traffic_class) {
2857 /* Version + traffic class + flow (0) */
2858 *(__be32 *)iph |= htonl(0x60000000 | (pkt_dev->traffic_class << 20));
2861 iph->hop_limit = 32;
2863 iph->payload_len = htons(sizeof(struct udphdr) + datalen);
2864 iph->nexthdr = IPPROTO_UDP;
2866 iph->daddr = pkt_dev->cur_in6_daddr;
2867 iph->saddr = pkt_dev->cur_in6_saddr;
2869 skb->mac_header = (skb->network_header - ETH_HLEN -
2870 pkt_dev->pkt_overhead);
2871 skb->protocol = protocol;
2872 skb->dev = odev;
2873 skb->pkt_type = PACKET_HOST;
2875 pktgen_finalize_skb(pkt_dev, skb, datalen);
2877 return skb;
2880 static struct sk_buff *fill_packet(struct net_device *odev,
2881 struct pktgen_dev *pkt_dev)
2883 if (pkt_dev->flags & F_IPV6)
2884 return fill_packet_ipv6(odev, pkt_dev);
2885 else
2886 return fill_packet_ipv4(odev, pkt_dev);
2889 static void pktgen_clear_counters(struct pktgen_dev *pkt_dev)
2891 pkt_dev->seq_num = 1;
2892 pkt_dev->idle_acc = 0;
2893 pkt_dev->sofar = 0;
2894 pkt_dev->tx_bytes = 0;
2895 pkt_dev->errors = 0;
2898 /* Set up structure for sending pkts, clear counters */
2900 static void pktgen_run(struct pktgen_thread *t)
2902 struct pktgen_dev *pkt_dev;
2903 int started = 0;
2905 func_enter();
2907 if_lock(t);
2908 list_for_each_entry(pkt_dev, &t->if_list, list) {
2911 * setup odev and create initial packet.
2913 pktgen_setup_inject(pkt_dev);
2915 if (pkt_dev->odev) {
2916 pktgen_clear_counters(pkt_dev);
2917 pkt_dev->running = 1; /* Cranke yeself! */
2918 pkt_dev->skb = NULL;
2919 pkt_dev->started_at =
2920 pkt_dev->next_tx = ktime_now();
2922 set_pkt_overhead(pkt_dev);
2924 strcpy(pkt_dev->result, "Starting");
2925 started++;
2926 } else
2927 strcpy(pkt_dev->result, "Error starting");
2929 if_unlock(t);
2930 if (started)
2931 t->control &= ~(T_STOP);
2934 static void pktgen_stop_all_threads_ifs(void)
2936 struct pktgen_thread *t;
2938 func_enter();
2940 mutex_lock(&pktgen_thread_lock);
2942 list_for_each_entry(t, &pktgen_threads, th_list)
2943 t->control |= T_STOP;
2945 mutex_unlock(&pktgen_thread_lock);
2948 static int thread_is_running(const struct pktgen_thread *t)
2950 const struct pktgen_dev *pkt_dev;
2952 list_for_each_entry(pkt_dev, &t->if_list, list)
2953 if (pkt_dev->running)
2954 return 1;
2955 return 0;
2958 static int pktgen_wait_thread_run(struct pktgen_thread *t)
2960 if_lock(t);
2962 while (thread_is_running(t)) {
2964 if_unlock(t);
2966 msleep_interruptible(100);
2968 if (signal_pending(current))
2969 goto signal;
2970 if_lock(t);
2972 if_unlock(t);
2973 return 1;
2974 signal:
2975 return 0;
2978 static int pktgen_wait_all_threads_run(void)
2980 struct pktgen_thread *t;
2981 int sig = 1;
2983 mutex_lock(&pktgen_thread_lock);
2985 list_for_each_entry(t, &pktgen_threads, th_list) {
2986 sig = pktgen_wait_thread_run(t);
2987 if (sig == 0)
2988 break;
2991 if (sig == 0)
2992 list_for_each_entry(t, &pktgen_threads, th_list)
2993 t->control |= (T_STOP);
2995 mutex_unlock(&pktgen_thread_lock);
2996 return sig;
2999 static void pktgen_run_all_threads(void)
3001 struct pktgen_thread *t;
3003 func_enter();
3005 mutex_lock(&pktgen_thread_lock);
3007 list_for_each_entry(t, &pktgen_threads, th_list)
3008 t->control |= (T_RUN);
3010 mutex_unlock(&pktgen_thread_lock);
3012 /* Propagate thread->control */
3013 schedule_timeout_interruptible(msecs_to_jiffies(125));
3015 pktgen_wait_all_threads_run();
3018 static void pktgen_reset_all_threads(void)
3020 struct pktgen_thread *t;
3022 func_enter();
3024 mutex_lock(&pktgen_thread_lock);
3026 list_for_each_entry(t, &pktgen_threads, th_list)
3027 t->control |= (T_REMDEVALL);
3029 mutex_unlock(&pktgen_thread_lock);
3031 /* Propagate thread->control */
3032 schedule_timeout_interruptible(msecs_to_jiffies(125));
3034 pktgen_wait_all_threads_run();
3037 static void show_results(struct pktgen_dev *pkt_dev, int nr_frags)
3039 __u64 bps, mbps, pps;
3040 char *p = pkt_dev->result;
3041 ktime_t elapsed = ktime_sub(pkt_dev->stopped_at,
3042 pkt_dev->started_at);
3043 ktime_t idle = ns_to_ktime(pkt_dev->idle_acc);
3045 p += sprintf(p, "OK: %llu(c%llu+d%llu) usec, %llu (%dbyte,%dfrags)\n",
3046 (unsigned long long)ktime_to_us(elapsed),
3047 (unsigned long long)ktime_to_us(ktime_sub(elapsed, idle)),
3048 (unsigned long long)ktime_to_us(idle),
3049 (unsigned long long)pkt_dev->sofar,
3050 pkt_dev->cur_pkt_size, nr_frags);
3052 pps = div64_u64(pkt_dev->sofar * NSEC_PER_SEC,
3053 ktime_to_ns(elapsed));
3055 bps = pps * 8 * pkt_dev->cur_pkt_size;
3057 mbps = bps;
3058 do_div(mbps, 1000000);
3059 p += sprintf(p, " %llupps %lluMb/sec (%llubps) errors: %llu",
3060 (unsigned long long)pps,
3061 (unsigned long long)mbps,
3062 (unsigned long long)bps,
3063 (unsigned long long)pkt_dev->errors);
3066 /* Set stopped-at timer, remove from running list, do counters & statistics */
3067 static int pktgen_stop_device(struct pktgen_dev *pkt_dev)
3069 int nr_frags = pkt_dev->skb ? skb_shinfo(pkt_dev->skb)->nr_frags : -1;
3071 if (!pkt_dev->running) {
3072 pr_warning("interface: %s is already stopped\n",
3073 pkt_dev->odevname);
3074 return -EINVAL;
3077 kfree_skb(pkt_dev->skb);
3078 pkt_dev->skb = NULL;
3079 pkt_dev->stopped_at = ktime_now();
3080 pkt_dev->running = 0;
3082 show_results(pkt_dev, nr_frags);
3084 return 0;
3087 static struct pktgen_dev *next_to_run(struct pktgen_thread *t)
3089 struct pktgen_dev *pkt_dev, *best = NULL;
3091 if_lock(t);
3093 list_for_each_entry(pkt_dev, &t->if_list, list) {
3094 if (!pkt_dev->running)
3095 continue;
3096 if (best == NULL)
3097 best = pkt_dev;
3098 else if (ktime_lt(pkt_dev->next_tx, best->next_tx))
3099 best = pkt_dev;
3101 if_unlock(t);
3102 return best;
3105 static void pktgen_stop(struct pktgen_thread *t)
3107 struct pktgen_dev *pkt_dev;
3109 func_enter();
3111 if_lock(t);
3113 list_for_each_entry(pkt_dev, &t->if_list, list) {
3114 pktgen_stop_device(pkt_dev);
3117 if_unlock(t);
3121 * one of our devices needs to be removed - find it
3122 * and remove it
3124 static void pktgen_rem_one_if(struct pktgen_thread *t)
3126 struct list_head *q, *n;
3127 struct pktgen_dev *cur;
3129 func_enter();
3131 if_lock(t);
3133 list_for_each_safe(q, n, &t->if_list) {
3134 cur = list_entry(q, struct pktgen_dev, list);
3136 if (!cur->removal_mark)
3137 continue;
3139 kfree_skb(cur->skb);
3140 cur->skb = NULL;
3142 pktgen_remove_device(t, cur);
3144 break;
3147 if_unlock(t);
3150 static void pktgen_rem_all_ifs(struct pktgen_thread *t)
3152 struct list_head *q, *n;
3153 struct pktgen_dev *cur;
3155 func_enter();
3157 /* Remove all devices, free mem */
3159 if_lock(t);
3161 list_for_each_safe(q, n, &t->if_list) {
3162 cur = list_entry(q, struct pktgen_dev, list);
3164 kfree_skb(cur->skb);
3165 cur->skb = NULL;
3167 pktgen_remove_device(t, cur);
3170 if_unlock(t);
3173 static void pktgen_rem_thread(struct pktgen_thread *t)
3175 /* Remove from the thread list */
3177 remove_proc_entry(t->tsk->comm, pg_proc_dir);
3181 static void pktgen_resched(struct pktgen_dev *pkt_dev)
3183 ktime_t idle_start = ktime_now();
3184 schedule();
3185 pkt_dev->idle_acc += ktime_to_ns(ktime_sub(ktime_now(), idle_start));
3188 static void pktgen_wait_for_skb(struct pktgen_dev *pkt_dev)
3190 ktime_t idle_start = ktime_now();
3192 while (atomic_read(&(pkt_dev->skb->users)) != 1) {
3193 if (signal_pending(current))
3194 break;
3196 if (need_resched())
3197 pktgen_resched(pkt_dev);
3198 else
3199 cpu_relax();
3201 pkt_dev->idle_acc += ktime_to_ns(ktime_sub(ktime_now(), idle_start));
3204 static void pktgen_xmit(struct pktgen_dev *pkt_dev)
3206 struct net_device *odev = pkt_dev->odev;
3207 netdev_tx_t (*xmit)(struct sk_buff *, struct net_device *)
3208 = odev->netdev_ops->ndo_start_xmit;
3209 struct netdev_queue *txq;
3210 u16 queue_map;
3211 int ret;
3213 /* If device is offline, then don't send */
3214 if (unlikely(!netif_running(odev) || !netif_carrier_ok(odev))) {
3215 pktgen_stop_device(pkt_dev);
3216 return;
3219 /* This is max DELAY, this has special meaning of
3220 * "never transmit"
3222 if (unlikely(pkt_dev->delay == ULLONG_MAX)) {
3223 pkt_dev->next_tx = ktime_add_ns(ktime_now(), ULONG_MAX);
3224 return;
3227 /* If no skb or clone count exhausted then get new one */
3228 if (!pkt_dev->skb || (pkt_dev->last_ok &&
3229 ++pkt_dev->clone_count >= pkt_dev->clone_skb)) {
3230 /* build a new pkt */
3231 kfree_skb(pkt_dev->skb);
3233 pkt_dev->skb = fill_packet(odev, pkt_dev);
3234 if (pkt_dev->skb == NULL) {
3235 pr_err("ERROR: couldn't allocate skb in fill_packet\n");
3236 schedule();
3237 pkt_dev->clone_count--; /* back out increment, OOM */
3238 return;
3240 pkt_dev->last_pkt_size = pkt_dev->skb->len;
3241 pkt_dev->allocated_skbs++;
3242 pkt_dev->clone_count = 0; /* reset counter */
3245 if (pkt_dev->delay && pkt_dev->last_ok)
3246 spin(pkt_dev, pkt_dev->next_tx);
3248 queue_map = skb_get_queue_mapping(pkt_dev->skb);
3249 txq = netdev_get_tx_queue(odev, queue_map);
3251 __netif_tx_lock_bh(txq);
3253 if (unlikely(netif_xmit_frozen_or_stopped(txq))) {
3254 ret = NETDEV_TX_BUSY;
3255 pkt_dev->last_ok = 0;
3256 goto unlock;
3258 atomic_inc(&(pkt_dev->skb->users));
3259 ret = (*xmit)(pkt_dev->skb, odev);
3261 switch (ret) {
3262 case NETDEV_TX_OK:
3263 txq_trans_update(txq);
3264 pkt_dev->last_ok = 1;
3265 pkt_dev->sofar++;
3266 pkt_dev->seq_num++;
3267 pkt_dev->tx_bytes += pkt_dev->last_pkt_size;
3268 break;
3269 case NET_XMIT_DROP:
3270 case NET_XMIT_CN:
3271 case NET_XMIT_POLICED:
3272 /* skb has been consumed */
3273 pkt_dev->errors++;
3274 break;
3275 default: /* Drivers are not supposed to return other values! */
3276 net_info_ratelimited("%s xmit error: %d\n",
3277 pkt_dev->odevname, ret);
3278 pkt_dev->errors++;
3279 /* fallthru */
3280 case NETDEV_TX_LOCKED:
3281 case NETDEV_TX_BUSY:
3282 /* Retry it next time */
3283 atomic_dec(&(pkt_dev->skb->users));
3284 pkt_dev->last_ok = 0;
3286 unlock:
3287 __netif_tx_unlock_bh(txq);
3289 /* If pkt_dev->count is zero, then run forever */
3290 if ((pkt_dev->count != 0) && (pkt_dev->sofar >= pkt_dev->count)) {
3291 pktgen_wait_for_skb(pkt_dev);
3293 /* Done with this */
3294 pktgen_stop_device(pkt_dev);
3299 * Main loop of the thread goes here
3302 static int pktgen_thread_worker(void *arg)
3304 DEFINE_WAIT(wait);
3305 struct pktgen_thread *t = arg;
3306 struct pktgen_dev *pkt_dev = NULL;
3307 int cpu = t->cpu;
3309 BUG_ON(smp_processor_id() != cpu);
3311 init_waitqueue_head(&t->queue);
3312 complete(&t->start_done);
3314 pr_debug("starting pktgen/%d: pid=%d\n", cpu, task_pid_nr(current));
3316 set_current_state(TASK_INTERRUPTIBLE);
3318 set_freezable();
3320 while (!kthread_should_stop()) {
3321 pkt_dev = next_to_run(t);
3323 if (unlikely(!pkt_dev && t->control == 0)) {
3324 if (pktgen_exiting)
3325 break;
3326 wait_event_interruptible_timeout(t->queue,
3327 t->control != 0,
3328 HZ/10);
3329 try_to_freeze();
3330 continue;
3333 __set_current_state(TASK_RUNNING);
3335 if (likely(pkt_dev)) {
3336 pktgen_xmit(pkt_dev);
3338 if (need_resched())
3339 pktgen_resched(pkt_dev);
3340 else
3341 cpu_relax();
3344 if (t->control & T_STOP) {
3345 pktgen_stop(t);
3346 t->control &= ~(T_STOP);
3349 if (t->control & T_RUN) {
3350 pktgen_run(t);
3351 t->control &= ~(T_RUN);
3354 if (t->control & T_REMDEVALL) {
3355 pktgen_rem_all_ifs(t);
3356 t->control &= ~(T_REMDEVALL);
3359 if (t->control & T_REMDEV) {
3360 pktgen_rem_one_if(t);
3361 t->control &= ~(T_REMDEV);
3364 try_to_freeze();
3366 set_current_state(TASK_INTERRUPTIBLE);
3369 pr_debug("%s stopping all device\n", t->tsk->comm);
3370 pktgen_stop(t);
3372 pr_debug("%s removing all device\n", t->tsk->comm);
3373 pktgen_rem_all_ifs(t);
3375 pr_debug("%s removing thread\n", t->tsk->comm);
3376 pktgen_rem_thread(t);
3378 /* Wait for kthread_stop */
3379 while (!kthread_should_stop()) {
3380 set_current_state(TASK_INTERRUPTIBLE);
3381 schedule();
3383 __set_current_state(TASK_RUNNING);
3385 return 0;
3388 static struct pktgen_dev *pktgen_find_dev(struct pktgen_thread *t,
3389 const char *ifname, bool exact)
3391 struct pktgen_dev *p, *pkt_dev = NULL;
3392 size_t len = strlen(ifname);
3394 if_lock(t);
3395 list_for_each_entry(p, &t->if_list, list)
3396 if (strncmp(p->odevname, ifname, len) == 0) {
3397 if (p->odevname[len]) {
3398 if (exact || p->odevname[len] != '@')
3399 continue;
3401 pkt_dev = p;
3402 break;
3405 if_unlock(t);
3406 pr_debug("find_dev(%s) returning %p\n", ifname, pkt_dev);
3407 return pkt_dev;
3411 * Adds a dev at front of if_list.
3414 static int add_dev_to_thread(struct pktgen_thread *t,
3415 struct pktgen_dev *pkt_dev)
3417 int rv = 0;
3419 if_lock(t);
3421 if (pkt_dev->pg_thread) {
3422 pr_err("ERROR: already assigned to a thread\n");
3423 rv = -EBUSY;
3424 goto out;
3427 list_add(&pkt_dev->list, &t->if_list);
3428 pkt_dev->pg_thread = t;
3429 pkt_dev->running = 0;
3431 out:
3432 if_unlock(t);
3433 return rv;
3436 /* Called under thread lock */
3438 static int pktgen_add_device(struct pktgen_thread *t, const char *ifname)
3440 struct pktgen_dev *pkt_dev;
3441 int err;
3442 int node = cpu_to_node(t->cpu);
3444 /* We don't allow a device to be on several threads */
3446 pkt_dev = __pktgen_NN_threads(ifname, FIND);
3447 if (pkt_dev) {
3448 pr_err("ERROR: interface already used\n");
3449 return -EBUSY;
3452 pkt_dev = kzalloc_node(sizeof(struct pktgen_dev), GFP_KERNEL, node);
3453 if (!pkt_dev)
3454 return -ENOMEM;
3456 strcpy(pkt_dev->odevname, ifname);
3457 pkt_dev->flows = vzalloc_node(MAX_CFLOWS * sizeof(struct flow_state),
3458 node);
3459 if (pkt_dev->flows == NULL) {
3460 kfree(pkt_dev);
3461 return -ENOMEM;
3464 pkt_dev->removal_mark = 0;
3465 pkt_dev->nfrags = 0;
3466 pkt_dev->delay = pg_delay_d;
3467 pkt_dev->count = pg_count_d;
3468 pkt_dev->sofar = 0;
3469 pkt_dev->udp_src_min = 9; /* sink port */
3470 pkt_dev->udp_src_max = 9;
3471 pkt_dev->udp_dst_min = 9;
3472 pkt_dev->udp_dst_max = 9;
3473 pkt_dev->vlan_p = 0;
3474 pkt_dev->vlan_cfi = 0;
3475 pkt_dev->vlan_id = 0xffff;
3476 pkt_dev->svlan_p = 0;
3477 pkt_dev->svlan_cfi = 0;
3478 pkt_dev->svlan_id = 0xffff;
3479 pkt_dev->node = -1;
3481 err = pktgen_setup_dev(pkt_dev, ifname);
3482 if (err)
3483 goto out1;
3484 if (pkt_dev->odev->priv_flags & IFF_TX_SKB_SHARING)
3485 pkt_dev->clone_skb = pg_clone_skb_d;
3487 pkt_dev->entry = proc_create_data(ifname, 0600, pg_proc_dir,
3488 &pktgen_if_fops, pkt_dev);
3489 if (!pkt_dev->entry) {
3490 pr_err("cannot create %s/%s procfs entry\n",
3491 PG_PROC_DIR, ifname);
3492 err = -EINVAL;
3493 goto out2;
3495 #ifdef CONFIG_XFRM
3496 pkt_dev->ipsmode = XFRM_MODE_TRANSPORT;
3497 pkt_dev->ipsproto = IPPROTO_ESP;
3498 #endif
3500 return add_dev_to_thread(t, pkt_dev);
3501 out2:
3502 dev_put(pkt_dev->odev);
3503 out1:
3504 #ifdef CONFIG_XFRM
3505 free_SAs(pkt_dev);
3506 #endif
3507 vfree(pkt_dev->flows);
3508 kfree(pkt_dev);
3509 return err;
3512 static int __init pktgen_create_thread(int cpu)
3514 struct pktgen_thread *t;
3515 struct proc_dir_entry *pe;
3516 struct task_struct *p;
3518 t = kzalloc_node(sizeof(struct pktgen_thread), GFP_KERNEL,
3519 cpu_to_node(cpu));
3520 if (!t) {
3521 pr_err("ERROR: out of memory, can't create new thread\n");
3522 return -ENOMEM;
3525 spin_lock_init(&t->if_lock);
3526 t->cpu = cpu;
3528 INIT_LIST_HEAD(&t->if_list);
3530 list_add_tail(&t->th_list, &pktgen_threads);
3531 init_completion(&t->start_done);
3533 p = kthread_create_on_node(pktgen_thread_worker,
3535 cpu_to_node(cpu),
3536 "kpktgend_%d", cpu);
3537 if (IS_ERR(p)) {
3538 pr_err("kernel_thread() failed for cpu %d\n", t->cpu);
3539 list_del(&t->th_list);
3540 kfree(t);
3541 return PTR_ERR(p);
3543 kthread_bind(p, cpu);
3544 t->tsk = p;
3546 pe = proc_create_data(t->tsk->comm, 0600, pg_proc_dir,
3547 &pktgen_thread_fops, t);
3548 if (!pe) {
3549 pr_err("cannot create %s/%s procfs entry\n",
3550 PG_PROC_DIR, t->tsk->comm);
3551 kthread_stop(p);
3552 list_del(&t->th_list);
3553 kfree(t);
3554 return -EINVAL;
3557 wake_up_process(p);
3558 wait_for_completion(&t->start_done);
3560 return 0;
3564 * Removes a device from the thread if_list.
3566 static void _rem_dev_from_if_list(struct pktgen_thread *t,
3567 struct pktgen_dev *pkt_dev)
3569 struct list_head *q, *n;
3570 struct pktgen_dev *p;
3572 list_for_each_safe(q, n, &t->if_list) {
3573 p = list_entry(q, struct pktgen_dev, list);
3574 if (p == pkt_dev)
3575 list_del(&p->list);
3579 static int pktgen_remove_device(struct pktgen_thread *t,
3580 struct pktgen_dev *pkt_dev)
3583 pr_debug("remove_device pkt_dev=%p\n", pkt_dev);
3585 if (pkt_dev->running) {
3586 pr_warning("WARNING: trying to remove a running interface, stopping it now\n");
3587 pktgen_stop_device(pkt_dev);
3590 /* Dis-associate from the interface */
3592 if (pkt_dev->odev) {
3593 dev_put(pkt_dev->odev);
3594 pkt_dev->odev = NULL;
3597 /* And update the thread if_list */
3599 _rem_dev_from_if_list(t, pkt_dev);
3601 if (pkt_dev->entry)
3602 remove_proc_entry(pkt_dev->entry->name, pg_proc_dir);
3604 #ifdef CONFIG_XFRM
3605 free_SAs(pkt_dev);
3606 #endif
3607 vfree(pkt_dev->flows);
3608 if (pkt_dev->page)
3609 put_page(pkt_dev->page);
3610 kfree(pkt_dev);
3611 return 0;
3614 static int __init pg_init(void)
3616 int cpu;
3617 struct proc_dir_entry *pe;
3618 int ret = 0;
3620 pr_info("%s", version);
3622 pg_proc_dir = proc_mkdir(PG_PROC_DIR, init_net.proc_net);
3623 if (!pg_proc_dir)
3624 return -ENODEV;
3626 pe = proc_create(PGCTRL, 0600, pg_proc_dir, &pktgen_fops);
3627 if (pe == NULL) {
3628 pr_err("ERROR: cannot create %s procfs entry\n", PGCTRL);
3629 ret = -EINVAL;
3630 goto remove_dir;
3633 register_netdevice_notifier(&pktgen_notifier_block);
3635 for_each_online_cpu(cpu) {
3636 int err;
3638 err = pktgen_create_thread(cpu);
3639 if (err)
3640 pr_warning("WARNING: Cannot create thread for cpu %d (%d)\n",
3641 cpu, err);
3644 if (list_empty(&pktgen_threads)) {
3645 pr_err("ERROR: Initialization failed for all threads\n");
3646 ret = -ENODEV;
3647 goto unregister;
3650 return 0;
3652 unregister:
3653 unregister_netdevice_notifier(&pktgen_notifier_block);
3654 remove_proc_entry(PGCTRL, pg_proc_dir);
3655 remove_dir:
3656 proc_net_remove(&init_net, PG_PROC_DIR);
3657 return ret;
3660 static void __exit pg_cleanup(void)
3662 struct pktgen_thread *t;
3663 struct list_head *q, *n;
3664 LIST_HEAD(list);
3666 /* Stop all interfaces & threads */
3667 pktgen_exiting = true;
3669 mutex_lock(&pktgen_thread_lock);
3670 list_splice_init(&pktgen_threads, &list);
3671 mutex_unlock(&pktgen_thread_lock);
3673 list_for_each_safe(q, n, &list) {
3674 t = list_entry(q, struct pktgen_thread, th_list);
3675 list_del(&t->th_list);
3676 kthread_stop(t->tsk);
3677 kfree(t);
3680 /* Un-register us from receiving netdevice events */
3681 unregister_netdevice_notifier(&pktgen_notifier_block);
3683 /* Clean up proc file system */
3684 remove_proc_entry(PGCTRL, pg_proc_dir);
3685 proc_net_remove(&init_net, PG_PROC_DIR);
3688 module_init(pg_init);
3689 module_exit(pg_cleanup);
3691 MODULE_AUTHOR("Robert Olsson <robert.olsson@its.uu.se>");
3692 MODULE_DESCRIPTION("Packet Generator tool");
3693 MODULE_LICENSE("GPL");
3694 MODULE_VERSION(VERSION);
3695 module_param(pg_count_d, int, 0);
3696 MODULE_PARM_DESC(pg_count_d, "Default number of packets to inject");
3697 module_param(pg_delay_d, int, 0);
3698 MODULE_PARM_DESC(pg_delay_d, "Default delay between packets (nanoseconds)");
3699 module_param(pg_clone_skb_d, int, 0);
3700 MODULE_PARM_DESC(pg_clone_skb_d, "Default number of copies of the same packet");
3701 module_param(debug, int, 0);
3702 MODULE_PARM_DESC(debug, "Enable debugging of pktgen module");