Merge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
[linux/fpc-iii.git] / net / core / pktgen.c
blobfdac61cac1bd11b15a69f78584029731e9310e4e
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/udp.h>
164 #include <net/ip6_checksum.h>
165 #include <net/addrconf.h>
166 #ifdef CONFIG_XFRM
167 #include <net/xfrm.h>
168 #endif
169 #include <net/netns/generic.h>
170 #include <asm/byteorder.h>
171 #include <linux/rcupdate.h>
172 #include <linux/bitops.h>
173 #include <linux/io.h>
174 #include <linux/timex.h>
175 #include <linux/uaccess.h>
176 #include <asm/dma.h>
177 #include <asm/div64.h> /* do_div */
179 #define VERSION "2.74"
180 #define IP_NAME_SZ 32
181 #define MAX_MPLS_LABELS 16 /* This is the max label stack depth */
182 #define MPLS_STACK_BOTTOM htonl(0x00000100)
184 #define func_enter() pr_debug("entering %s\n", __func__);
186 /* Device flag bits */
187 #define F_IPSRC_RND (1<<0) /* IP-Src Random */
188 #define F_IPDST_RND (1<<1) /* IP-Dst Random */
189 #define F_UDPSRC_RND (1<<2) /* UDP-Src Random */
190 #define F_UDPDST_RND (1<<3) /* UDP-Dst Random */
191 #define F_MACSRC_RND (1<<4) /* MAC-Src Random */
192 #define F_MACDST_RND (1<<5) /* MAC-Dst Random */
193 #define F_TXSIZE_RND (1<<6) /* Transmit size is random */
194 #define F_IPV6 (1<<7) /* Interface in IPV6 Mode */
195 #define F_MPLS_RND (1<<8) /* Random MPLS labels */
196 #define F_VID_RND (1<<9) /* Random VLAN ID */
197 #define F_SVID_RND (1<<10) /* Random SVLAN ID */
198 #define F_FLOW_SEQ (1<<11) /* Sequential flows */
199 #define F_IPSEC_ON (1<<12) /* ipsec on for flows */
200 #define F_QUEUE_MAP_RND (1<<13) /* queue map Random */
201 #define F_QUEUE_MAP_CPU (1<<14) /* queue map mirrors smp_processor_id() */
202 #define F_NODE (1<<15) /* Node memory alloc*/
203 #define F_UDPCSUM (1<<16) /* Include UDP checksum */
205 /* Thread control flag bits */
206 #define T_STOP (1<<0) /* Stop run */
207 #define T_RUN (1<<1) /* Start run */
208 #define T_REMDEVALL (1<<2) /* Remove all devs */
209 #define T_REMDEV (1<<3) /* Remove one dev */
211 /* If lock -- can be removed after some work */
212 #define if_lock(t) spin_lock(&(t->if_lock));
213 #define if_unlock(t) spin_unlock(&(t->if_lock));
215 /* Used to help with determining the pkts on receive */
216 #define PKTGEN_MAGIC 0xbe9be955
217 #define PG_PROC_DIR "pktgen"
218 #define PGCTRL "pgctrl"
220 #define MAX_CFLOWS 65536
222 #define VLAN_TAG_SIZE(x) ((x)->vlan_id == 0xffff ? 0 : 4)
223 #define SVLAN_TAG_SIZE(x) ((x)->svlan_id == 0xffff ? 0 : 4)
225 struct flow_state {
226 __be32 cur_daddr;
227 int count;
228 #ifdef CONFIG_XFRM
229 struct xfrm_state *x;
230 #endif
231 __u32 flags;
234 /* flow flag bits */
235 #define F_INIT (1<<0) /* flow has been initialized */
237 struct pktgen_dev {
239 * Try to keep frequent/infrequent used vars. separated.
241 struct proc_dir_entry *entry; /* proc file */
242 struct pktgen_thread *pg_thread;/* the owner */
243 struct list_head list; /* chaining in the thread's run-queue */
245 int running; /* if false, the test will stop */
247 /* If min != max, then we will either do a linear iteration, or
248 * we will do a random selection from within the range.
250 __u32 flags;
251 int removal_mark; /* non-zero => the device is marked for
252 * removal by worker thread */
254 int min_pkt_size;
255 int max_pkt_size;
256 int pkt_overhead; /* overhead for MPLS, VLANs, IPSEC etc */
257 int nfrags;
258 struct page *page;
259 u64 delay; /* nano-seconds */
261 __u64 count; /* Default No packets to send */
262 __u64 sofar; /* How many pkts we've sent so far */
263 __u64 tx_bytes; /* How many bytes we've transmitted */
264 __u64 errors; /* Errors when trying to transmit, */
266 /* runtime counters relating to clone_skb */
268 __u64 allocated_skbs;
269 __u32 clone_count;
270 int last_ok; /* Was last skb sent?
271 * Or a failed transmit of some sort?
272 * This will keep sequence numbers in order
274 ktime_t next_tx;
275 ktime_t started_at;
276 ktime_t stopped_at;
277 u64 idle_acc; /* nano-seconds */
279 __u32 seq_num;
281 int clone_skb; /*
282 * Use multiple SKBs during packet gen.
283 * If this number is greater than 1, then
284 * that many copies of the same packet will be
285 * sent before a new packet is allocated.
286 * If you want to send 1024 identical packets
287 * before creating a new packet,
288 * set clone_skb to 1024.
291 char dst_min[IP_NAME_SZ]; /* IP, ie 1.2.3.4 */
292 char dst_max[IP_NAME_SZ]; /* IP, ie 1.2.3.4 */
293 char src_min[IP_NAME_SZ]; /* IP, ie 1.2.3.4 */
294 char src_max[IP_NAME_SZ]; /* IP, ie 1.2.3.4 */
296 struct in6_addr in6_saddr;
297 struct in6_addr in6_daddr;
298 struct in6_addr cur_in6_daddr;
299 struct in6_addr cur_in6_saddr;
300 /* For ranges */
301 struct in6_addr min_in6_daddr;
302 struct in6_addr max_in6_daddr;
303 struct in6_addr min_in6_saddr;
304 struct in6_addr max_in6_saddr;
306 /* If we're doing ranges, random or incremental, then this
307 * defines the min/max for those ranges.
309 __be32 saddr_min; /* inclusive, source IP address */
310 __be32 saddr_max; /* exclusive, source IP address */
311 __be32 daddr_min; /* inclusive, dest IP address */
312 __be32 daddr_max; /* exclusive, dest IP address */
314 __u16 udp_src_min; /* inclusive, source UDP port */
315 __u16 udp_src_max; /* exclusive, source UDP port */
316 __u16 udp_dst_min; /* inclusive, dest UDP port */
317 __u16 udp_dst_max; /* exclusive, dest UDP port */
319 /* DSCP + ECN */
320 __u8 tos; /* six MSB of (former) IPv4 TOS
321 are for dscp codepoint */
322 __u8 traffic_class; /* ditto for the (former) Traffic Class in IPv6
323 (see RFC 3260, sec. 4) */
325 /* MPLS */
326 unsigned int nr_labels; /* Depth of stack, 0 = no MPLS */
327 __be32 labels[MAX_MPLS_LABELS];
329 /* VLAN/SVLAN (802.1Q/Q-in-Q) */
330 __u8 vlan_p;
331 __u8 vlan_cfi;
332 __u16 vlan_id; /* 0xffff means no vlan tag */
334 __u8 svlan_p;
335 __u8 svlan_cfi;
336 __u16 svlan_id; /* 0xffff means no svlan tag */
338 __u32 src_mac_count; /* How many MACs to iterate through */
339 __u32 dst_mac_count; /* How many MACs to iterate through */
341 unsigned char dst_mac[ETH_ALEN];
342 unsigned char src_mac[ETH_ALEN];
344 __u32 cur_dst_mac_offset;
345 __u32 cur_src_mac_offset;
346 __be32 cur_saddr;
347 __be32 cur_daddr;
348 __u16 ip_id;
349 __u16 cur_udp_dst;
350 __u16 cur_udp_src;
351 __u16 cur_queue_map;
352 __u32 cur_pkt_size;
353 __u32 last_pkt_size;
355 __u8 hh[14];
356 /* = {
357 0x00, 0x80, 0xC8, 0x79, 0xB3, 0xCB,
359 We fill in SRC address later
360 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
361 0x08, 0x00
364 __u16 pad; /* pad out the hh struct to an even 16 bytes */
366 struct sk_buff *skb; /* skb we are to transmit next, used for when we
367 * are transmitting the same one multiple times
369 struct net_device *odev; /* The out-going device.
370 * Note that the device should have it's
371 * pg_info pointer pointing back to this
372 * device.
373 * Set when the user specifies the out-going
374 * device name (not when the inject is
375 * started as it used to do.)
377 char odevname[32];
378 struct flow_state *flows;
379 unsigned int cflows; /* Concurrent flows (config) */
380 unsigned int lflow; /* Flow length (config) */
381 unsigned int nflows; /* accumulated flows (stats) */
382 unsigned int curfl; /* current sequenced flow (state)*/
384 u16 queue_map_min;
385 u16 queue_map_max;
386 __u32 skb_priority; /* skb priority field */
387 int node; /* Memory node */
389 #ifdef CONFIG_XFRM
390 __u8 ipsmode; /* IPSEC mode (config) */
391 __u8 ipsproto; /* IPSEC type (config) */
392 __u32 spi;
393 struct dst_entry dst;
394 struct dst_ops dstops;
395 #endif
396 char result[512];
399 struct pktgen_hdr {
400 __be32 pgh_magic;
401 __be32 seq_num;
402 __be32 tv_sec;
403 __be32 tv_usec;
407 static int pg_net_id __read_mostly;
409 struct pktgen_net {
410 struct net *net;
411 struct proc_dir_entry *proc_dir;
412 struct list_head pktgen_threads;
413 bool pktgen_exiting;
416 struct pktgen_thread {
417 spinlock_t if_lock; /* for list of devices */
418 struct list_head if_list; /* All device here */
419 struct list_head th_list;
420 struct task_struct *tsk;
421 char result[512];
423 /* Field for thread to receive "posted" events terminate,
424 stop ifs etc. */
426 u32 control;
427 int cpu;
429 wait_queue_head_t queue;
430 struct completion start_done;
431 struct pktgen_net *net;
434 #define REMOVE 1
435 #define FIND 0
437 static const char version[] =
438 "Packet Generator for packet performance testing. "
439 "Version: " VERSION "\n";
441 static int pktgen_remove_device(struct pktgen_thread *t, struct pktgen_dev *i);
442 static int pktgen_add_device(struct pktgen_thread *t, const char *ifname);
443 static struct pktgen_dev *pktgen_find_dev(struct pktgen_thread *t,
444 const char *ifname, bool exact);
445 static int pktgen_device_event(struct notifier_block *, unsigned long, void *);
446 static void pktgen_run_all_threads(struct pktgen_net *pn);
447 static void pktgen_reset_all_threads(struct pktgen_net *pn);
448 static void pktgen_stop_all_threads_ifs(struct pktgen_net *pn);
450 static void pktgen_stop(struct pktgen_thread *t);
451 static void pktgen_clear_counters(struct pktgen_dev *pkt_dev);
453 /* Module parameters, defaults. */
454 static int pg_count_d __read_mostly = 1000;
455 static int pg_delay_d __read_mostly;
456 static int pg_clone_skb_d __read_mostly;
457 static int debug __read_mostly;
459 static DEFINE_MUTEX(pktgen_thread_lock);
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];
481 struct pktgen_net *pn = net_generic(current->nsproxy->net_ns, pg_net_id);
483 if (!capable(CAP_NET_ADMIN)) {
484 err = -EPERM;
485 goto out;
488 if (count > sizeof(data))
489 count = sizeof(data);
491 if (copy_from_user(data, buf, count)) {
492 err = -EFAULT;
493 goto out;
495 data[count - 1] = 0; /* Make string */
497 if (!strcmp(data, "stop"))
498 pktgen_stop_all_threads_ifs(pn);
500 else if (!strcmp(data, "start"))
501 pktgen_run_all_threads(pn);
503 else if (!strcmp(data, "reset"))
504 pktgen_reset_all_threads(pn);
506 else
507 pr_warning("Unknown command: %s\n", data);
509 err = count;
511 out:
512 return err;
515 static int pgctrl_open(struct inode *inode, struct file *file)
517 return single_open(file, pgctrl_show, PDE_DATA(inode));
520 static const struct file_operations pktgen_fops = {
521 .owner = THIS_MODULE,
522 .open = pgctrl_open,
523 .read = seq_read,
524 .llseek = seq_lseek,
525 .write = pgctrl_write,
526 .release = single_release,
529 static int pktgen_if_show(struct seq_file *seq, void *v)
531 const struct pktgen_dev *pkt_dev = seq->private;
532 ktime_t stopped;
533 u64 idle;
535 seq_printf(seq,
536 "Params: count %llu min_pkt_size: %u max_pkt_size: %u\n",
537 (unsigned long long)pkt_dev->count, pkt_dev->min_pkt_size,
538 pkt_dev->max_pkt_size);
540 seq_printf(seq,
541 " frags: %d delay: %llu clone_skb: %d ifname: %s\n",
542 pkt_dev->nfrags, (unsigned long long) pkt_dev->delay,
543 pkt_dev->clone_skb, pkt_dev->odevname);
545 seq_printf(seq, " flows: %u flowlen: %u\n", pkt_dev->cflows,
546 pkt_dev->lflow);
548 seq_printf(seq,
549 " queue_map_min: %u queue_map_max: %u\n",
550 pkt_dev->queue_map_min,
551 pkt_dev->queue_map_max);
553 if (pkt_dev->skb_priority)
554 seq_printf(seq, " skb_priority: %u\n",
555 pkt_dev->skb_priority);
557 if (pkt_dev->flags & F_IPV6) {
558 seq_printf(seq,
559 " saddr: %pI6c min_saddr: %pI6c max_saddr: %pI6c\n"
560 " daddr: %pI6c min_daddr: %pI6c max_daddr: %pI6c\n",
561 &pkt_dev->in6_saddr,
562 &pkt_dev->min_in6_saddr, &pkt_dev->max_in6_saddr,
563 &pkt_dev->in6_daddr,
564 &pkt_dev->min_in6_daddr, &pkt_dev->max_in6_daddr);
565 } else {
566 seq_printf(seq,
567 " dst_min: %s dst_max: %s\n",
568 pkt_dev->dst_min, pkt_dev->dst_max);
569 seq_printf(seq,
570 " src_min: %s src_max: %s\n",
571 pkt_dev->src_min, pkt_dev->src_max);
574 seq_puts(seq, " src_mac: ");
576 seq_printf(seq, "%pM ",
577 is_zero_ether_addr(pkt_dev->src_mac) ?
578 pkt_dev->odev->dev_addr : pkt_dev->src_mac);
580 seq_printf(seq, "dst_mac: ");
581 seq_printf(seq, "%pM\n", pkt_dev->dst_mac);
583 seq_printf(seq,
584 " udp_src_min: %d udp_src_max: %d"
585 " udp_dst_min: %d udp_dst_max: %d\n",
586 pkt_dev->udp_src_min, pkt_dev->udp_src_max,
587 pkt_dev->udp_dst_min, pkt_dev->udp_dst_max);
589 seq_printf(seq,
590 " src_mac_count: %d dst_mac_count: %d\n",
591 pkt_dev->src_mac_count, pkt_dev->dst_mac_count);
593 if (pkt_dev->nr_labels) {
594 unsigned int i;
595 seq_printf(seq, " mpls: ");
596 for (i = 0; i < pkt_dev->nr_labels; i++)
597 seq_printf(seq, "%08x%s", ntohl(pkt_dev->labels[i]),
598 i == pkt_dev->nr_labels-1 ? "\n" : ", ");
601 if (pkt_dev->vlan_id != 0xffff)
602 seq_printf(seq, " vlan_id: %u vlan_p: %u vlan_cfi: %u\n",
603 pkt_dev->vlan_id, pkt_dev->vlan_p,
604 pkt_dev->vlan_cfi);
606 if (pkt_dev->svlan_id != 0xffff)
607 seq_printf(seq, " svlan_id: %u vlan_p: %u vlan_cfi: %u\n",
608 pkt_dev->svlan_id, pkt_dev->svlan_p,
609 pkt_dev->svlan_cfi);
611 if (pkt_dev->tos)
612 seq_printf(seq, " tos: 0x%02x\n", pkt_dev->tos);
614 if (pkt_dev->traffic_class)
615 seq_printf(seq, " traffic_class: 0x%02x\n", pkt_dev->traffic_class);
617 if (pkt_dev->node >= 0)
618 seq_printf(seq, " node: %d\n", pkt_dev->node);
620 seq_printf(seq, " Flags: ");
622 if (pkt_dev->flags & F_IPV6)
623 seq_printf(seq, "IPV6 ");
625 if (pkt_dev->flags & F_IPSRC_RND)
626 seq_printf(seq, "IPSRC_RND ");
628 if (pkt_dev->flags & F_IPDST_RND)
629 seq_printf(seq, "IPDST_RND ");
631 if (pkt_dev->flags & F_TXSIZE_RND)
632 seq_printf(seq, "TXSIZE_RND ");
634 if (pkt_dev->flags & F_UDPSRC_RND)
635 seq_printf(seq, "UDPSRC_RND ");
637 if (pkt_dev->flags & F_UDPDST_RND)
638 seq_printf(seq, "UDPDST_RND ");
640 if (pkt_dev->flags & F_UDPCSUM)
641 seq_printf(seq, "UDPCSUM ");
643 if (pkt_dev->flags & F_MPLS_RND)
644 seq_printf(seq, "MPLS_RND ");
646 if (pkt_dev->flags & F_QUEUE_MAP_RND)
647 seq_printf(seq, "QUEUE_MAP_RND ");
649 if (pkt_dev->flags & F_QUEUE_MAP_CPU)
650 seq_printf(seq, "QUEUE_MAP_CPU ");
652 if (pkt_dev->cflows) {
653 if (pkt_dev->flags & F_FLOW_SEQ)
654 seq_printf(seq, "FLOW_SEQ "); /*in sequence flows*/
655 else
656 seq_printf(seq, "FLOW_RND ");
659 #ifdef CONFIG_XFRM
660 if (pkt_dev->flags & F_IPSEC_ON) {
661 seq_printf(seq, "IPSEC ");
662 if (pkt_dev->spi)
663 seq_printf(seq, "spi:%u", pkt_dev->spi);
665 #endif
667 if (pkt_dev->flags & F_MACSRC_RND)
668 seq_printf(seq, "MACSRC_RND ");
670 if (pkt_dev->flags & F_MACDST_RND)
671 seq_printf(seq, "MACDST_RND ");
673 if (pkt_dev->flags & F_VID_RND)
674 seq_printf(seq, "VID_RND ");
676 if (pkt_dev->flags & F_SVID_RND)
677 seq_printf(seq, "SVID_RND ");
679 if (pkt_dev->flags & F_NODE)
680 seq_printf(seq, "NODE_ALLOC ");
682 seq_puts(seq, "\n");
684 /* not really stopped, more like last-running-at */
685 stopped = pkt_dev->running ? ktime_get() : pkt_dev->stopped_at;
686 idle = pkt_dev->idle_acc;
687 do_div(idle, NSEC_PER_USEC);
689 seq_printf(seq,
690 "Current:\n pkts-sofar: %llu errors: %llu\n",
691 (unsigned long long)pkt_dev->sofar,
692 (unsigned long long)pkt_dev->errors);
694 seq_printf(seq,
695 " started: %lluus stopped: %lluus idle: %lluus\n",
696 (unsigned long long) ktime_to_us(pkt_dev->started_at),
697 (unsigned long long) ktime_to_us(stopped),
698 (unsigned long long) idle);
700 seq_printf(seq,
701 " seq_num: %d cur_dst_mac_offset: %d cur_src_mac_offset: %d\n",
702 pkt_dev->seq_num, pkt_dev->cur_dst_mac_offset,
703 pkt_dev->cur_src_mac_offset);
705 if (pkt_dev->flags & F_IPV6) {
706 seq_printf(seq, " cur_saddr: %pI6c cur_daddr: %pI6c\n",
707 &pkt_dev->cur_in6_saddr,
708 &pkt_dev->cur_in6_daddr);
709 } else
710 seq_printf(seq, " cur_saddr: %pI4 cur_daddr: %pI4\n",
711 &pkt_dev->cur_saddr, &pkt_dev->cur_daddr);
713 seq_printf(seq, " cur_udp_dst: %d cur_udp_src: %d\n",
714 pkt_dev->cur_udp_dst, pkt_dev->cur_udp_src);
716 seq_printf(seq, " cur_queue_map: %u\n", pkt_dev->cur_queue_map);
718 seq_printf(seq, " flows: %u\n", pkt_dev->nflows);
720 if (pkt_dev->result[0])
721 seq_printf(seq, "Result: %s\n", pkt_dev->result);
722 else
723 seq_printf(seq, "Result: Idle\n");
725 return 0;
729 static int hex32_arg(const char __user *user_buffer, unsigned long maxlen,
730 __u32 *num)
732 int i = 0;
733 *num = 0;
735 for (; i < maxlen; i++) {
736 int value;
737 char c;
738 *num <<= 4;
739 if (get_user(c, &user_buffer[i]))
740 return -EFAULT;
741 value = hex_to_bin(c);
742 if (value >= 0)
743 *num |= value;
744 else
745 break;
747 return i;
750 static int count_trail_chars(const char __user * user_buffer,
751 unsigned int maxlen)
753 int i;
755 for (i = 0; i < maxlen; i++) {
756 char c;
757 if (get_user(c, &user_buffer[i]))
758 return -EFAULT;
759 switch (c) {
760 case '\"':
761 case '\n':
762 case '\r':
763 case '\t':
764 case ' ':
765 case '=':
766 break;
767 default:
768 goto done;
771 done:
772 return i;
775 static long num_arg(const char __user *user_buffer, unsigned long maxlen,
776 unsigned long *num)
778 int i;
779 *num = 0;
781 for (i = 0; i < maxlen; i++) {
782 char c;
783 if (get_user(c, &user_buffer[i]))
784 return -EFAULT;
785 if ((c >= '0') && (c <= '9')) {
786 *num *= 10;
787 *num += c - '0';
788 } else
789 break;
791 return i;
794 static int strn_len(const char __user * user_buffer, unsigned int maxlen)
796 int i;
798 for (i = 0; i < maxlen; i++) {
799 char c;
800 if (get_user(c, &user_buffer[i]))
801 return -EFAULT;
802 switch (c) {
803 case '\"':
804 case '\n':
805 case '\r':
806 case '\t':
807 case ' ':
808 goto done_str;
809 break;
810 default:
811 break;
814 done_str:
815 return i;
818 static ssize_t get_labels(const char __user *buffer, struct pktgen_dev *pkt_dev)
820 unsigned int n = 0;
821 char c;
822 ssize_t i = 0;
823 int len;
825 pkt_dev->nr_labels = 0;
826 do {
827 __u32 tmp;
828 len = hex32_arg(&buffer[i], 8, &tmp);
829 if (len <= 0)
830 return len;
831 pkt_dev->labels[n] = htonl(tmp);
832 if (pkt_dev->labels[n] & MPLS_STACK_BOTTOM)
833 pkt_dev->flags |= F_MPLS_RND;
834 i += len;
835 if (get_user(c, &buffer[i]))
836 return -EFAULT;
837 i++;
838 n++;
839 if (n >= MAX_MPLS_LABELS)
840 return -E2BIG;
841 } while (c == ',');
843 pkt_dev->nr_labels = n;
844 return i;
847 static ssize_t pktgen_if_write(struct file *file,
848 const char __user * user_buffer, size_t count,
849 loff_t * offset)
851 struct seq_file *seq = file->private_data;
852 struct pktgen_dev *pkt_dev = seq->private;
853 int i, max, len;
854 char name[16], valstr[32];
855 unsigned long value = 0;
856 char *pg_result = NULL;
857 int tmp = 0;
858 char buf[128];
860 pg_result = &(pkt_dev->result[0]);
862 if (count < 1) {
863 pr_warning("wrong command format\n");
864 return -EINVAL;
867 max = count;
868 tmp = count_trail_chars(user_buffer, max);
869 if (tmp < 0) {
870 pr_warning("illegal format\n");
871 return tmp;
873 i = tmp;
875 /* Read variable name */
877 len = strn_len(&user_buffer[i], sizeof(name) - 1);
878 if (len < 0)
879 return len;
881 memset(name, 0, sizeof(name));
882 if (copy_from_user(name, &user_buffer[i], len))
883 return -EFAULT;
884 i += len;
886 max = count - i;
887 len = count_trail_chars(&user_buffer[i], max);
888 if (len < 0)
889 return len;
891 i += len;
893 if (debug) {
894 size_t copy = min_t(size_t, count, 1023);
895 char tb[copy + 1];
896 if (copy_from_user(tb, user_buffer, copy))
897 return -EFAULT;
898 tb[copy] = 0;
899 pr_debug("%s,%lu buffer -:%s:-\n",
900 name, (unsigned long)count, tb);
903 if (!strcmp(name, "min_pkt_size")) {
904 len = num_arg(&user_buffer[i], 10, &value);
905 if (len < 0)
906 return len;
908 i += len;
909 if (value < 14 + 20 + 8)
910 value = 14 + 20 + 8;
911 if (value != pkt_dev->min_pkt_size) {
912 pkt_dev->min_pkt_size = value;
913 pkt_dev->cur_pkt_size = value;
915 sprintf(pg_result, "OK: min_pkt_size=%u",
916 pkt_dev->min_pkt_size);
917 return count;
920 if (!strcmp(name, "max_pkt_size")) {
921 len = num_arg(&user_buffer[i], 10, &value);
922 if (len < 0)
923 return len;
925 i += len;
926 if (value < 14 + 20 + 8)
927 value = 14 + 20 + 8;
928 if (value != pkt_dev->max_pkt_size) {
929 pkt_dev->max_pkt_size = value;
930 pkt_dev->cur_pkt_size = value;
932 sprintf(pg_result, "OK: max_pkt_size=%u",
933 pkt_dev->max_pkt_size);
934 return count;
937 /* Shortcut for min = max */
939 if (!strcmp(name, "pkt_size")) {
940 len = num_arg(&user_buffer[i], 10, &value);
941 if (len < 0)
942 return len;
944 i += len;
945 if (value < 14 + 20 + 8)
946 value = 14 + 20 + 8;
947 if (value != pkt_dev->min_pkt_size) {
948 pkt_dev->min_pkt_size = value;
949 pkt_dev->max_pkt_size = value;
950 pkt_dev->cur_pkt_size = value;
952 sprintf(pg_result, "OK: pkt_size=%u", pkt_dev->min_pkt_size);
953 return count;
956 if (!strcmp(name, "debug")) {
957 len = num_arg(&user_buffer[i], 10, &value);
958 if (len < 0)
959 return len;
961 i += len;
962 debug = value;
963 sprintf(pg_result, "OK: debug=%u", debug);
964 return count;
967 if (!strcmp(name, "frags")) {
968 len = num_arg(&user_buffer[i], 10, &value);
969 if (len < 0)
970 return len;
972 i += len;
973 pkt_dev->nfrags = value;
974 sprintf(pg_result, "OK: frags=%u", pkt_dev->nfrags);
975 return count;
977 if (!strcmp(name, "delay")) {
978 len = num_arg(&user_buffer[i], 10, &value);
979 if (len < 0)
980 return len;
982 i += len;
983 if (value == 0x7FFFFFFF)
984 pkt_dev->delay = ULLONG_MAX;
985 else
986 pkt_dev->delay = (u64)value;
988 sprintf(pg_result, "OK: delay=%llu",
989 (unsigned long long) pkt_dev->delay);
990 return count;
992 if (!strcmp(name, "rate")) {
993 len = num_arg(&user_buffer[i], 10, &value);
994 if (len < 0)
995 return len;
997 i += len;
998 if (!value)
999 return len;
1000 pkt_dev->delay = pkt_dev->min_pkt_size*8*NSEC_PER_USEC/value;
1001 if (debug)
1002 pr_info("Delay set at: %llu ns\n", pkt_dev->delay);
1004 sprintf(pg_result, "OK: rate=%lu", value);
1005 return count;
1007 if (!strcmp(name, "ratep")) {
1008 len = num_arg(&user_buffer[i], 10, &value);
1009 if (len < 0)
1010 return len;
1012 i += len;
1013 if (!value)
1014 return len;
1015 pkt_dev->delay = NSEC_PER_SEC/value;
1016 if (debug)
1017 pr_info("Delay set at: %llu ns\n", pkt_dev->delay);
1019 sprintf(pg_result, "OK: rate=%lu", value);
1020 return count;
1022 if (!strcmp(name, "udp_src_min")) {
1023 len = num_arg(&user_buffer[i], 10, &value);
1024 if (len < 0)
1025 return len;
1027 i += len;
1028 if (value != pkt_dev->udp_src_min) {
1029 pkt_dev->udp_src_min = value;
1030 pkt_dev->cur_udp_src = value;
1032 sprintf(pg_result, "OK: udp_src_min=%u", pkt_dev->udp_src_min);
1033 return count;
1035 if (!strcmp(name, "udp_dst_min")) {
1036 len = num_arg(&user_buffer[i], 10, &value);
1037 if (len < 0)
1038 return len;
1040 i += len;
1041 if (value != pkt_dev->udp_dst_min) {
1042 pkt_dev->udp_dst_min = value;
1043 pkt_dev->cur_udp_dst = value;
1045 sprintf(pg_result, "OK: udp_dst_min=%u", pkt_dev->udp_dst_min);
1046 return count;
1048 if (!strcmp(name, "udp_src_max")) {
1049 len = num_arg(&user_buffer[i], 10, &value);
1050 if (len < 0)
1051 return len;
1053 i += len;
1054 if (value != pkt_dev->udp_src_max) {
1055 pkt_dev->udp_src_max = value;
1056 pkt_dev->cur_udp_src = value;
1058 sprintf(pg_result, "OK: udp_src_max=%u", pkt_dev->udp_src_max);
1059 return count;
1061 if (!strcmp(name, "udp_dst_max")) {
1062 len = num_arg(&user_buffer[i], 10, &value);
1063 if (len < 0)
1064 return len;
1066 i += len;
1067 if (value != pkt_dev->udp_dst_max) {
1068 pkt_dev->udp_dst_max = value;
1069 pkt_dev->cur_udp_dst = value;
1071 sprintf(pg_result, "OK: udp_dst_max=%u", pkt_dev->udp_dst_max);
1072 return count;
1074 if (!strcmp(name, "clone_skb")) {
1075 len = num_arg(&user_buffer[i], 10, &value);
1076 if (len < 0)
1077 return len;
1078 if ((value > 0) &&
1079 (!(pkt_dev->odev->priv_flags & IFF_TX_SKB_SHARING)))
1080 return -ENOTSUPP;
1081 i += len;
1082 pkt_dev->clone_skb = value;
1084 sprintf(pg_result, "OK: clone_skb=%d", pkt_dev->clone_skb);
1085 return count;
1087 if (!strcmp(name, "count")) {
1088 len = num_arg(&user_buffer[i], 10, &value);
1089 if (len < 0)
1090 return len;
1092 i += len;
1093 pkt_dev->count = value;
1094 sprintf(pg_result, "OK: count=%llu",
1095 (unsigned long long)pkt_dev->count);
1096 return count;
1098 if (!strcmp(name, "src_mac_count")) {
1099 len = num_arg(&user_buffer[i], 10, &value);
1100 if (len < 0)
1101 return len;
1103 i += len;
1104 if (pkt_dev->src_mac_count != value) {
1105 pkt_dev->src_mac_count = value;
1106 pkt_dev->cur_src_mac_offset = 0;
1108 sprintf(pg_result, "OK: src_mac_count=%d",
1109 pkt_dev->src_mac_count);
1110 return count;
1112 if (!strcmp(name, "dst_mac_count")) {
1113 len = num_arg(&user_buffer[i], 10, &value);
1114 if (len < 0)
1115 return len;
1117 i += len;
1118 if (pkt_dev->dst_mac_count != value) {
1119 pkt_dev->dst_mac_count = value;
1120 pkt_dev->cur_dst_mac_offset = 0;
1122 sprintf(pg_result, "OK: dst_mac_count=%d",
1123 pkt_dev->dst_mac_count);
1124 return count;
1126 if (!strcmp(name, "node")) {
1127 len = num_arg(&user_buffer[i], 10, &value);
1128 if (len < 0)
1129 return len;
1131 i += len;
1133 if (node_possible(value)) {
1134 pkt_dev->node = value;
1135 sprintf(pg_result, "OK: node=%d", pkt_dev->node);
1136 if (pkt_dev->page) {
1137 put_page(pkt_dev->page);
1138 pkt_dev->page = NULL;
1141 else
1142 sprintf(pg_result, "ERROR: node not possible");
1143 return count;
1145 if (!strcmp(name, "flag")) {
1146 char f[32];
1147 memset(f, 0, 32);
1148 len = strn_len(&user_buffer[i], sizeof(f) - 1);
1149 if (len < 0)
1150 return len;
1152 if (copy_from_user(f, &user_buffer[i], len))
1153 return -EFAULT;
1154 i += len;
1155 if (strcmp(f, "IPSRC_RND") == 0)
1156 pkt_dev->flags |= F_IPSRC_RND;
1158 else if (strcmp(f, "!IPSRC_RND") == 0)
1159 pkt_dev->flags &= ~F_IPSRC_RND;
1161 else if (strcmp(f, "TXSIZE_RND") == 0)
1162 pkt_dev->flags |= F_TXSIZE_RND;
1164 else if (strcmp(f, "!TXSIZE_RND") == 0)
1165 pkt_dev->flags &= ~F_TXSIZE_RND;
1167 else if (strcmp(f, "IPDST_RND") == 0)
1168 pkt_dev->flags |= F_IPDST_RND;
1170 else if (strcmp(f, "!IPDST_RND") == 0)
1171 pkt_dev->flags &= ~F_IPDST_RND;
1173 else if (strcmp(f, "UDPSRC_RND") == 0)
1174 pkt_dev->flags |= F_UDPSRC_RND;
1176 else if (strcmp(f, "!UDPSRC_RND") == 0)
1177 pkt_dev->flags &= ~F_UDPSRC_RND;
1179 else if (strcmp(f, "UDPDST_RND") == 0)
1180 pkt_dev->flags |= F_UDPDST_RND;
1182 else if (strcmp(f, "!UDPDST_RND") == 0)
1183 pkt_dev->flags &= ~F_UDPDST_RND;
1185 else if (strcmp(f, "MACSRC_RND") == 0)
1186 pkt_dev->flags |= F_MACSRC_RND;
1188 else if (strcmp(f, "!MACSRC_RND") == 0)
1189 pkt_dev->flags &= ~F_MACSRC_RND;
1191 else if (strcmp(f, "MACDST_RND") == 0)
1192 pkt_dev->flags |= F_MACDST_RND;
1194 else if (strcmp(f, "!MACDST_RND") == 0)
1195 pkt_dev->flags &= ~F_MACDST_RND;
1197 else if (strcmp(f, "MPLS_RND") == 0)
1198 pkt_dev->flags |= F_MPLS_RND;
1200 else if (strcmp(f, "!MPLS_RND") == 0)
1201 pkt_dev->flags &= ~F_MPLS_RND;
1203 else if (strcmp(f, "VID_RND") == 0)
1204 pkt_dev->flags |= F_VID_RND;
1206 else if (strcmp(f, "!VID_RND") == 0)
1207 pkt_dev->flags &= ~F_VID_RND;
1209 else if (strcmp(f, "SVID_RND") == 0)
1210 pkt_dev->flags |= F_SVID_RND;
1212 else if (strcmp(f, "!SVID_RND") == 0)
1213 pkt_dev->flags &= ~F_SVID_RND;
1215 else if (strcmp(f, "FLOW_SEQ") == 0)
1216 pkt_dev->flags |= F_FLOW_SEQ;
1218 else if (strcmp(f, "QUEUE_MAP_RND") == 0)
1219 pkt_dev->flags |= F_QUEUE_MAP_RND;
1221 else if (strcmp(f, "!QUEUE_MAP_RND") == 0)
1222 pkt_dev->flags &= ~F_QUEUE_MAP_RND;
1224 else if (strcmp(f, "QUEUE_MAP_CPU") == 0)
1225 pkt_dev->flags |= F_QUEUE_MAP_CPU;
1227 else if (strcmp(f, "!QUEUE_MAP_CPU") == 0)
1228 pkt_dev->flags &= ~F_QUEUE_MAP_CPU;
1229 #ifdef CONFIG_XFRM
1230 else if (strcmp(f, "IPSEC") == 0)
1231 pkt_dev->flags |= F_IPSEC_ON;
1232 #endif
1234 else if (strcmp(f, "!IPV6") == 0)
1235 pkt_dev->flags &= ~F_IPV6;
1237 else if (strcmp(f, "NODE_ALLOC") == 0)
1238 pkt_dev->flags |= F_NODE;
1240 else if (strcmp(f, "!NODE_ALLOC") == 0)
1241 pkt_dev->flags &= ~F_NODE;
1243 else if (strcmp(f, "UDPCSUM") == 0)
1244 pkt_dev->flags |= F_UDPCSUM;
1246 else if (strcmp(f, "!UDPCSUM") == 0)
1247 pkt_dev->flags &= ~F_UDPCSUM;
1249 else {
1250 sprintf(pg_result,
1251 "Flag -:%s:- unknown\nAvailable flags, (prepend ! to un-set flag):\n%s",
1253 "IPSRC_RND, IPDST_RND, UDPSRC_RND, UDPDST_RND, "
1254 "MACSRC_RND, MACDST_RND, TXSIZE_RND, IPV6, MPLS_RND, VID_RND, SVID_RND, FLOW_SEQ, IPSEC, NODE_ALLOC\n");
1255 return count;
1257 sprintf(pg_result, "OK: flags=0x%x", pkt_dev->flags);
1258 return count;
1260 if (!strcmp(name, "dst_min") || !strcmp(name, "dst")) {
1261 len = strn_len(&user_buffer[i], sizeof(pkt_dev->dst_min) - 1);
1262 if (len < 0)
1263 return len;
1265 if (copy_from_user(buf, &user_buffer[i], len))
1266 return -EFAULT;
1267 buf[len] = 0;
1268 if (strcmp(buf, pkt_dev->dst_min) != 0) {
1269 memset(pkt_dev->dst_min, 0, sizeof(pkt_dev->dst_min));
1270 strncpy(pkt_dev->dst_min, buf, len);
1271 pkt_dev->daddr_min = in_aton(pkt_dev->dst_min);
1272 pkt_dev->cur_daddr = pkt_dev->daddr_min;
1274 if (debug)
1275 pr_debug("dst_min set to: %s\n", pkt_dev->dst_min);
1276 i += len;
1277 sprintf(pg_result, "OK: dst_min=%s", pkt_dev->dst_min);
1278 return count;
1280 if (!strcmp(name, "dst_max")) {
1281 len = strn_len(&user_buffer[i], sizeof(pkt_dev->dst_max) - 1);
1282 if (len < 0)
1283 return len;
1286 if (copy_from_user(buf, &user_buffer[i], len))
1287 return -EFAULT;
1289 buf[len] = 0;
1290 if (strcmp(buf, pkt_dev->dst_max) != 0) {
1291 memset(pkt_dev->dst_max, 0, sizeof(pkt_dev->dst_max));
1292 strncpy(pkt_dev->dst_max, buf, len);
1293 pkt_dev->daddr_max = in_aton(pkt_dev->dst_max);
1294 pkt_dev->cur_daddr = pkt_dev->daddr_max;
1296 if (debug)
1297 pr_debug("dst_max set to: %s\n", pkt_dev->dst_max);
1298 i += len;
1299 sprintf(pg_result, "OK: dst_max=%s", pkt_dev->dst_max);
1300 return count;
1302 if (!strcmp(name, "dst6")) {
1303 len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1304 if (len < 0)
1305 return len;
1307 pkt_dev->flags |= F_IPV6;
1309 if (copy_from_user(buf, &user_buffer[i], len))
1310 return -EFAULT;
1311 buf[len] = 0;
1313 in6_pton(buf, -1, pkt_dev->in6_daddr.s6_addr, -1, NULL);
1314 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->in6_daddr);
1316 pkt_dev->cur_in6_daddr = pkt_dev->in6_daddr;
1318 if (debug)
1319 pr_debug("dst6 set to: %s\n", buf);
1321 i += len;
1322 sprintf(pg_result, "OK: dst6=%s", buf);
1323 return count;
1325 if (!strcmp(name, "dst6_min")) {
1326 len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1327 if (len < 0)
1328 return len;
1330 pkt_dev->flags |= F_IPV6;
1332 if (copy_from_user(buf, &user_buffer[i], len))
1333 return -EFAULT;
1334 buf[len] = 0;
1336 in6_pton(buf, -1, pkt_dev->min_in6_daddr.s6_addr, -1, NULL);
1337 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->min_in6_daddr);
1339 pkt_dev->cur_in6_daddr = pkt_dev->min_in6_daddr;
1340 if (debug)
1341 pr_debug("dst6_min set to: %s\n", buf);
1343 i += len;
1344 sprintf(pg_result, "OK: dst6_min=%s", buf);
1345 return count;
1347 if (!strcmp(name, "dst6_max")) {
1348 len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1349 if (len < 0)
1350 return len;
1352 pkt_dev->flags |= F_IPV6;
1354 if (copy_from_user(buf, &user_buffer[i], len))
1355 return -EFAULT;
1356 buf[len] = 0;
1358 in6_pton(buf, -1, pkt_dev->max_in6_daddr.s6_addr, -1, NULL);
1359 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->max_in6_daddr);
1361 if (debug)
1362 pr_debug("dst6_max set to: %s\n", buf);
1364 i += len;
1365 sprintf(pg_result, "OK: dst6_max=%s", buf);
1366 return count;
1368 if (!strcmp(name, "src6")) {
1369 len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1370 if (len < 0)
1371 return len;
1373 pkt_dev->flags |= F_IPV6;
1375 if (copy_from_user(buf, &user_buffer[i], len))
1376 return -EFAULT;
1377 buf[len] = 0;
1379 in6_pton(buf, -1, pkt_dev->in6_saddr.s6_addr, -1, NULL);
1380 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->in6_saddr);
1382 pkt_dev->cur_in6_saddr = pkt_dev->in6_saddr;
1384 if (debug)
1385 pr_debug("src6 set to: %s\n", buf);
1387 i += len;
1388 sprintf(pg_result, "OK: src6=%s", buf);
1389 return count;
1391 if (!strcmp(name, "src_min")) {
1392 len = strn_len(&user_buffer[i], sizeof(pkt_dev->src_min) - 1);
1393 if (len < 0)
1394 return len;
1396 if (copy_from_user(buf, &user_buffer[i], len))
1397 return -EFAULT;
1398 buf[len] = 0;
1399 if (strcmp(buf, pkt_dev->src_min) != 0) {
1400 memset(pkt_dev->src_min, 0, sizeof(pkt_dev->src_min));
1401 strncpy(pkt_dev->src_min, buf, len);
1402 pkt_dev->saddr_min = in_aton(pkt_dev->src_min);
1403 pkt_dev->cur_saddr = pkt_dev->saddr_min;
1405 if (debug)
1406 pr_debug("src_min set to: %s\n", pkt_dev->src_min);
1407 i += len;
1408 sprintf(pg_result, "OK: src_min=%s", pkt_dev->src_min);
1409 return count;
1411 if (!strcmp(name, "src_max")) {
1412 len = strn_len(&user_buffer[i], sizeof(pkt_dev->src_max) - 1);
1413 if (len < 0)
1414 return len;
1416 if (copy_from_user(buf, &user_buffer[i], len))
1417 return -EFAULT;
1418 buf[len] = 0;
1419 if (strcmp(buf, pkt_dev->src_max) != 0) {
1420 memset(pkt_dev->src_max, 0, sizeof(pkt_dev->src_max));
1421 strncpy(pkt_dev->src_max, buf, len);
1422 pkt_dev->saddr_max = in_aton(pkt_dev->src_max);
1423 pkt_dev->cur_saddr = pkt_dev->saddr_max;
1425 if (debug)
1426 pr_debug("src_max set to: %s\n", pkt_dev->src_max);
1427 i += len;
1428 sprintf(pg_result, "OK: src_max=%s", pkt_dev->src_max);
1429 return count;
1431 if (!strcmp(name, "dst_mac")) {
1432 len = strn_len(&user_buffer[i], sizeof(valstr) - 1);
1433 if (len < 0)
1434 return len;
1436 memset(valstr, 0, sizeof(valstr));
1437 if (copy_from_user(valstr, &user_buffer[i], len))
1438 return -EFAULT;
1440 if (!mac_pton(valstr, pkt_dev->dst_mac))
1441 return -EINVAL;
1442 /* Set up Dest MAC */
1443 ether_addr_copy(&pkt_dev->hh[0], pkt_dev->dst_mac);
1445 sprintf(pg_result, "OK: dstmac %pM", pkt_dev->dst_mac);
1446 return count;
1448 if (!strcmp(name, "src_mac")) {
1449 len = strn_len(&user_buffer[i], sizeof(valstr) - 1);
1450 if (len < 0)
1451 return len;
1453 memset(valstr, 0, sizeof(valstr));
1454 if (copy_from_user(valstr, &user_buffer[i], len))
1455 return -EFAULT;
1457 if (!mac_pton(valstr, pkt_dev->src_mac))
1458 return -EINVAL;
1459 /* Set up Src MAC */
1460 ether_addr_copy(&pkt_dev->hh[6], pkt_dev->src_mac);
1462 sprintf(pg_result, "OK: srcmac %pM", pkt_dev->src_mac);
1463 return count;
1466 if (!strcmp(name, "clear_counters")) {
1467 pktgen_clear_counters(pkt_dev);
1468 sprintf(pg_result, "OK: Clearing counters.\n");
1469 return count;
1472 if (!strcmp(name, "flows")) {
1473 len = num_arg(&user_buffer[i], 10, &value);
1474 if (len < 0)
1475 return len;
1477 i += len;
1478 if (value > MAX_CFLOWS)
1479 value = MAX_CFLOWS;
1481 pkt_dev->cflows = value;
1482 sprintf(pg_result, "OK: flows=%u", pkt_dev->cflows);
1483 return count;
1485 #ifdef CONFIG_XFRM
1486 if (!strcmp(name, "spi")) {
1487 len = num_arg(&user_buffer[i], 10, &value);
1488 if (len < 0)
1489 return len;
1491 i += len;
1492 pkt_dev->spi = value;
1493 sprintf(pg_result, "OK: spi=%u", pkt_dev->spi);
1494 return count;
1496 #endif
1497 if (!strcmp(name, "flowlen")) {
1498 len = num_arg(&user_buffer[i], 10, &value);
1499 if (len < 0)
1500 return len;
1502 i += len;
1503 pkt_dev->lflow = value;
1504 sprintf(pg_result, "OK: flowlen=%u", pkt_dev->lflow);
1505 return count;
1508 if (!strcmp(name, "queue_map_min")) {
1509 len = num_arg(&user_buffer[i], 5, &value);
1510 if (len < 0)
1511 return len;
1513 i += len;
1514 pkt_dev->queue_map_min = value;
1515 sprintf(pg_result, "OK: queue_map_min=%u", pkt_dev->queue_map_min);
1516 return count;
1519 if (!strcmp(name, "queue_map_max")) {
1520 len = num_arg(&user_buffer[i], 5, &value);
1521 if (len < 0)
1522 return len;
1524 i += len;
1525 pkt_dev->queue_map_max = value;
1526 sprintf(pg_result, "OK: queue_map_max=%u", pkt_dev->queue_map_max);
1527 return count;
1530 if (!strcmp(name, "mpls")) {
1531 unsigned int n, cnt;
1533 len = get_labels(&user_buffer[i], pkt_dev);
1534 if (len < 0)
1535 return len;
1536 i += len;
1537 cnt = sprintf(pg_result, "OK: mpls=");
1538 for (n = 0; n < pkt_dev->nr_labels; n++)
1539 cnt += sprintf(pg_result + cnt,
1540 "%08x%s", ntohl(pkt_dev->labels[n]),
1541 n == pkt_dev->nr_labels-1 ? "" : ",");
1543 if (pkt_dev->nr_labels && pkt_dev->vlan_id != 0xffff) {
1544 pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1545 pkt_dev->svlan_id = 0xffff;
1547 if (debug)
1548 pr_debug("VLAN/SVLAN auto turned off\n");
1550 return count;
1553 if (!strcmp(name, "vlan_id")) {
1554 len = num_arg(&user_buffer[i], 4, &value);
1555 if (len < 0)
1556 return len;
1558 i += len;
1559 if (value <= 4095) {
1560 pkt_dev->vlan_id = value; /* turn on VLAN */
1562 if (debug)
1563 pr_debug("VLAN turned on\n");
1565 if (debug && pkt_dev->nr_labels)
1566 pr_debug("MPLS auto turned off\n");
1568 pkt_dev->nr_labels = 0; /* turn off MPLS */
1569 sprintf(pg_result, "OK: vlan_id=%u", pkt_dev->vlan_id);
1570 } else {
1571 pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1572 pkt_dev->svlan_id = 0xffff;
1574 if (debug)
1575 pr_debug("VLAN/SVLAN turned off\n");
1577 return count;
1580 if (!strcmp(name, "vlan_p")) {
1581 len = num_arg(&user_buffer[i], 1, &value);
1582 if (len < 0)
1583 return len;
1585 i += len;
1586 if ((value <= 7) && (pkt_dev->vlan_id != 0xffff)) {
1587 pkt_dev->vlan_p = value;
1588 sprintf(pg_result, "OK: vlan_p=%u", pkt_dev->vlan_p);
1589 } else {
1590 sprintf(pg_result, "ERROR: vlan_p must be 0-7");
1592 return count;
1595 if (!strcmp(name, "vlan_cfi")) {
1596 len = num_arg(&user_buffer[i], 1, &value);
1597 if (len < 0)
1598 return len;
1600 i += len;
1601 if ((value <= 1) && (pkt_dev->vlan_id != 0xffff)) {
1602 pkt_dev->vlan_cfi = value;
1603 sprintf(pg_result, "OK: vlan_cfi=%u", pkt_dev->vlan_cfi);
1604 } else {
1605 sprintf(pg_result, "ERROR: vlan_cfi must be 0-1");
1607 return count;
1610 if (!strcmp(name, "svlan_id")) {
1611 len = num_arg(&user_buffer[i], 4, &value);
1612 if (len < 0)
1613 return len;
1615 i += len;
1616 if ((value <= 4095) && ((pkt_dev->vlan_id != 0xffff))) {
1617 pkt_dev->svlan_id = value; /* turn on SVLAN */
1619 if (debug)
1620 pr_debug("SVLAN turned on\n");
1622 if (debug && pkt_dev->nr_labels)
1623 pr_debug("MPLS auto turned off\n");
1625 pkt_dev->nr_labels = 0; /* turn off MPLS */
1626 sprintf(pg_result, "OK: svlan_id=%u", pkt_dev->svlan_id);
1627 } else {
1628 pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1629 pkt_dev->svlan_id = 0xffff;
1631 if (debug)
1632 pr_debug("VLAN/SVLAN turned off\n");
1634 return count;
1637 if (!strcmp(name, "svlan_p")) {
1638 len = num_arg(&user_buffer[i], 1, &value);
1639 if (len < 0)
1640 return len;
1642 i += len;
1643 if ((value <= 7) && (pkt_dev->svlan_id != 0xffff)) {
1644 pkt_dev->svlan_p = value;
1645 sprintf(pg_result, "OK: svlan_p=%u", pkt_dev->svlan_p);
1646 } else {
1647 sprintf(pg_result, "ERROR: svlan_p must be 0-7");
1649 return count;
1652 if (!strcmp(name, "svlan_cfi")) {
1653 len = num_arg(&user_buffer[i], 1, &value);
1654 if (len < 0)
1655 return len;
1657 i += len;
1658 if ((value <= 1) && (pkt_dev->svlan_id != 0xffff)) {
1659 pkt_dev->svlan_cfi = value;
1660 sprintf(pg_result, "OK: svlan_cfi=%u", pkt_dev->svlan_cfi);
1661 } else {
1662 sprintf(pg_result, "ERROR: svlan_cfi must be 0-1");
1664 return count;
1667 if (!strcmp(name, "tos")) {
1668 __u32 tmp_value = 0;
1669 len = hex32_arg(&user_buffer[i], 2, &tmp_value);
1670 if (len < 0)
1671 return len;
1673 i += len;
1674 if (len == 2) {
1675 pkt_dev->tos = tmp_value;
1676 sprintf(pg_result, "OK: tos=0x%02x", pkt_dev->tos);
1677 } else {
1678 sprintf(pg_result, "ERROR: tos must be 00-ff");
1680 return count;
1683 if (!strcmp(name, "traffic_class")) {
1684 __u32 tmp_value = 0;
1685 len = hex32_arg(&user_buffer[i], 2, &tmp_value);
1686 if (len < 0)
1687 return len;
1689 i += len;
1690 if (len == 2) {
1691 pkt_dev->traffic_class = tmp_value;
1692 sprintf(pg_result, "OK: traffic_class=0x%02x", pkt_dev->traffic_class);
1693 } else {
1694 sprintf(pg_result, "ERROR: traffic_class must be 00-ff");
1696 return count;
1699 if (!strcmp(name, "skb_priority")) {
1700 len = num_arg(&user_buffer[i], 9, &value);
1701 if (len < 0)
1702 return len;
1704 i += len;
1705 pkt_dev->skb_priority = value;
1706 sprintf(pg_result, "OK: skb_priority=%i",
1707 pkt_dev->skb_priority);
1708 return count;
1711 sprintf(pkt_dev->result, "No such parameter \"%s\"", name);
1712 return -EINVAL;
1715 static int pktgen_if_open(struct inode *inode, struct file *file)
1717 return single_open(file, pktgen_if_show, PDE_DATA(inode));
1720 static const struct file_operations pktgen_if_fops = {
1721 .owner = THIS_MODULE,
1722 .open = pktgen_if_open,
1723 .read = seq_read,
1724 .llseek = seq_lseek,
1725 .write = pktgen_if_write,
1726 .release = single_release,
1729 static int pktgen_thread_show(struct seq_file *seq, void *v)
1731 struct pktgen_thread *t = seq->private;
1732 const struct pktgen_dev *pkt_dev;
1734 BUG_ON(!t);
1736 seq_printf(seq, "Running: ");
1738 if_lock(t);
1739 list_for_each_entry(pkt_dev, &t->if_list, list)
1740 if (pkt_dev->running)
1741 seq_printf(seq, "%s ", pkt_dev->odevname);
1743 seq_printf(seq, "\nStopped: ");
1745 list_for_each_entry(pkt_dev, &t->if_list, list)
1746 if (!pkt_dev->running)
1747 seq_printf(seq, "%s ", pkt_dev->odevname);
1749 if (t->result[0])
1750 seq_printf(seq, "\nResult: %s\n", t->result);
1751 else
1752 seq_printf(seq, "\nResult: NA\n");
1754 if_unlock(t);
1756 return 0;
1759 static ssize_t pktgen_thread_write(struct file *file,
1760 const char __user * user_buffer,
1761 size_t count, loff_t * offset)
1763 struct seq_file *seq = file->private_data;
1764 struct pktgen_thread *t = seq->private;
1765 int i, max, len, ret;
1766 char name[40];
1767 char *pg_result;
1769 if (count < 1) {
1770 // sprintf(pg_result, "Wrong command format");
1771 return -EINVAL;
1774 max = count;
1775 len = count_trail_chars(user_buffer, max);
1776 if (len < 0)
1777 return len;
1779 i = len;
1781 /* Read variable name */
1783 len = strn_len(&user_buffer[i], sizeof(name) - 1);
1784 if (len < 0)
1785 return len;
1787 memset(name, 0, sizeof(name));
1788 if (copy_from_user(name, &user_buffer[i], len))
1789 return -EFAULT;
1790 i += len;
1792 max = count - i;
1793 len = count_trail_chars(&user_buffer[i], max);
1794 if (len < 0)
1795 return len;
1797 i += len;
1799 if (debug)
1800 pr_debug("t=%s, count=%lu\n", name, (unsigned long)count);
1802 if (!t) {
1803 pr_err("ERROR: No thread\n");
1804 ret = -EINVAL;
1805 goto out;
1808 pg_result = &(t->result[0]);
1810 if (!strcmp(name, "add_device")) {
1811 char f[32];
1812 memset(f, 0, 32);
1813 len = strn_len(&user_buffer[i], sizeof(f) - 1);
1814 if (len < 0) {
1815 ret = len;
1816 goto out;
1818 if (copy_from_user(f, &user_buffer[i], len))
1819 return -EFAULT;
1820 i += len;
1821 mutex_lock(&pktgen_thread_lock);
1822 ret = pktgen_add_device(t, f);
1823 mutex_unlock(&pktgen_thread_lock);
1824 if (!ret) {
1825 ret = count;
1826 sprintf(pg_result, "OK: add_device=%s", f);
1827 } else
1828 sprintf(pg_result, "ERROR: can not add device %s", f);
1829 goto out;
1832 if (!strcmp(name, "rem_device_all")) {
1833 mutex_lock(&pktgen_thread_lock);
1834 t->control |= T_REMDEVALL;
1835 mutex_unlock(&pktgen_thread_lock);
1836 schedule_timeout_interruptible(msecs_to_jiffies(125)); /* Propagate thread->control */
1837 ret = count;
1838 sprintf(pg_result, "OK: rem_device_all");
1839 goto out;
1842 if (!strcmp(name, "max_before_softirq")) {
1843 sprintf(pg_result, "OK: Note! max_before_softirq is obsoleted -- Do not use");
1844 ret = count;
1845 goto out;
1848 ret = -EINVAL;
1849 out:
1850 return ret;
1853 static int pktgen_thread_open(struct inode *inode, struct file *file)
1855 return single_open(file, pktgen_thread_show, PDE_DATA(inode));
1858 static const struct file_operations pktgen_thread_fops = {
1859 .owner = THIS_MODULE,
1860 .open = pktgen_thread_open,
1861 .read = seq_read,
1862 .llseek = seq_lseek,
1863 .write = pktgen_thread_write,
1864 .release = single_release,
1867 /* Think find or remove for NN */
1868 static struct pktgen_dev *__pktgen_NN_threads(const struct pktgen_net *pn,
1869 const char *ifname, int remove)
1871 struct pktgen_thread *t;
1872 struct pktgen_dev *pkt_dev = NULL;
1873 bool exact = (remove == FIND);
1875 list_for_each_entry(t, &pn->pktgen_threads, th_list) {
1876 pkt_dev = pktgen_find_dev(t, ifname, exact);
1877 if (pkt_dev) {
1878 if (remove) {
1879 if_lock(t);
1880 pkt_dev->removal_mark = 1;
1881 t->control |= T_REMDEV;
1882 if_unlock(t);
1884 break;
1887 return pkt_dev;
1891 * mark a device for removal
1893 static void pktgen_mark_device(const struct pktgen_net *pn, const char *ifname)
1895 struct pktgen_dev *pkt_dev = NULL;
1896 const int max_tries = 10, msec_per_try = 125;
1897 int i = 0;
1899 mutex_lock(&pktgen_thread_lock);
1900 pr_debug("%s: marking %s for removal\n", __func__, ifname);
1902 while (1) {
1904 pkt_dev = __pktgen_NN_threads(pn, ifname, REMOVE);
1905 if (pkt_dev == NULL)
1906 break; /* success */
1908 mutex_unlock(&pktgen_thread_lock);
1909 pr_debug("%s: waiting for %s to disappear....\n",
1910 __func__, ifname);
1911 schedule_timeout_interruptible(msecs_to_jiffies(msec_per_try));
1912 mutex_lock(&pktgen_thread_lock);
1914 if (++i >= max_tries) {
1915 pr_err("%s: timed out after waiting %d msec for device %s to be removed\n",
1916 __func__, msec_per_try * i, ifname);
1917 break;
1922 mutex_unlock(&pktgen_thread_lock);
1925 static void pktgen_change_name(const struct pktgen_net *pn, struct net_device *dev)
1927 struct pktgen_thread *t;
1929 list_for_each_entry(t, &pn->pktgen_threads, th_list) {
1930 struct pktgen_dev *pkt_dev;
1932 list_for_each_entry(pkt_dev, &t->if_list, list) {
1933 if (pkt_dev->odev != dev)
1934 continue;
1936 proc_remove(pkt_dev->entry);
1938 pkt_dev->entry = proc_create_data(dev->name, 0600,
1939 pn->proc_dir,
1940 &pktgen_if_fops,
1941 pkt_dev);
1942 if (!pkt_dev->entry)
1943 pr_err("can't move proc entry for '%s'\n",
1944 dev->name);
1945 break;
1950 static int pktgen_device_event(struct notifier_block *unused,
1951 unsigned long event, void *ptr)
1953 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1954 struct pktgen_net *pn = net_generic(dev_net(dev), pg_net_id);
1956 if (pn->pktgen_exiting)
1957 return NOTIFY_DONE;
1959 /* It is OK that we do not hold the group lock right now,
1960 * as we run under the RTNL lock.
1963 switch (event) {
1964 case NETDEV_CHANGENAME:
1965 pktgen_change_name(pn, dev);
1966 break;
1968 case NETDEV_UNREGISTER:
1969 pktgen_mark_device(pn, dev->name);
1970 break;
1973 return NOTIFY_DONE;
1976 static struct net_device *pktgen_dev_get_by_name(const struct pktgen_net *pn,
1977 struct pktgen_dev *pkt_dev,
1978 const char *ifname)
1980 char b[IFNAMSIZ+5];
1981 int i;
1983 for (i = 0; ifname[i] != '@'; i++) {
1984 if (i == IFNAMSIZ)
1985 break;
1987 b[i] = ifname[i];
1989 b[i] = 0;
1991 return dev_get_by_name(pn->net, b);
1995 /* Associate pktgen_dev with a device. */
1997 static int pktgen_setup_dev(const struct pktgen_net *pn,
1998 struct pktgen_dev *pkt_dev, const char *ifname)
2000 struct net_device *odev;
2001 int err;
2003 /* Clean old setups */
2004 if (pkt_dev->odev) {
2005 dev_put(pkt_dev->odev);
2006 pkt_dev->odev = NULL;
2009 odev = pktgen_dev_get_by_name(pn, pkt_dev, ifname);
2010 if (!odev) {
2011 pr_err("no such netdevice: \"%s\"\n", ifname);
2012 return -ENODEV;
2015 if (odev->type != ARPHRD_ETHER) {
2016 pr_err("not an ethernet device: \"%s\"\n", ifname);
2017 err = -EINVAL;
2018 } else if (!netif_running(odev)) {
2019 pr_err("device is down: \"%s\"\n", ifname);
2020 err = -ENETDOWN;
2021 } else {
2022 pkt_dev->odev = odev;
2023 return 0;
2026 dev_put(odev);
2027 return err;
2030 /* Read pkt_dev from the interface and set up internal pktgen_dev
2031 * structure to have the right information to create/send packets
2033 static void pktgen_setup_inject(struct pktgen_dev *pkt_dev)
2035 int ntxq;
2037 if (!pkt_dev->odev) {
2038 pr_err("ERROR: pkt_dev->odev == NULL in setup_inject\n");
2039 sprintf(pkt_dev->result,
2040 "ERROR: pkt_dev->odev == NULL in setup_inject.\n");
2041 return;
2044 /* make sure that we don't pick a non-existing transmit queue */
2045 ntxq = pkt_dev->odev->real_num_tx_queues;
2047 if (ntxq <= pkt_dev->queue_map_min) {
2048 pr_warning("WARNING: Requested queue_map_min (zero-based) (%d) exceeds valid range [0 - %d] for (%d) queues on %s, resetting\n",
2049 pkt_dev->queue_map_min, (ntxq ?: 1) - 1, ntxq,
2050 pkt_dev->odevname);
2051 pkt_dev->queue_map_min = (ntxq ?: 1) - 1;
2053 if (pkt_dev->queue_map_max >= ntxq) {
2054 pr_warning("WARNING: Requested queue_map_max (zero-based) (%d) exceeds valid range [0 - %d] for (%d) queues on %s, resetting\n",
2055 pkt_dev->queue_map_max, (ntxq ?: 1) - 1, ntxq,
2056 pkt_dev->odevname);
2057 pkt_dev->queue_map_max = (ntxq ?: 1) - 1;
2060 /* Default to the interface's mac if not explicitly set. */
2062 if (is_zero_ether_addr(pkt_dev->src_mac))
2063 ether_addr_copy(&(pkt_dev->hh[6]), pkt_dev->odev->dev_addr);
2065 /* Set up Dest MAC */
2066 ether_addr_copy(&(pkt_dev->hh[0]), pkt_dev->dst_mac);
2068 if (pkt_dev->flags & F_IPV6) {
2069 int i, set = 0, err = 1;
2070 struct inet6_dev *idev;
2072 if (pkt_dev->min_pkt_size == 0) {
2073 pkt_dev->min_pkt_size = 14 + sizeof(struct ipv6hdr)
2074 + sizeof(struct udphdr)
2075 + sizeof(struct pktgen_hdr)
2076 + pkt_dev->pkt_overhead;
2079 for (i = 0; i < IN6_ADDR_HSIZE; i++)
2080 if (pkt_dev->cur_in6_saddr.s6_addr[i]) {
2081 set = 1;
2082 break;
2085 if (!set) {
2088 * Use linklevel address if unconfigured.
2090 * use ipv6_get_lladdr if/when it's get exported
2093 rcu_read_lock();
2094 idev = __in6_dev_get(pkt_dev->odev);
2095 if (idev) {
2096 struct inet6_ifaddr *ifp;
2098 read_lock_bh(&idev->lock);
2099 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2100 if ((ifp->scope & IFA_LINK) &&
2101 !(ifp->flags & IFA_F_TENTATIVE)) {
2102 pkt_dev->cur_in6_saddr = ifp->addr;
2103 err = 0;
2104 break;
2107 read_unlock_bh(&idev->lock);
2109 rcu_read_unlock();
2110 if (err)
2111 pr_err("ERROR: IPv6 link address not available\n");
2113 } else {
2114 if (pkt_dev->min_pkt_size == 0) {
2115 pkt_dev->min_pkt_size = 14 + sizeof(struct iphdr)
2116 + sizeof(struct udphdr)
2117 + sizeof(struct pktgen_hdr)
2118 + pkt_dev->pkt_overhead;
2121 pkt_dev->saddr_min = 0;
2122 pkt_dev->saddr_max = 0;
2123 if (strlen(pkt_dev->src_min) == 0) {
2125 struct in_device *in_dev;
2127 rcu_read_lock();
2128 in_dev = __in_dev_get_rcu(pkt_dev->odev);
2129 if (in_dev) {
2130 if (in_dev->ifa_list) {
2131 pkt_dev->saddr_min =
2132 in_dev->ifa_list->ifa_address;
2133 pkt_dev->saddr_max = pkt_dev->saddr_min;
2136 rcu_read_unlock();
2137 } else {
2138 pkt_dev->saddr_min = in_aton(pkt_dev->src_min);
2139 pkt_dev->saddr_max = in_aton(pkt_dev->src_max);
2142 pkt_dev->daddr_min = in_aton(pkt_dev->dst_min);
2143 pkt_dev->daddr_max = in_aton(pkt_dev->dst_max);
2145 /* Initialize current values. */
2146 pkt_dev->cur_pkt_size = pkt_dev->min_pkt_size;
2147 if (pkt_dev->min_pkt_size > pkt_dev->max_pkt_size)
2148 pkt_dev->max_pkt_size = pkt_dev->min_pkt_size;
2150 pkt_dev->cur_dst_mac_offset = 0;
2151 pkt_dev->cur_src_mac_offset = 0;
2152 pkt_dev->cur_saddr = pkt_dev->saddr_min;
2153 pkt_dev->cur_daddr = pkt_dev->daddr_min;
2154 pkt_dev->cur_udp_dst = pkt_dev->udp_dst_min;
2155 pkt_dev->cur_udp_src = pkt_dev->udp_src_min;
2156 pkt_dev->nflows = 0;
2160 static void spin(struct pktgen_dev *pkt_dev, ktime_t spin_until)
2162 ktime_t start_time, end_time;
2163 s64 remaining;
2164 struct hrtimer_sleeper t;
2166 hrtimer_init_on_stack(&t.timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS);
2167 hrtimer_set_expires(&t.timer, spin_until);
2169 remaining = ktime_to_ns(hrtimer_expires_remaining(&t.timer));
2170 if (remaining <= 0) {
2171 pkt_dev->next_tx = ktime_add_ns(spin_until, pkt_dev->delay);
2172 return;
2175 start_time = ktime_get();
2176 if (remaining < 100000) {
2177 /* for small delays (<100us), just loop until limit is reached */
2178 do {
2179 end_time = ktime_get();
2180 } while (ktime_compare(end_time, spin_until) < 0);
2181 } else {
2182 /* see do_nanosleep */
2183 hrtimer_init_sleeper(&t, current);
2184 do {
2185 set_current_state(TASK_INTERRUPTIBLE);
2186 hrtimer_start_expires(&t.timer, HRTIMER_MODE_ABS);
2187 if (!hrtimer_active(&t.timer))
2188 t.task = NULL;
2190 if (likely(t.task))
2191 schedule();
2193 hrtimer_cancel(&t.timer);
2194 } while (t.task && pkt_dev->running && !signal_pending(current));
2195 __set_current_state(TASK_RUNNING);
2196 end_time = ktime_get();
2199 pkt_dev->idle_acc += ktime_to_ns(ktime_sub(end_time, start_time));
2200 pkt_dev->next_tx = ktime_add_ns(spin_until, pkt_dev->delay);
2203 static inline void set_pkt_overhead(struct pktgen_dev *pkt_dev)
2205 pkt_dev->pkt_overhead = 0;
2206 pkt_dev->pkt_overhead += pkt_dev->nr_labels*sizeof(u32);
2207 pkt_dev->pkt_overhead += VLAN_TAG_SIZE(pkt_dev);
2208 pkt_dev->pkt_overhead += SVLAN_TAG_SIZE(pkt_dev);
2211 static inline int f_seen(const struct pktgen_dev *pkt_dev, int flow)
2213 return !!(pkt_dev->flows[flow].flags & F_INIT);
2216 static inline int f_pick(struct pktgen_dev *pkt_dev)
2218 int flow = pkt_dev->curfl;
2220 if (pkt_dev->flags & F_FLOW_SEQ) {
2221 if (pkt_dev->flows[flow].count >= pkt_dev->lflow) {
2222 /* reset time */
2223 pkt_dev->flows[flow].count = 0;
2224 pkt_dev->flows[flow].flags = 0;
2225 pkt_dev->curfl += 1;
2226 if (pkt_dev->curfl >= pkt_dev->cflows)
2227 pkt_dev->curfl = 0; /*reset */
2229 } else {
2230 flow = prandom_u32() % pkt_dev->cflows;
2231 pkt_dev->curfl = flow;
2233 if (pkt_dev->flows[flow].count > pkt_dev->lflow) {
2234 pkt_dev->flows[flow].count = 0;
2235 pkt_dev->flows[flow].flags = 0;
2239 return pkt_dev->curfl;
2243 #ifdef CONFIG_XFRM
2244 /* If there was already an IPSEC SA, we keep it as is, else
2245 * we go look for it ...
2247 #define DUMMY_MARK 0
2248 static void get_ipsec_sa(struct pktgen_dev *pkt_dev, int flow)
2250 struct xfrm_state *x = pkt_dev->flows[flow].x;
2251 struct pktgen_net *pn = net_generic(dev_net(pkt_dev->odev), pg_net_id);
2252 if (!x) {
2254 if (pkt_dev->spi) {
2255 /* We need as quick as possible to find the right SA
2256 * Searching with minimum criteria to archieve this.
2258 x = xfrm_state_lookup_byspi(pn->net, htonl(pkt_dev->spi), AF_INET);
2259 } else {
2260 /* slow path: we dont already have xfrm_state */
2261 x = xfrm_stateonly_find(pn->net, DUMMY_MARK,
2262 (xfrm_address_t *)&pkt_dev->cur_daddr,
2263 (xfrm_address_t *)&pkt_dev->cur_saddr,
2264 AF_INET,
2265 pkt_dev->ipsmode,
2266 pkt_dev->ipsproto, 0);
2268 if (x) {
2269 pkt_dev->flows[flow].x = x;
2270 set_pkt_overhead(pkt_dev);
2271 pkt_dev->pkt_overhead += x->props.header_len;
2276 #endif
2277 static void set_cur_queue_map(struct pktgen_dev *pkt_dev)
2280 if (pkt_dev->flags & F_QUEUE_MAP_CPU)
2281 pkt_dev->cur_queue_map = smp_processor_id();
2283 else if (pkt_dev->queue_map_min <= pkt_dev->queue_map_max) {
2284 __u16 t;
2285 if (pkt_dev->flags & F_QUEUE_MAP_RND) {
2286 t = prandom_u32() %
2287 (pkt_dev->queue_map_max -
2288 pkt_dev->queue_map_min + 1)
2289 + pkt_dev->queue_map_min;
2290 } else {
2291 t = pkt_dev->cur_queue_map + 1;
2292 if (t > pkt_dev->queue_map_max)
2293 t = pkt_dev->queue_map_min;
2295 pkt_dev->cur_queue_map = t;
2297 pkt_dev->cur_queue_map = pkt_dev->cur_queue_map % pkt_dev->odev->real_num_tx_queues;
2300 /* Increment/randomize headers according to flags and current values
2301 * for IP src/dest, UDP src/dst port, MAC-Addr src/dst
2303 static void mod_cur_headers(struct pktgen_dev *pkt_dev)
2305 __u32 imn;
2306 __u32 imx;
2307 int flow = 0;
2309 if (pkt_dev->cflows)
2310 flow = f_pick(pkt_dev);
2312 /* Deal with source MAC */
2313 if (pkt_dev->src_mac_count > 1) {
2314 __u32 mc;
2315 __u32 tmp;
2317 if (pkt_dev->flags & F_MACSRC_RND)
2318 mc = prandom_u32() % pkt_dev->src_mac_count;
2319 else {
2320 mc = pkt_dev->cur_src_mac_offset++;
2321 if (pkt_dev->cur_src_mac_offset >=
2322 pkt_dev->src_mac_count)
2323 pkt_dev->cur_src_mac_offset = 0;
2326 tmp = pkt_dev->src_mac[5] + (mc & 0xFF);
2327 pkt_dev->hh[11] = tmp;
2328 tmp = (pkt_dev->src_mac[4] + ((mc >> 8) & 0xFF) + (tmp >> 8));
2329 pkt_dev->hh[10] = tmp;
2330 tmp = (pkt_dev->src_mac[3] + ((mc >> 16) & 0xFF) + (tmp >> 8));
2331 pkt_dev->hh[9] = tmp;
2332 tmp = (pkt_dev->src_mac[2] + ((mc >> 24) & 0xFF) + (tmp >> 8));
2333 pkt_dev->hh[8] = tmp;
2334 tmp = (pkt_dev->src_mac[1] + (tmp >> 8));
2335 pkt_dev->hh[7] = tmp;
2338 /* Deal with Destination MAC */
2339 if (pkt_dev->dst_mac_count > 1) {
2340 __u32 mc;
2341 __u32 tmp;
2343 if (pkt_dev->flags & F_MACDST_RND)
2344 mc = prandom_u32() % pkt_dev->dst_mac_count;
2346 else {
2347 mc = pkt_dev->cur_dst_mac_offset++;
2348 if (pkt_dev->cur_dst_mac_offset >=
2349 pkt_dev->dst_mac_count) {
2350 pkt_dev->cur_dst_mac_offset = 0;
2354 tmp = pkt_dev->dst_mac[5] + (mc & 0xFF);
2355 pkt_dev->hh[5] = tmp;
2356 tmp = (pkt_dev->dst_mac[4] + ((mc >> 8) & 0xFF) + (tmp >> 8));
2357 pkt_dev->hh[4] = tmp;
2358 tmp = (pkt_dev->dst_mac[3] + ((mc >> 16) & 0xFF) + (tmp >> 8));
2359 pkt_dev->hh[3] = tmp;
2360 tmp = (pkt_dev->dst_mac[2] + ((mc >> 24) & 0xFF) + (tmp >> 8));
2361 pkt_dev->hh[2] = tmp;
2362 tmp = (pkt_dev->dst_mac[1] + (tmp >> 8));
2363 pkt_dev->hh[1] = tmp;
2366 if (pkt_dev->flags & F_MPLS_RND) {
2367 unsigned int i;
2368 for (i = 0; i < pkt_dev->nr_labels; i++)
2369 if (pkt_dev->labels[i] & MPLS_STACK_BOTTOM)
2370 pkt_dev->labels[i] = MPLS_STACK_BOTTOM |
2371 ((__force __be32)prandom_u32() &
2372 htonl(0x000fffff));
2375 if ((pkt_dev->flags & F_VID_RND) && (pkt_dev->vlan_id != 0xffff)) {
2376 pkt_dev->vlan_id = prandom_u32() & (4096 - 1);
2379 if ((pkt_dev->flags & F_SVID_RND) && (pkt_dev->svlan_id != 0xffff)) {
2380 pkt_dev->svlan_id = prandom_u32() & (4096 - 1);
2383 if (pkt_dev->udp_src_min < pkt_dev->udp_src_max) {
2384 if (pkt_dev->flags & F_UDPSRC_RND)
2385 pkt_dev->cur_udp_src = prandom_u32() %
2386 (pkt_dev->udp_src_max - pkt_dev->udp_src_min)
2387 + pkt_dev->udp_src_min;
2389 else {
2390 pkt_dev->cur_udp_src++;
2391 if (pkt_dev->cur_udp_src >= pkt_dev->udp_src_max)
2392 pkt_dev->cur_udp_src = pkt_dev->udp_src_min;
2396 if (pkt_dev->udp_dst_min < pkt_dev->udp_dst_max) {
2397 if (pkt_dev->flags & F_UDPDST_RND) {
2398 pkt_dev->cur_udp_dst = prandom_u32() %
2399 (pkt_dev->udp_dst_max - pkt_dev->udp_dst_min)
2400 + pkt_dev->udp_dst_min;
2401 } else {
2402 pkt_dev->cur_udp_dst++;
2403 if (pkt_dev->cur_udp_dst >= pkt_dev->udp_dst_max)
2404 pkt_dev->cur_udp_dst = pkt_dev->udp_dst_min;
2408 if (!(pkt_dev->flags & F_IPV6)) {
2410 imn = ntohl(pkt_dev->saddr_min);
2411 imx = ntohl(pkt_dev->saddr_max);
2412 if (imn < imx) {
2413 __u32 t;
2414 if (pkt_dev->flags & F_IPSRC_RND)
2415 t = prandom_u32() % (imx - imn) + imn;
2416 else {
2417 t = ntohl(pkt_dev->cur_saddr);
2418 t++;
2419 if (t > imx)
2420 t = imn;
2423 pkt_dev->cur_saddr = htonl(t);
2426 if (pkt_dev->cflows && f_seen(pkt_dev, flow)) {
2427 pkt_dev->cur_daddr = pkt_dev->flows[flow].cur_daddr;
2428 } else {
2429 imn = ntohl(pkt_dev->daddr_min);
2430 imx = ntohl(pkt_dev->daddr_max);
2431 if (imn < imx) {
2432 __u32 t;
2433 __be32 s;
2434 if (pkt_dev->flags & F_IPDST_RND) {
2436 do {
2437 t = prandom_u32() %
2438 (imx - imn) + imn;
2439 s = htonl(t);
2440 } while (ipv4_is_loopback(s) ||
2441 ipv4_is_multicast(s) ||
2442 ipv4_is_lbcast(s) ||
2443 ipv4_is_zeronet(s) ||
2444 ipv4_is_local_multicast(s));
2445 pkt_dev->cur_daddr = s;
2446 } else {
2447 t = ntohl(pkt_dev->cur_daddr);
2448 t++;
2449 if (t > imx) {
2450 t = imn;
2452 pkt_dev->cur_daddr = htonl(t);
2455 if (pkt_dev->cflows) {
2456 pkt_dev->flows[flow].flags |= F_INIT;
2457 pkt_dev->flows[flow].cur_daddr =
2458 pkt_dev->cur_daddr;
2459 #ifdef CONFIG_XFRM
2460 if (pkt_dev->flags & F_IPSEC_ON)
2461 get_ipsec_sa(pkt_dev, flow);
2462 #endif
2463 pkt_dev->nflows++;
2466 } else { /* IPV6 * */
2468 if (!ipv6_addr_any(&pkt_dev->min_in6_daddr)) {
2469 int i;
2471 /* Only random destinations yet */
2473 for (i = 0; i < 4; i++) {
2474 pkt_dev->cur_in6_daddr.s6_addr32[i] =
2475 (((__force __be32)prandom_u32() |
2476 pkt_dev->min_in6_daddr.s6_addr32[i]) &
2477 pkt_dev->max_in6_daddr.s6_addr32[i]);
2482 if (pkt_dev->min_pkt_size < pkt_dev->max_pkt_size) {
2483 __u32 t;
2484 if (pkt_dev->flags & F_TXSIZE_RND) {
2485 t = prandom_u32() %
2486 (pkt_dev->max_pkt_size - pkt_dev->min_pkt_size)
2487 + pkt_dev->min_pkt_size;
2488 } else {
2489 t = pkt_dev->cur_pkt_size + 1;
2490 if (t > pkt_dev->max_pkt_size)
2491 t = pkt_dev->min_pkt_size;
2493 pkt_dev->cur_pkt_size = t;
2496 set_cur_queue_map(pkt_dev);
2498 pkt_dev->flows[flow].count++;
2502 #ifdef CONFIG_XFRM
2503 static u32 pktgen_dst_metrics[RTAX_MAX + 1] = {
2505 [RTAX_HOPLIMIT] = 0x5, /* Set a static hoplimit */
2508 static int pktgen_output_ipsec(struct sk_buff *skb, struct pktgen_dev *pkt_dev)
2510 struct xfrm_state *x = pkt_dev->flows[pkt_dev->curfl].x;
2511 int err = 0;
2512 struct net *net = dev_net(pkt_dev->odev);
2514 if (!x)
2515 return 0;
2516 /* XXX: we dont support tunnel mode for now until
2517 * we resolve the dst issue */
2518 if ((x->props.mode != XFRM_MODE_TRANSPORT) && (pkt_dev->spi == 0))
2519 return 0;
2521 /* But when user specify an valid SPI, transformation
2522 * supports both transport/tunnel mode + ESP/AH type.
2524 if ((x->props.mode == XFRM_MODE_TUNNEL) && (pkt_dev->spi != 0))
2525 skb->_skb_refdst = (unsigned long)&pkt_dev->dst | SKB_DST_NOREF;
2527 rcu_read_lock_bh();
2528 err = x->outer_mode->output(x, skb);
2529 rcu_read_unlock_bh();
2530 if (err) {
2531 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATEMODEERROR);
2532 goto error;
2534 err = x->type->output(x, skb);
2535 if (err) {
2536 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATEPROTOERROR);
2537 goto error;
2539 spin_lock_bh(&x->lock);
2540 x->curlft.bytes += skb->len;
2541 x->curlft.packets++;
2542 spin_unlock_bh(&x->lock);
2543 error:
2544 return err;
2547 static void free_SAs(struct pktgen_dev *pkt_dev)
2549 if (pkt_dev->cflows) {
2550 /* let go of the SAs if we have them */
2551 int i;
2552 for (i = 0; i < pkt_dev->cflows; i++) {
2553 struct xfrm_state *x = pkt_dev->flows[i].x;
2554 if (x) {
2555 xfrm_state_put(x);
2556 pkt_dev->flows[i].x = NULL;
2562 static int process_ipsec(struct pktgen_dev *pkt_dev,
2563 struct sk_buff *skb, __be16 protocol)
2565 if (pkt_dev->flags & F_IPSEC_ON) {
2566 struct xfrm_state *x = pkt_dev->flows[pkt_dev->curfl].x;
2567 int nhead = 0;
2568 if (x) {
2569 int ret;
2570 __u8 *eth;
2571 struct iphdr *iph;
2573 nhead = x->props.header_len - skb_headroom(skb);
2574 if (nhead > 0) {
2575 ret = pskb_expand_head(skb, nhead, 0, GFP_ATOMIC);
2576 if (ret < 0) {
2577 pr_err("Error expanding ipsec packet %d\n",
2578 ret);
2579 goto err;
2583 /* ipsec is not expecting ll header */
2584 skb_pull(skb, ETH_HLEN);
2585 ret = pktgen_output_ipsec(skb, pkt_dev);
2586 if (ret) {
2587 pr_err("Error creating ipsec packet %d\n", ret);
2588 goto err;
2590 /* restore ll */
2591 eth = (__u8 *) skb_push(skb, ETH_HLEN);
2592 memcpy(eth, pkt_dev->hh, 12);
2593 *(u16 *) &eth[12] = protocol;
2595 /* Update IPv4 header len as well as checksum value */
2596 iph = ip_hdr(skb);
2597 iph->tot_len = htons(skb->len - ETH_HLEN);
2598 ip_send_check(iph);
2601 return 1;
2602 err:
2603 kfree_skb(skb);
2604 return 0;
2606 #endif
2608 static void mpls_push(__be32 *mpls, struct pktgen_dev *pkt_dev)
2610 unsigned int i;
2611 for (i = 0; i < pkt_dev->nr_labels; i++)
2612 *mpls++ = pkt_dev->labels[i] & ~MPLS_STACK_BOTTOM;
2614 mpls--;
2615 *mpls |= MPLS_STACK_BOTTOM;
2618 static inline __be16 build_tci(unsigned int id, unsigned int cfi,
2619 unsigned int prio)
2621 return htons(id | (cfi << 12) | (prio << 13));
2624 static void pktgen_finalize_skb(struct pktgen_dev *pkt_dev, struct sk_buff *skb,
2625 int datalen)
2627 struct timeval timestamp;
2628 struct pktgen_hdr *pgh;
2630 pgh = (struct pktgen_hdr *)skb_put(skb, sizeof(*pgh));
2631 datalen -= sizeof(*pgh);
2633 if (pkt_dev->nfrags <= 0) {
2634 memset(skb_put(skb, datalen), 0, datalen);
2635 } else {
2636 int frags = pkt_dev->nfrags;
2637 int i, len;
2638 int frag_len;
2641 if (frags > MAX_SKB_FRAGS)
2642 frags = MAX_SKB_FRAGS;
2643 len = datalen - frags * PAGE_SIZE;
2644 if (len > 0) {
2645 memset(skb_put(skb, len), 0, len);
2646 datalen = frags * PAGE_SIZE;
2649 i = 0;
2650 frag_len = (datalen/frags) < PAGE_SIZE ?
2651 (datalen/frags) : PAGE_SIZE;
2652 while (datalen > 0) {
2653 if (unlikely(!pkt_dev->page)) {
2654 int node = numa_node_id();
2656 if (pkt_dev->node >= 0 && (pkt_dev->flags & F_NODE))
2657 node = pkt_dev->node;
2658 pkt_dev->page = alloc_pages_node(node, GFP_KERNEL | __GFP_ZERO, 0);
2659 if (!pkt_dev->page)
2660 break;
2662 get_page(pkt_dev->page);
2663 skb_frag_set_page(skb, i, pkt_dev->page);
2664 skb_shinfo(skb)->frags[i].page_offset = 0;
2665 /*last fragment, fill rest of data*/
2666 if (i == (frags - 1))
2667 skb_frag_size_set(&skb_shinfo(skb)->frags[i],
2668 (datalen < PAGE_SIZE ? datalen : PAGE_SIZE));
2669 else
2670 skb_frag_size_set(&skb_shinfo(skb)->frags[i], frag_len);
2671 datalen -= skb_frag_size(&skb_shinfo(skb)->frags[i]);
2672 skb->len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2673 skb->data_len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2674 i++;
2675 skb_shinfo(skb)->nr_frags = i;
2679 /* Stamp the time, and sequence number,
2680 * convert them to network byte order
2682 pgh->pgh_magic = htonl(PKTGEN_MAGIC);
2683 pgh->seq_num = htonl(pkt_dev->seq_num);
2685 do_gettimeofday(&timestamp);
2686 pgh->tv_sec = htonl(timestamp.tv_sec);
2687 pgh->tv_usec = htonl(timestamp.tv_usec);
2690 static struct sk_buff *pktgen_alloc_skb(struct net_device *dev,
2691 struct pktgen_dev *pkt_dev,
2692 unsigned int extralen)
2694 struct sk_buff *skb = NULL;
2695 unsigned int size = pkt_dev->cur_pkt_size + 64 + extralen +
2696 pkt_dev->pkt_overhead;
2698 if (pkt_dev->flags & F_NODE) {
2699 int node = pkt_dev->node >= 0 ? pkt_dev->node : numa_node_id();
2701 skb = __alloc_skb(NET_SKB_PAD + size, GFP_NOWAIT, 0, node);
2702 if (likely(skb)) {
2703 skb_reserve(skb, NET_SKB_PAD);
2704 skb->dev = dev;
2706 } else {
2707 skb = __netdev_alloc_skb(dev, size, GFP_NOWAIT);
2710 return skb;
2713 static struct sk_buff *fill_packet_ipv4(struct net_device *odev,
2714 struct pktgen_dev *pkt_dev)
2716 struct sk_buff *skb = NULL;
2717 __u8 *eth;
2718 struct udphdr *udph;
2719 int datalen, iplen;
2720 struct iphdr *iph;
2721 __be16 protocol = htons(ETH_P_IP);
2722 __be32 *mpls;
2723 __be16 *vlan_tci = NULL; /* Encapsulates priority and VLAN ID */
2724 __be16 *vlan_encapsulated_proto = NULL; /* packet type ID field (or len) for VLAN tag */
2725 __be16 *svlan_tci = NULL; /* Encapsulates priority and SVLAN ID */
2726 __be16 *svlan_encapsulated_proto = NULL; /* packet type ID field (or len) for SVLAN tag */
2727 u16 queue_map;
2729 if (pkt_dev->nr_labels)
2730 protocol = htons(ETH_P_MPLS_UC);
2732 if (pkt_dev->vlan_id != 0xffff)
2733 protocol = htons(ETH_P_8021Q);
2735 /* Update any of the values, used when we're incrementing various
2736 * fields.
2738 mod_cur_headers(pkt_dev);
2739 queue_map = pkt_dev->cur_queue_map;
2741 datalen = (odev->hard_header_len + 16) & ~0xf;
2743 skb = pktgen_alloc_skb(odev, pkt_dev, datalen);
2744 if (!skb) {
2745 sprintf(pkt_dev->result, "No memory");
2746 return NULL;
2749 prefetchw(skb->data);
2750 skb_reserve(skb, datalen);
2752 /* Reserve for ethernet and IP header */
2753 eth = (__u8 *) skb_push(skb, 14);
2754 mpls = (__be32 *)skb_put(skb, pkt_dev->nr_labels*sizeof(__u32));
2755 if (pkt_dev->nr_labels)
2756 mpls_push(mpls, pkt_dev);
2758 if (pkt_dev->vlan_id != 0xffff) {
2759 if (pkt_dev->svlan_id != 0xffff) {
2760 svlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2761 *svlan_tci = build_tci(pkt_dev->svlan_id,
2762 pkt_dev->svlan_cfi,
2763 pkt_dev->svlan_p);
2764 svlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2765 *svlan_encapsulated_proto = htons(ETH_P_8021Q);
2767 vlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2768 *vlan_tci = build_tci(pkt_dev->vlan_id,
2769 pkt_dev->vlan_cfi,
2770 pkt_dev->vlan_p);
2771 vlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2772 *vlan_encapsulated_proto = htons(ETH_P_IP);
2775 skb_set_mac_header(skb, 0);
2776 skb_set_network_header(skb, skb->len);
2777 iph = (struct iphdr *) skb_put(skb, sizeof(struct iphdr));
2779 skb_set_transport_header(skb, skb->len);
2780 udph = (struct udphdr *) skb_put(skb, sizeof(struct udphdr));
2781 skb_set_queue_mapping(skb, queue_map);
2782 skb->priority = pkt_dev->skb_priority;
2784 memcpy(eth, pkt_dev->hh, 12);
2785 *(__be16 *) & eth[12] = protocol;
2787 /* Eth + IPh + UDPh + mpls */
2788 datalen = pkt_dev->cur_pkt_size - 14 - 20 - 8 -
2789 pkt_dev->pkt_overhead;
2790 if (datalen < 0 || datalen < sizeof(struct pktgen_hdr))
2791 datalen = sizeof(struct pktgen_hdr);
2793 udph->source = htons(pkt_dev->cur_udp_src);
2794 udph->dest = htons(pkt_dev->cur_udp_dst);
2795 udph->len = htons(datalen + 8); /* DATA + udphdr */
2796 udph->check = 0;
2798 iph->ihl = 5;
2799 iph->version = 4;
2800 iph->ttl = 32;
2801 iph->tos = pkt_dev->tos;
2802 iph->protocol = IPPROTO_UDP; /* UDP */
2803 iph->saddr = pkt_dev->cur_saddr;
2804 iph->daddr = pkt_dev->cur_daddr;
2805 iph->id = htons(pkt_dev->ip_id);
2806 pkt_dev->ip_id++;
2807 iph->frag_off = 0;
2808 iplen = 20 + 8 + datalen;
2809 iph->tot_len = htons(iplen);
2810 ip_send_check(iph);
2811 skb->protocol = protocol;
2812 skb->dev = odev;
2813 skb->pkt_type = PACKET_HOST;
2815 if (!(pkt_dev->flags & F_UDPCSUM)) {
2816 skb->ip_summed = CHECKSUM_NONE;
2817 } else if (odev->features & NETIF_F_V4_CSUM) {
2818 skb->ip_summed = CHECKSUM_PARTIAL;
2819 skb->csum = 0;
2820 udp4_hwcsum(skb, udph->source, udph->dest);
2821 } else {
2822 __wsum csum = udp_csum(skb);
2824 /* add protocol-dependent pseudo-header */
2825 udph->check = csum_tcpudp_magic(udph->source, udph->dest,
2826 datalen + 8, IPPROTO_UDP, csum);
2828 if (udph->check == 0)
2829 udph->check = CSUM_MANGLED_0;
2832 pktgen_finalize_skb(pkt_dev, skb, datalen);
2834 #ifdef CONFIG_XFRM
2835 if (!process_ipsec(pkt_dev, skb, protocol))
2836 return NULL;
2837 #endif
2839 return skb;
2842 static struct sk_buff *fill_packet_ipv6(struct net_device *odev,
2843 struct pktgen_dev *pkt_dev)
2845 struct sk_buff *skb = NULL;
2846 __u8 *eth;
2847 struct udphdr *udph;
2848 int datalen, udplen;
2849 struct ipv6hdr *iph;
2850 __be16 protocol = htons(ETH_P_IPV6);
2851 __be32 *mpls;
2852 __be16 *vlan_tci = NULL; /* Encapsulates priority and VLAN ID */
2853 __be16 *vlan_encapsulated_proto = NULL; /* packet type ID field (or len) for VLAN tag */
2854 __be16 *svlan_tci = NULL; /* Encapsulates priority and SVLAN ID */
2855 __be16 *svlan_encapsulated_proto = NULL; /* packet type ID field (or len) for SVLAN tag */
2856 u16 queue_map;
2858 if (pkt_dev->nr_labels)
2859 protocol = htons(ETH_P_MPLS_UC);
2861 if (pkt_dev->vlan_id != 0xffff)
2862 protocol = htons(ETH_P_8021Q);
2864 /* Update any of the values, used when we're incrementing various
2865 * fields.
2867 mod_cur_headers(pkt_dev);
2868 queue_map = pkt_dev->cur_queue_map;
2870 skb = pktgen_alloc_skb(odev, pkt_dev, 16);
2871 if (!skb) {
2872 sprintf(pkt_dev->result, "No memory");
2873 return NULL;
2876 prefetchw(skb->data);
2877 skb_reserve(skb, 16);
2879 /* Reserve for ethernet and IP header */
2880 eth = (__u8 *) skb_push(skb, 14);
2881 mpls = (__be32 *)skb_put(skb, pkt_dev->nr_labels*sizeof(__u32));
2882 if (pkt_dev->nr_labels)
2883 mpls_push(mpls, pkt_dev);
2885 if (pkt_dev->vlan_id != 0xffff) {
2886 if (pkt_dev->svlan_id != 0xffff) {
2887 svlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2888 *svlan_tci = build_tci(pkt_dev->svlan_id,
2889 pkt_dev->svlan_cfi,
2890 pkt_dev->svlan_p);
2891 svlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2892 *svlan_encapsulated_proto = htons(ETH_P_8021Q);
2894 vlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2895 *vlan_tci = build_tci(pkt_dev->vlan_id,
2896 pkt_dev->vlan_cfi,
2897 pkt_dev->vlan_p);
2898 vlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2899 *vlan_encapsulated_proto = htons(ETH_P_IPV6);
2902 skb_set_mac_header(skb, 0);
2903 skb_set_network_header(skb, skb->len);
2904 iph = (struct ipv6hdr *) skb_put(skb, sizeof(struct ipv6hdr));
2906 skb_set_transport_header(skb, skb->len);
2907 udph = (struct udphdr *) skb_put(skb, sizeof(struct udphdr));
2908 skb_set_queue_mapping(skb, queue_map);
2909 skb->priority = pkt_dev->skb_priority;
2911 memcpy(eth, pkt_dev->hh, 12);
2912 *(__be16 *) &eth[12] = protocol;
2914 /* Eth + IPh + UDPh + mpls */
2915 datalen = pkt_dev->cur_pkt_size - 14 -
2916 sizeof(struct ipv6hdr) - sizeof(struct udphdr) -
2917 pkt_dev->pkt_overhead;
2919 if (datalen < 0 || datalen < sizeof(struct pktgen_hdr)) {
2920 datalen = sizeof(struct pktgen_hdr);
2921 net_info_ratelimited("increased datalen to %d\n", datalen);
2924 udplen = datalen + sizeof(struct udphdr);
2925 udph->source = htons(pkt_dev->cur_udp_src);
2926 udph->dest = htons(pkt_dev->cur_udp_dst);
2927 udph->len = htons(udplen);
2928 udph->check = 0;
2930 *(__be32 *) iph = htonl(0x60000000); /* Version + flow */
2932 if (pkt_dev->traffic_class) {
2933 /* Version + traffic class + flow (0) */
2934 *(__be32 *)iph |= htonl(0x60000000 | (pkt_dev->traffic_class << 20));
2937 iph->hop_limit = 32;
2939 iph->payload_len = htons(udplen);
2940 iph->nexthdr = IPPROTO_UDP;
2942 iph->daddr = pkt_dev->cur_in6_daddr;
2943 iph->saddr = pkt_dev->cur_in6_saddr;
2945 skb->protocol = protocol;
2946 skb->dev = odev;
2947 skb->pkt_type = PACKET_HOST;
2949 if (!(pkt_dev->flags & F_UDPCSUM)) {
2950 skb->ip_summed = CHECKSUM_NONE;
2951 } else if (odev->features & NETIF_F_V6_CSUM) {
2952 skb->ip_summed = CHECKSUM_PARTIAL;
2953 skb->csum_start = skb_transport_header(skb) - skb->head;
2954 skb->csum_offset = offsetof(struct udphdr, check);
2955 udph->check = ~csum_ipv6_magic(&iph->saddr, &iph->daddr, udplen, IPPROTO_UDP, 0);
2956 } else {
2957 __wsum csum = udp_csum(skb);
2959 /* add protocol-dependent pseudo-header */
2960 udph->check = csum_ipv6_magic(&iph->saddr, &iph->daddr, udplen, IPPROTO_UDP, csum);
2962 if (udph->check == 0)
2963 udph->check = CSUM_MANGLED_0;
2966 pktgen_finalize_skb(pkt_dev, skb, datalen);
2968 return skb;
2971 static struct sk_buff *fill_packet(struct net_device *odev,
2972 struct pktgen_dev *pkt_dev)
2974 if (pkt_dev->flags & F_IPV6)
2975 return fill_packet_ipv6(odev, pkt_dev);
2976 else
2977 return fill_packet_ipv4(odev, pkt_dev);
2980 static void pktgen_clear_counters(struct pktgen_dev *pkt_dev)
2982 pkt_dev->seq_num = 1;
2983 pkt_dev->idle_acc = 0;
2984 pkt_dev->sofar = 0;
2985 pkt_dev->tx_bytes = 0;
2986 pkt_dev->errors = 0;
2989 /* Set up structure for sending pkts, clear counters */
2991 static void pktgen_run(struct pktgen_thread *t)
2993 struct pktgen_dev *pkt_dev;
2994 int started = 0;
2996 func_enter();
2998 if_lock(t);
2999 list_for_each_entry(pkt_dev, &t->if_list, list) {
3002 * setup odev and create initial packet.
3004 pktgen_setup_inject(pkt_dev);
3006 if (pkt_dev->odev) {
3007 pktgen_clear_counters(pkt_dev);
3008 pkt_dev->running = 1; /* Cranke yeself! */
3009 pkt_dev->skb = NULL;
3010 pkt_dev->started_at = pkt_dev->next_tx = ktime_get();
3012 set_pkt_overhead(pkt_dev);
3014 strcpy(pkt_dev->result, "Starting");
3015 started++;
3016 } else
3017 strcpy(pkt_dev->result, "Error starting");
3019 if_unlock(t);
3020 if (started)
3021 t->control &= ~(T_STOP);
3024 static void pktgen_stop_all_threads_ifs(struct pktgen_net *pn)
3026 struct pktgen_thread *t;
3028 func_enter();
3030 mutex_lock(&pktgen_thread_lock);
3032 list_for_each_entry(t, &pn->pktgen_threads, th_list)
3033 t->control |= T_STOP;
3035 mutex_unlock(&pktgen_thread_lock);
3038 static int thread_is_running(const struct pktgen_thread *t)
3040 const struct pktgen_dev *pkt_dev;
3042 list_for_each_entry(pkt_dev, &t->if_list, list)
3043 if (pkt_dev->running)
3044 return 1;
3045 return 0;
3048 static int pktgen_wait_thread_run(struct pktgen_thread *t)
3050 if_lock(t);
3052 while (thread_is_running(t)) {
3054 if_unlock(t);
3056 msleep_interruptible(100);
3058 if (signal_pending(current))
3059 goto signal;
3060 if_lock(t);
3062 if_unlock(t);
3063 return 1;
3064 signal:
3065 return 0;
3068 static int pktgen_wait_all_threads_run(struct pktgen_net *pn)
3070 struct pktgen_thread *t;
3071 int sig = 1;
3073 mutex_lock(&pktgen_thread_lock);
3075 list_for_each_entry(t, &pn->pktgen_threads, th_list) {
3076 sig = pktgen_wait_thread_run(t);
3077 if (sig == 0)
3078 break;
3081 if (sig == 0)
3082 list_for_each_entry(t, &pn->pktgen_threads, th_list)
3083 t->control |= (T_STOP);
3085 mutex_unlock(&pktgen_thread_lock);
3086 return sig;
3089 static void pktgen_run_all_threads(struct pktgen_net *pn)
3091 struct pktgen_thread *t;
3093 func_enter();
3095 mutex_lock(&pktgen_thread_lock);
3097 list_for_each_entry(t, &pn->pktgen_threads, th_list)
3098 t->control |= (T_RUN);
3100 mutex_unlock(&pktgen_thread_lock);
3102 /* Propagate thread->control */
3103 schedule_timeout_interruptible(msecs_to_jiffies(125));
3105 pktgen_wait_all_threads_run(pn);
3108 static void pktgen_reset_all_threads(struct pktgen_net *pn)
3110 struct pktgen_thread *t;
3112 func_enter();
3114 mutex_lock(&pktgen_thread_lock);
3116 list_for_each_entry(t, &pn->pktgen_threads, th_list)
3117 t->control |= (T_REMDEVALL);
3119 mutex_unlock(&pktgen_thread_lock);
3121 /* Propagate thread->control */
3122 schedule_timeout_interruptible(msecs_to_jiffies(125));
3124 pktgen_wait_all_threads_run(pn);
3127 static void show_results(struct pktgen_dev *pkt_dev, int nr_frags)
3129 __u64 bps, mbps, pps;
3130 char *p = pkt_dev->result;
3131 ktime_t elapsed = ktime_sub(pkt_dev->stopped_at,
3132 pkt_dev->started_at);
3133 ktime_t idle = ns_to_ktime(pkt_dev->idle_acc);
3135 p += sprintf(p, "OK: %llu(c%llu+d%llu) usec, %llu (%dbyte,%dfrags)\n",
3136 (unsigned long long)ktime_to_us(elapsed),
3137 (unsigned long long)ktime_to_us(ktime_sub(elapsed, idle)),
3138 (unsigned long long)ktime_to_us(idle),
3139 (unsigned long long)pkt_dev->sofar,
3140 pkt_dev->cur_pkt_size, nr_frags);
3142 pps = div64_u64(pkt_dev->sofar * NSEC_PER_SEC,
3143 ktime_to_ns(elapsed));
3145 bps = pps * 8 * pkt_dev->cur_pkt_size;
3147 mbps = bps;
3148 do_div(mbps, 1000000);
3149 p += sprintf(p, " %llupps %lluMb/sec (%llubps) errors: %llu",
3150 (unsigned long long)pps,
3151 (unsigned long long)mbps,
3152 (unsigned long long)bps,
3153 (unsigned long long)pkt_dev->errors);
3156 /* Set stopped-at timer, remove from running list, do counters & statistics */
3157 static int pktgen_stop_device(struct pktgen_dev *pkt_dev)
3159 int nr_frags = pkt_dev->skb ? skb_shinfo(pkt_dev->skb)->nr_frags : -1;
3161 if (!pkt_dev->running) {
3162 pr_warning("interface: %s is already stopped\n",
3163 pkt_dev->odevname);
3164 return -EINVAL;
3167 kfree_skb(pkt_dev->skb);
3168 pkt_dev->skb = NULL;
3169 pkt_dev->stopped_at = ktime_get();
3170 pkt_dev->running = 0;
3172 show_results(pkt_dev, nr_frags);
3174 return 0;
3177 static struct pktgen_dev *next_to_run(struct pktgen_thread *t)
3179 struct pktgen_dev *pkt_dev, *best = NULL;
3181 if_lock(t);
3183 list_for_each_entry(pkt_dev, &t->if_list, list) {
3184 if (!pkt_dev->running)
3185 continue;
3186 if (best == NULL)
3187 best = pkt_dev;
3188 else if (ktime_compare(pkt_dev->next_tx, best->next_tx) < 0)
3189 best = pkt_dev;
3191 if_unlock(t);
3192 return best;
3195 static void pktgen_stop(struct pktgen_thread *t)
3197 struct pktgen_dev *pkt_dev;
3199 func_enter();
3201 if_lock(t);
3203 list_for_each_entry(pkt_dev, &t->if_list, list) {
3204 pktgen_stop_device(pkt_dev);
3207 if_unlock(t);
3211 * one of our devices needs to be removed - find it
3212 * and remove it
3214 static void pktgen_rem_one_if(struct pktgen_thread *t)
3216 struct list_head *q, *n;
3217 struct pktgen_dev *cur;
3219 func_enter();
3221 if_lock(t);
3223 list_for_each_safe(q, n, &t->if_list) {
3224 cur = list_entry(q, struct pktgen_dev, list);
3226 if (!cur->removal_mark)
3227 continue;
3229 kfree_skb(cur->skb);
3230 cur->skb = NULL;
3232 pktgen_remove_device(t, cur);
3234 break;
3237 if_unlock(t);
3240 static void pktgen_rem_all_ifs(struct pktgen_thread *t)
3242 struct list_head *q, *n;
3243 struct pktgen_dev *cur;
3245 func_enter();
3247 /* Remove all devices, free mem */
3249 if_lock(t);
3251 list_for_each_safe(q, n, &t->if_list) {
3252 cur = list_entry(q, struct pktgen_dev, list);
3254 kfree_skb(cur->skb);
3255 cur->skb = NULL;
3257 pktgen_remove_device(t, cur);
3260 if_unlock(t);
3263 static void pktgen_rem_thread(struct pktgen_thread *t)
3265 /* Remove from the thread list */
3266 remove_proc_entry(t->tsk->comm, t->net->proc_dir);
3269 static void pktgen_resched(struct pktgen_dev *pkt_dev)
3271 ktime_t idle_start = ktime_get();
3272 schedule();
3273 pkt_dev->idle_acc += ktime_to_ns(ktime_sub(ktime_get(), idle_start));
3276 static void pktgen_wait_for_skb(struct pktgen_dev *pkt_dev)
3278 ktime_t idle_start = ktime_get();
3280 while (atomic_read(&(pkt_dev->skb->users)) != 1) {
3281 if (signal_pending(current))
3282 break;
3284 if (need_resched())
3285 pktgen_resched(pkt_dev);
3286 else
3287 cpu_relax();
3289 pkt_dev->idle_acc += ktime_to_ns(ktime_sub(ktime_get(), idle_start));
3292 static void pktgen_xmit(struct pktgen_dev *pkt_dev)
3294 struct net_device *odev = pkt_dev->odev;
3295 netdev_tx_t (*xmit)(struct sk_buff *, struct net_device *)
3296 = odev->netdev_ops->ndo_start_xmit;
3297 struct netdev_queue *txq;
3298 u16 queue_map;
3299 int ret;
3301 /* If device is offline, then don't send */
3302 if (unlikely(!netif_running(odev) || !netif_carrier_ok(odev))) {
3303 pktgen_stop_device(pkt_dev);
3304 return;
3307 /* This is max DELAY, this has special meaning of
3308 * "never transmit"
3310 if (unlikely(pkt_dev->delay == ULLONG_MAX)) {
3311 pkt_dev->next_tx = ktime_add_ns(ktime_get(), ULONG_MAX);
3312 return;
3315 /* If no skb or clone count exhausted then get new one */
3316 if (!pkt_dev->skb || (pkt_dev->last_ok &&
3317 ++pkt_dev->clone_count >= pkt_dev->clone_skb)) {
3318 /* build a new pkt */
3319 kfree_skb(pkt_dev->skb);
3321 pkt_dev->skb = fill_packet(odev, pkt_dev);
3322 if (pkt_dev->skb == NULL) {
3323 pr_err("ERROR: couldn't allocate skb in fill_packet\n");
3324 schedule();
3325 pkt_dev->clone_count--; /* back out increment, OOM */
3326 return;
3328 pkt_dev->last_pkt_size = pkt_dev->skb->len;
3329 pkt_dev->allocated_skbs++;
3330 pkt_dev->clone_count = 0; /* reset counter */
3333 if (pkt_dev->delay && pkt_dev->last_ok)
3334 spin(pkt_dev, pkt_dev->next_tx);
3336 queue_map = skb_get_queue_mapping(pkt_dev->skb);
3337 txq = netdev_get_tx_queue(odev, queue_map);
3339 __netif_tx_lock_bh(txq);
3341 if (unlikely(netif_xmit_frozen_or_stopped(txq))) {
3342 ret = NETDEV_TX_BUSY;
3343 pkt_dev->last_ok = 0;
3344 goto unlock;
3346 atomic_inc(&(pkt_dev->skb->users));
3347 ret = (*xmit)(pkt_dev->skb, odev);
3349 switch (ret) {
3350 case NETDEV_TX_OK:
3351 txq_trans_update(txq);
3352 pkt_dev->last_ok = 1;
3353 pkt_dev->sofar++;
3354 pkt_dev->seq_num++;
3355 pkt_dev->tx_bytes += pkt_dev->last_pkt_size;
3356 break;
3357 case NET_XMIT_DROP:
3358 case NET_XMIT_CN:
3359 case NET_XMIT_POLICED:
3360 /* skb has been consumed */
3361 pkt_dev->errors++;
3362 break;
3363 default: /* Drivers are not supposed to return other values! */
3364 net_info_ratelimited("%s xmit error: %d\n",
3365 pkt_dev->odevname, ret);
3366 pkt_dev->errors++;
3367 /* fallthru */
3368 case NETDEV_TX_LOCKED:
3369 case NETDEV_TX_BUSY:
3370 /* Retry it next time */
3371 atomic_dec(&(pkt_dev->skb->users));
3372 pkt_dev->last_ok = 0;
3374 unlock:
3375 __netif_tx_unlock_bh(txq);
3377 /* If pkt_dev->count is zero, then run forever */
3378 if ((pkt_dev->count != 0) && (pkt_dev->sofar >= pkt_dev->count)) {
3379 pktgen_wait_for_skb(pkt_dev);
3381 /* Done with this */
3382 pktgen_stop_device(pkt_dev);
3387 * Main loop of the thread goes here
3390 static int pktgen_thread_worker(void *arg)
3392 DEFINE_WAIT(wait);
3393 struct pktgen_thread *t = arg;
3394 struct pktgen_dev *pkt_dev = NULL;
3395 int cpu = t->cpu;
3397 BUG_ON(smp_processor_id() != cpu);
3399 init_waitqueue_head(&t->queue);
3400 complete(&t->start_done);
3402 pr_debug("starting pktgen/%d: pid=%d\n", cpu, task_pid_nr(current));
3404 set_current_state(TASK_INTERRUPTIBLE);
3406 set_freezable();
3408 while (!kthread_should_stop()) {
3409 pkt_dev = next_to_run(t);
3411 if (unlikely(!pkt_dev && t->control == 0)) {
3412 if (t->net->pktgen_exiting)
3413 break;
3414 wait_event_interruptible_timeout(t->queue,
3415 t->control != 0,
3416 HZ/10);
3417 try_to_freeze();
3418 continue;
3421 __set_current_state(TASK_RUNNING);
3423 if (likely(pkt_dev)) {
3424 pktgen_xmit(pkt_dev);
3426 if (need_resched())
3427 pktgen_resched(pkt_dev);
3428 else
3429 cpu_relax();
3432 if (t->control & T_STOP) {
3433 pktgen_stop(t);
3434 t->control &= ~(T_STOP);
3437 if (t->control & T_RUN) {
3438 pktgen_run(t);
3439 t->control &= ~(T_RUN);
3442 if (t->control & T_REMDEVALL) {
3443 pktgen_rem_all_ifs(t);
3444 t->control &= ~(T_REMDEVALL);
3447 if (t->control & T_REMDEV) {
3448 pktgen_rem_one_if(t);
3449 t->control &= ~(T_REMDEV);
3452 try_to_freeze();
3454 set_current_state(TASK_INTERRUPTIBLE);
3457 pr_debug("%s stopping all device\n", t->tsk->comm);
3458 pktgen_stop(t);
3460 pr_debug("%s removing all device\n", t->tsk->comm);
3461 pktgen_rem_all_ifs(t);
3463 pr_debug("%s removing thread\n", t->tsk->comm);
3464 pktgen_rem_thread(t);
3466 /* Wait for kthread_stop */
3467 while (!kthread_should_stop()) {
3468 set_current_state(TASK_INTERRUPTIBLE);
3469 schedule();
3471 __set_current_state(TASK_RUNNING);
3473 return 0;
3476 static struct pktgen_dev *pktgen_find_dev(struct pktgen_thread *t,
3477 const char *ifname, bool exact)
3479 struct pktgen_dev *p, *pkt_dev = NULL;
3480 size_t len = strlen(ifname);
3482 if_lock(t);
3483 list_for_each_entry(p, &t->if_list, list)
3484 if (strncmp(p->odevname, ifname, len) == 0) {
3485 if (p->odevname[len]) {
3486 if (exact || p->odevname[len] != '@')
3487 continue;
3489 pkt_dev = p;
3490 break;
3493 if_unlock(t);
3494 pr_debug("find_dev(%s) returning %p\n", ifname, pkt_dev);
3495 return pkt_dev;
3499 * Adds a dev at front of if_list.
3502 static int add_dev_to_thread(struct pktgen_thread *t,
3503 struct pktgen_dev *pkt_dev)
3505 int rv = 0;
3507 if_lock(t);
3509 if (pkt_dev->pg_thread) {
3510 pr_err("ERROR: already assigned to a thread\n");
3511 rv = -EBUSY;
3512 goto out;
3515 list_add(&pkt_dev->list, &t->if_list);
3516 pkt_dev->pg_thread = t;
3517 pkt_dev->running = 0;
3519 out:
3520 if_unlock(t);
3521 return rv;
3524 /* Called under thread lock */
3526 static int pktgen_add_device(struct pktgen_thread *t, const char *ifname)
3528 struct pktgen_dev *pkt_dev;
3529 int err;
3530 int node = cpu_to_node(t->cpu);
3532 /* We don't allow a device to be on several threads */
3534 pkt_dev = __pktgen_NN_threads(t->net, ifname, FIND);
3535 if (pkt_dev) {
3536 pr_err("ERROR: interface already used\n");
3537 return -EBUSY;
3540 pkt_dev = kzalloc_node(sizeof(struct pktgen_dev), GFP_KERNEL, node);
3541 if (!pkt_dev)
3542 return -ENOMEM;
3544 strcpy(pkt_dev->odevname, ifname);
3545 pkt_dev->flows = vzalloc_node(MAX_CFLOWS * sizeof(struct flow_state),
3546 node);
3547 if (pkt_dev->flows == NULL) {
3548 kfree(pkt_dev);
3549 return -ENOMEM;
3552 pkt_dev->removal_mark = 0;
3553 pkt_dev->nfrags = 0;
3554 pkt_dev->delay = pg_delay_d;
3555 pkt_dev->count = pg_count_d;
3556 pkt_dev->sofar = 0;
3557 pkt_dev->udp_src_min = 9; /* sink port */
3558 pkt_dev->udp_src_max = 9;
3559 pkt_dev->udp_dst_min = 9;
3560 pkt_dev->udp_dst_max = 9;
3561 pkt_dev->vlan_p = 0;
3562 pkt_dev->vlan_cfi = 0;
3563 pkt_dev->vlan_id = 0xffff;
3564 pkt_dev->svlan_p = 0;
3565 pkt_dev->svlan_cfi = 0;
3566 pkt_dev->svlan_id = 0xffff;
3567 pkt_dev->node = -1;
3569 err = pktgen_setup_dev(t->net, pkt_dev, ifname);
3570 if (err)
3571 goto out1;
3572 if (pkt_dev->odev->priv_flags & IFF_TX_SKB_SHARING)
3573 pkt_dev->clone_skb = pg_clone_skb_d;
3575 pkt_dev->entry = proc_create_data(ifname, 0600, t->net->proc_dir,
3576 &pktgen_if_fops, pkt_dev);
3577 if (!pkt_dev->entry) {
3578 pr_err("cannot create %s/%s procfs entry\n",
3579 PG_PROC_DIR, ifname);
3580 err = -EINVAL;
3581 goto out2;
3583 #ifdef CONFIG_XFRM
3584 pkt_dev->ipsmode = XFRM_MODE_TRANSPORT;
3585 pkt_dev->ipsproto = IPPROTO_ESP;
3587 /* xfrm tunnel mode needs additional dst to extract outter
3588 * ip header protocol/ttl/id field, here creat a phony one.
3589 * instead of looking for a valid rt, which definitely hurting
3590 * performance under such circumstance.
3592 pkt_dev->dstops.family = AF_INET;
3593 pkt_dev->dst.dev = pkt_dev->odev;
3594 dst_init_metrics(&pkt_dev->dst, pktgen_dst_metrics, false);
3595 pkt_dev->dst.child = &pkt_dev->dst;
3596 pkt_dev->dst.ops = &pkt_dev->dstops;
3597 #endif
3599 return add_dev_to_thread(t, pkt_dev);
3600 out2:
3601 dev_put(pkt_dev->odev);
3602 out1:
3603 #ifdef CONFIG_XFRM
3604 free_SAs(pkt_dev);
3605 #endif
3606 vfree(pkt_dev->flows);
3607 kfree(pkt_dev);
3608 return err;
3611 static int __net_init pktgen_create_thread(int cpu, struct pktgen_net *pn)
3613 struct pktgen_thread *t;
3614 struct proc_dir_entry *pe;
3615 struct task_struct *p;
3617 t = kzalloc_node(sizeof(struct pktgen_thread), GFP_KERNEL,
3618 cpu_to_node(cpu));
3619 if (!t) {
3620 pr_err("ERROR: out of memory, can't create new thread\n");
3621 return -ENOMEM;
3624 spin_lock_init(&t->if_lock);
3625 t->cpu = cpu;
3627 INIT_LIST_HEAD(&t->if_list);
3629 list_add_tail(&t->th_list, &pn->pktgen_threads);
3630 init_completion(&t->start_done);
3632 p = kthread_create_on_node(pktgen_thread_worker,
3634 cpu_to_node(cpu),
3635 "kpktgend_%d", cpu);
3636 if (IS_ERR(p)) {
3637 pr_err("kernel_thread() failed for cpu %d\n", t->cpu);
3638 list_del(&t->th_list);
3639 kfree(t);
3640 return PTR_ERR(p);
3642 kthread_bind(p, cpu);
3643 t->tsk = p;
3645 pe = proc_create_data(t->tsk->comm, 0600, pn->proc_dir,
3646 &pktgen_thread_fops, t);
3647 if (!pe) {
3648 pr_err("cannot create %s/%s procfs entry\n",
3649 PG_PROC_DIR, t->tsk->comm);
3650 kthread_stop(p);
3651 list_del(&t->th_list);
3652 kfree(t);
3653 return -EINVAL;
3656 t->net = pn;
3657 wake_up_process(p);
3658 wait_for_completion(&t->start_done);
3660 return 0;
3664 * Removes a device from the thread if_list.
3666 static void _rem_dev_from_if_list(struct pktgen_thread *t,
3667 struct pktgen_dev *pkt_dev)
3669 struct list_head *q, *n;
3670 struct pktgen_dev *p;
3672 list_for_each_safe(q, n, &t->if_list) {
3673 p = list_entry(q, struct pktgen_dev, list);
3674 if (p == pkt_dev)
3675 list_del(&p->list);
3679 static int pktgen_remove_device(struct pktgen_thread *t,
3680 struct pktgen_dev *pkt_dev)
3682 pr_debug("remove_device pkt_dev=%p\n", pkt_dev);
3684 if (pkt_dev->running) {
3685 pr_warning("WARNING: trying to remove a running interface, stopping it now\n");
3686 pktgen_stop_device(pkt_dev);
3689 /* Dis-associate from the interface */
3691 if (pkt_dev->odev) {
3692 dev_put(pkt_dev->odev);
3693 pkt_dev->odev = NULL;
3696 /* And update the thread if_list */
3698 _rem_dev_from_if_list(t, pkt_dev);
3700 if (pkt_dev->entry)
3701 proc_remove(pkt_dev->entry);
3703 #ifdef CONFIG_XFRM
3704 free_SAs(pkt_dev);
3705 #endif
3706 vfree(pkt_dev->flows);
3707 if (pkt_dev->page)
3708 put_page(pkt_dev->page);
3709 kfree(pkt_dev);
3710 return 0;
3713 static int __net_init pg_net_init(struct net *net)
3715 struct pktgen_net *pn = net_generic(net, pg_net_id);
3716 struct proc_dir_entry *pe;
3717 int cpu, ret = 0;
3719 pn->net = net;
3720 INIT_LIST_HEAD(&pn->pktgen_threads);
3721 pn->pktgen_exiting = false;
3722 pn->proc_dir = proc_mkdir(PG_PROC_DIR, pn->net->proc_net);
3723 if (!pn->proc_dir) {
3724 pr_warn("cannot create /proc/net/%s\n", PG_PROC_DIR);
3725 return -ENODEV;
3727 pe = proc_create(PGCTRL, 0600, pn->proc_dir, &pktgen_fops);
3728 if (pe == NULL) {
3729 pr_err("cannot create %s procfs entry\n", PGCTRL);
3730 ret = -EINVAL;
3731 goto remove;
3734 for_each_online_cpu(cpu) {
3735 int err;
3737 err = pktgen_create_thread(cpu, pn);
3738 if (err)
3739 pr_warn("Cannot create thread for cpu %d (%d)\n",
3740 cpu, err);
3743 if (list_empty(&pn->pktgen_threads)) {
3744 pr_err("Initialization failed for all threads\n");
3745 ret = -ENODEV;
3746 goto remove_entry;
3749 return 0;
3751 remove_entry:
3752 remove_proc_entry(PGCTRL, pn->proc_dir);
3753 remove:
3754 remove_proc_entry(PG_PROC_DIR, pn->net->proc_net);
3755 return ret;
3758 static void __net_exit pg_net_exit(struct net *net)
3760 struct pktgen_net *pn = net_generic(net, pg_net_id);
3761 struct pktgen_thread *t;
3762 struct list_head *q, *n;
3763 LIST_HEAD(list);
3765 /* Stop all interfaces & threads */
3766 pn->pktgen_exiting = true;
3768 mutex_lock(&pktgen_thread_lock);
3769 list_splice_init(&pn->pktgen_threads, &list);
3770 mutex_unlock(&pktgen_thread_lock);
3772 list_for_each_safe(q, n, &list) {
3773 t = list_entry(q, struct pktgen_thread, th_list);
3774 list_del(&t->th_list);
3775 kthread_stop(t->tsk);
3776 kfree(t);
3779 remove_proc_entry(PGCTRL, pn->proc_dir);
3780 remove_proc_entry(PG_PROC_DIR, pn->net->proc_net);
3783 static struct pernet_operations pg_net_ops = {
3784 .init = pg_net_init,
3785 .exit = pg_net_exit,
3786 .id = &pg_net_id,
3787 .size = sizeof(struct pktgen_net),
3790 static int __init pg_init(void)
3792 int ret = 0;
3794 pr_info("%s", version);
3795 ret = register_pernet_subsys(&pg_net_ops);
3796 if (ret)
3797 return ret;
3798 ret = register_netdevice_notifier(&pktgen_notifier_block);
3799 if (ret)
3800 unregister_pernet_subsys(&pg_net_ops);
3802 return ret;
3805 static void __exit pg_cleanup(void)
3807 unregister_netdevice_notifier(&pktgen_notifier_block);
3808 unregister_pernet_subsys(&pg_net_ops);
3811 module_init(pg_init);
3812 module_exit(pg_cleanup);
3814 MODULE_AUTHOR("Robert Olsson <robert.olsson@its.uu.se>");
3815 MODULE_DESCRIPTION("Packet Generator tool");
3816 MODULE_LICENSE("GPL");
3817 MODULE_VERSION(VERSION);
3818 module_param(pg_count_d, int, 0);
3819 MODULE_PARM_DESC(pg_count_d, "Default number of packets to inject");
3820 module_param(pg_delay_d, int, 0);
3821 MODULE_PARM_DESC(pg_delay_d, "Default delay between packets (nanoseconds)");
3822 module_param(pg_clone_skb_d, int, 0);
3823 MODULE_PARM_DESC(pg_clone_skb_d, "Default number of copies of the same packet");
3824 module_param(debug, int, 0);
3825 MODULE_PARM_DESC(debug, "Enable debugging of pktgen module");