net: add busy_poll device feature
[linux-2.6/btrfs-unstable.git] / net / core / pktgen.c
blobd0dac57291afab943db9719aa5818ebb6611cf65
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 char data[128];
480 struct pktgen_net *pn = net_generic(current->nsproxy->net_ns, pg_net_id);
482 if (!capable(CAP_NET_ADMIN))
483 return -EPERM;
485 if (count == 0)
486 return -EINVAL;
488 if (count > sizeof(data))
489 count = sizeof(data);
491 if (copy_from_user(data, buf, count))
492 return -EFAULT;
494 data[count - 1] = 0; /* Strip trailing '\n' and terminate string */
496 if (!strcmp(data, "stop"))
497 pktgen_stop_all_threads_ifs(pn);
499 else if (!strcmp(data, "start"))
500 pktgen_run_all_threads(pn);
502 else if (!strcmp(data, "reset"))
503 pktgen_reset_all_threads(pn);
505 else
506 pr_warning("Unknown command: %s\n", data);
508 return count;
511 static int pgctrl_open(struct inode *inode, struct file *file)
513 return single_open(file, pgctrl_show, PDE_DATA(inode));
516 static const struct file_operations pktgen_fops = {
517 .owner = THIS_MODULE,
518 .open = pgctrl_open,
519 .read = seq_read,
520 .llseek = seq_lseek,
521 .write = pgctrl_write,
522 .release = single_release,
525 static int pktgen_if_show(struct seq_file *seq, void *v)
527 const struct pktgen_dev *pkt_dev = seq->private;
528 ktime_t stopped;
529 u64 idle;
531 seq_printf(seq,
532 "Params: count %llu min_pkt_size: %u max_pkt_size: %u\n",
533 (unsigned long long)pkt_dev->count, pkt_dev->min_pkt_size,
534 pkt_dev->max_pkt_size);
536 seq_printf(seq,
537 " frags: %d delay: %llu clone_skb: %d ifname: %s\n",
538 pkt_dev->nfrags, (unsigned long long) pkt_dev->delay,
539 pkt_dev->clone_skb, pkt_dev->odevname);
541 seq_printf(seq, " flows: %u flowlen: %u\n", pkt_dev->cflows,
542 pkt_dev->lflow);
544 seq_printf(seq,
545 " queue_map_min: %u queue_map_max: %u\n",
546 pkt_dev->queue_map_min,
547 pkt_dev->queue_map_max);
549 if (pkt_dev->skb_priority)
550 seq_printf(seq, " skb_priority: %u\n",
551 pkt_dev->skb_priority);
553 if (pkt_dev->flags & F_IPV6) {
554 seq_printf(seq,
555 " saddr: %pI6c min_saddr: %pI6c max_saddr: %pI6c\n"
556 " daddr: %pI6c min_daddr: %pI6c max_daddr: %pI6c\n",
557 &pkt_dev->in6_saddr,
558 &pkt_dev->min_in6_saddr, &pkt_dev->max_in6_saddr,
559 &pkt_dev->in6_daddr,
560 &pkt_dev->min_in6_daddr, &pkt_dev->max_in6_daddr);
561 } else {
562 seq_printf(seq,
563 " dst_min: %s dst_max: %s\n",
564 pkt_dev->dst_min, pkt_dev->dst_max);
565 seq_printf(seq,
566 " src_min: %s src_max: %s\n",
567 pkt_dev->src_min, pkt_dev->src_max);
570 seq_puts(seq, " src_mac: ");
572 seq_printf(seq, "%pM ",
573 is_zero_ether_addr(pkt_dev->src_mac) ?
574 pkt_dev->odev->dev_addr : pkt_dev->src_mac);
576 seq_printf(seq, "dst_mac: ");
577 seq_printf(seq, "%pM\n", pkt_dev->dst_mac);
579 seq_printf(seq,
580 " udp_src_min: %d udp_src_max: %d"
581 " udp_dst_min: %d udp_dst_max: %d\n",
582 pkt_dev->udp_src_min, pkt_dev->udp_src_max,
583 pkt_dev->udp_dst_min, pkt_dev->udp_dst_max);
585 seq_printf(seq,
586 " src_mac_count: %d dst_mac_count: %d\n",
587 pkt_dev->src_mac_count, pkt_dev->dst_mac_count);
589 if (pkt_dev->nr_labels) {
590 unsigned int i;
591 seq_printf(seq, " mpls: ");
592 for (i = 0; i < pkt_dev->nr_labels; i++)
593 seq_printf(seq, "%08x%s", ntohl(pkt_dev->labels[i]),
594 i == pkt_dev->nr_labels-1 ? "\n" : ", ");
597 if (pkt_dev->vlan_id != 0xffff)
598 seq_printf(seq, " vlan_id: %u vlan_p: %u vlan_cfi: %u\n",
599 pkt_dev->vlan_id, pkt_dev->vlan_p,
600 pkt_dev->vlan_cfi);
602 if (pkt_dev->svlan_id != 0xffff)
603 seq_printf(seq, " svlan_id: %u vlan_p: %u vlan_cfi: %u\n",
604 pkt_dev->svlan_id, pkt_dev->svlan_p,
605 pkt_dev->svlan_cfi);
607 if (pkt_dev->tos)
608 seq_printf(seq, " tos: 0x%02x\n", pkt_dev->tos);
610 if (pkt_dev->traffic_class)
611 seq_printf(seq, " traffic_class: 0x%02x\n", pkt_dev->traffic_class);
613 if (pkt_dev->node >= 0)
614 seq_printf(seq, " node: %d\n", pkt_dev->node);
616 seq_printf(seq, " Flags: ");
618 if (pkt_dev->flags & F_IPV6)
619 seq_printf(seq, "IPV6 ");
621 if (pkt_dev->flags & F_IPSRC_RND)
622 seq_printf(seq, "IPSRC_RND ");
624 if (pkt_dev->flags & F_IPDST_RND)
625 seq_printf(seq, "IPDST_RND ");
627 if (pkt_dev->flags & F_TXSIZE_RND)
628 seq_printf(seq, "TXSIZE_RND ");
630 if (pkt_dev->flags & F_UDPSRC_RND)
631 seq_printf(seq, "UDPSRC_RND ");
633 if (pkt_dev->flags & F_UDPDST_RND)
634 seq_printf(seq, "UDPDST_RND ");
636 if (pkt_dev->flags & F_UDPCSUM)
637 seq_printf(seq, "UDPCSUM ");
639 if (pkt_dev->flags & F_MPLS_RND)
640 seq_printf(seq, "MPLS_RND ");
642 if (pkt_dev->flags & F_QUEUE_MAP_RND)
643 seq_printf(seq, "QUEUE_MAP_RND ");
645 if (pkt_dev->flags & F_QUEUE_MAP_CPU)
646 seq_printf(seq, "QUEUE_MAP_CPU ");
648 if (pkt_dev->cflows) {
649 if (pkt_dev->flags & F_FLOW_SEQ)
650 seq_printf(seq, "FLOW_SEQ "); /*in sequence flows*/
651 else
652 seq_printf(seq, "FLOW_RND ");
655 #ifdef CONFIG_XFRM
656 if (pkt_dev->flags & F_IPSEC_ON) {
657 seq_printf(seq, "IPSEC ");
658 if (pkt_dev->spi)
659 seq_printf(seq, "spi:%u", pkt_dev->spi);
661 #endif
663 if (pkt_dev->flags & F_MACSRC_RND)
664 seq_printf(seq, "MACSRC_RND ");
666 if (pkt_dev->flags & F_MACDST_RND)
667 seq_printf(seq, "MACDST_RND ");
669 if (pkt_dev->flags & F_VID_RND)
670 seq_printf(seq, "VID_RND ");
672 if (pkt_dev->flags & F_SVID_RND)
673 seq_printf(seq, "SVID_RND ");
675 if (pkt_dev->flags & F_NODE)
676 seq_printf(seq, "NODE_ALLOC ");
678 seq_puts(seq, "\n");
680 /* not really stopped, more like last-running-at */
681 stopped = pkt_dev->running ? ktime_get() : pkt_dev->stopped_at;
682 idle = pkt_dev->idle_acc;
683 do_div(idle, NSEC_PER_USEC);
685 seq_printf(seq,
686 "Current:\n pkts-sofar: %llu errors: %llu\n",
687 (unsigned long long)pkt_dev->sofar,
688 (unsigned long long)pkt_dev->errors);
690 seq_printf(seq,
691 " started: %lluus stopped: %lluus idle: %lluus\n",
692 (unsigned long long) ktime_to_us(pkt_dev->started_at),
693 (unsigned long long) ktime_to_us(stopped),
694 (unsigned long long) idle);
696 seq_printf(seq,
697 " seq_num: %d cur_dst_mac_offset: %d cur_src_mac_offset: %d\n",
698 pkt_dev->seq_num, pkt_dev->cur_dst_mac_offset,
699 pkt_dev->cur_src_mac_offset);
701 if (pkt_dev->flags & F_IPV6) {
702 seq_printf(seq, " cur_saddr: %pI6c cur_daddr: %pI6c\n",
703 &pkt_dev->cur_in6_saddr,
704 &pkt_dev->cur_in6_daddr);
705 } else
706 seq_printf(seq, " cur_saddr: %pI4 cur_daddr: %pI4\n",
707 &pkt_dev->cur_saddr, &pkt_dev->cur_daddr);
709 seq_printf(seq, " cur_udp_dst: %d cur_udp_src: %d\n",
710 pkt_dev->cur_udp_dst, pkt_dev->cur_udp_src);
712 seq_printf(seq, " cur_queue_map: %u\n", pkt_dev->cur_queue_map);
714 seq_printf(seq, " flows: %u\n", pkt_dev->nflows);
716 if (pkt_dev->result[0])
717 seq_printf(seq, "Result: %s\n", pkt_dev->result);
718 else
719 seq_printf(seq, "Result: Idle\n");
721 return 0;
725 static int hex32_arg(const char __user *user_buffer, unsigned long maxlen,
726 __u32 *num)
728 int i = 0;
729 *num = 0;
731 for (; i < maxlen; i++) {
732 int value;
733 char c;
734 *num <<= 4;
735 if (get_user(c, &user_buffer[i]))
736 return -EFAULT;
737 value = hex_to_bin(c);
738 if (value >= 0)
739 *num |= value;
740 else
741 break;
743 return i;
746 static int count_trail_chars(const char __user * user_buffer,
747 unsigned int maxlen)
749 int i;
751 for (i = 0; i < maxlen; i++) {
752 char c;
753 if (get_user(c, &user_buffer[i]))
754 return -EFAULT;
755 switch (c) {
756 case '\"':
757 case '\n':
758 case '\r':
759 case '\t':
760 case ' ':
761 case '=':
762 break;
763 default:
764 goto done;
767 done:
768 return i;
771 static long num_arg(const char __user *user_buffer, unsigned long maxlen,
772 unsigned long *num)
774 int i;
775 *num = 0;
777 for (i = 0; i < maxlen; i++) {
778 char c;
779 if (get_user(c, &user_buffer[i]))
780 return -EFAULT;
781 if ((c >= '0') && (c <= '9')) {
782 *num *= 10;
783 *num += c - '0';
784 } else
785 break;
787 return i;
790 static int strn_len(const char __user * user_buffer, unsigned int maxlen)
792 int i;
794 for (i = 0; i < maxlen; i++) {
795 char c;
796 if (get_user(c, &user_buffer[i]))
797 return -EFAULT;
798 switch (c) {
799 case '\"':
800 case '\n':
801 case '\r':
802 case '\t':
803 case ' ':
804 goto done_str;
805 break;
806 default:
807 break;
810 done_str:
811 return i;
814 static ssize_t get_labels(const char __user *buffer, struct pktgen_dev *pkt_dev)
816 unsigned int n = 0;
817 char c;
818 ssize_t i = 0;
819 int len;
821 pkt_dev->nr_labels = 0;
822 do {
823 __u32 tmp;
824 len = hex32_arg(&buffer[i], 8, &tmp);
825 if (len <= 0)
826 return len;
827 pkt_dev->labels[n] = htonl(tmp);
828 if (pkt_dev->labels[n] & MPLS_STACK_BOTTOM)
829 pkt_dev->flags |= F_MPLS_RND;
830 i += len;
831 if (get_user(c, &buffer[i]))
832 return -EFAULT;
833 i++;
834 n++;
835 if (n >= MAX_MPLS_LABELS)
836 return -E2BIG;
837 } while (c == ',');
839 pkt_dev->nr_labels = n;
840 return i;
843 static ssize_t pktgen_if_write(struct file *file,
844 const char __user * user_buffer, size_t count,
845 loff_t * offset)
847 struct seq_file *seq = file->private_data;
848 struct pktgen_dev *pkt_dev = seq->private;
849 int i, max, len;
850 char name[16], valstr[32];
851 unsigned long value = 0;
852 char *pg_result = NULL;
853 int tmp = 0;
854 char buf[128];
856 pg_result = &(pkt_dev->result[0]);
858 if (count < 1) {
859 pr_warning("wrong command format\n");
860 return -EINVAL;
863 max = count;
864 tmp = count_trail_chars(user_buffer, max);
865 if (tmp < 0) {
866 pr_warning("illegal format\n");
867 return tmp;
869 i = tmp;
871 /* Read variable name */
873 len = strn_len(&user_buffer[i], sizeof(name) - 1);
874 if (len < 0)
875 return len;
877 memset(name, 0, sizeof(name));
878 if (copy_from_user(name, &user_buffer[i], len))
879 return -EFAULT;
880 i += len;
882 max = count - i;
883 len = count_trail_chars(&user_buffer[i], max);
884 if (len < 0)
885 return len;
887 i += len;
889 if (debug) {
890 size_t copy = min_t(size_t, count, 1023);
891 char tb[copy + 1];
892 if (copy_from_user(tb, user_buffer, copy))
893 return -EFAULT;
894 tb[copy] = 0;
895 pr_debug("%s,%lu buffer -:%s:-\n",
896 name, (unsigned long)count, tb);
899 if (!strcmp(name, "min_pkt_size")) {
900 len = num_arg(&user_buffer[i], 10, &value);
901 if (len < 0)
902 return len;
904 i += len;
905 if (value < 14 + 20 + 8)
906 value = 14 + 20 + 8;
907 if (value != pkt_dev->min_pkt_size) {
908 pkt_dev->min_pkt_size = value;
909 pkt_dev->cur_pkt_size = value;
911 sprintf(pg_result, "OK: min_pkt_size=%u",
912 pkt_dev->min_pkt_size);
913 return count;
916 if (!strcmp(name, "max_pkt_size")) {
917 len = num_arg(&user_buffer[i], 10, &value);
918 if (len < 0)
919 return len;
921 i += len;
922 if (value < 14 + 20 + 8)
923 value = 14 + 20 + 8;
924 if (value != pkt_dev->max_pkt_size) {
925 pkt_dev->max_pkt_size = value;
926 pkt_dev->cur_pkt_size = value;
928 sprintf(pg_result, "OK: max_pkt_size=%u",
929 pkt_dev->max_pkt_size);
930 return count;
933 /* Shortcut for min = max */
935 if (!strcmp(name, "pkt_size")) {
936 len = num_arg(&user_buffer[i], 10, &value);
937 if (len < 0)
938 return len;
940 i += len;
941 if (value < 14 + 20 + 8)
942 value = 14 + 20 + 8;
943 if (value != pkt_dev->min_pkt_size) {
944 pkt_dev->min_pkt_size = value;
945 pkt_dev->max_pkt_size = value;
946 pkt_dev->cur_pkt_size = value;
948 sprintf(pg_result, "OK: pkt_size=%u", pkt_dev->min_pkt_size);
949 return count;
952 if (!strcmp(name, "debug")) {
953 len = num_arg(&user_buffer[i], 10, &value);
954 if (len < 0)
955 return len;
957 i += len;
958 debug = value;
959 sprintf(pg_result, "OK: debug=%u", debug);
960 return count;
963 if (!strcmp(name, "frags")) {
964 len = num_arg(&user_buffer[i], 10, &value);
965 if (len < 0)
966 return len;
968 i += len;
969 pkt_dev->nfrags = value;
970 sprintf(pg_result, "OK: frags=%u", pkt_dev->nfrags);
971 return count;
973 if (!strcmp(name, "delay")) {
974 len = num_arg(&user_buffer[i], 10, &value);
975 if (len < 0)
976 return len;
978 i += len;
979 if (value == 0x7FFFFFFF)
980 pkt_dev->delay = ULLONG_MAX;
981 else
982 pkt_dev->delay = (u64)value;
984 sprintf(pg_result, "OK: delay=%llu",
985 (unsigned long long) pkt_dev->delay);
986 return count;
988 if (!strcmp(name, "rate")) {
989 len = num_arg(&user_buffer[i], 10, &value);
990 if (len < 0)
991 return len;
993 i += len;
994 if (!value)
995 return len;
996 pkt_dev->delay = pkt_dev->min_pkt_size*8*NSEC_PER_USEC/value;
997 if (debug)
998 pr_info("Delay set at: %llu ns\n", pkt_dev->delay);
1000 sprintf(pg_result, "OK: rate=%lu", value);
1001 return count;
1003 if (!strcmp(name, "ratep")) {
1004 len = num_arg(&user_buffer[i], 10, &value);
1005 if (len < 0)
1006 return len;
1008 i += len;
1009 if (!value)
1010 return len;
1011 pkt_dev->delay = NSEC_PER_SEC/value;
1012 if (debug)
1013 pr_info("Delay set at: %llu ns\n", pkt_dev->delay);
1015 sprintf(pg_result, "OK: rate=%lu", value);
1016 return count;
1018 if (!strcmp(name, "udp_src_min")) {
1019 len = num_arg(&user_buffer[i], 10, &value);
1020 if (len < 0)
1021 return len;
1023 i += len;
1024 if (value != pkt_dev->udp_src_min) {
1025 pkt_dev->udp_src_min = value;
1026 pkt_dev->cur_udp_src = value;
1028 sprintf(pg_result, "OK: udp_src_min=%u", pkt_dev->udp_src_min);
1029 return count;
1031 if (!strcmp(name, "udp_dst_min")) {
1032 len = num_arg(&user_buffer[i], 10, &value);
1033 if (len < 0)
1034 return len;
1036 i += len;
1037 if (value != pkt_dev->udp_dst_min) {
1038 pkt_dev->udp_dst_min = value;
1039 pkt_dev->cur_udp_dst = value;
1041 sprintf(pg_result, "OK: udp_dst_min=%u", pkt_dev->udp_dst_min);
1042 return count;
1044 if (!strcmp(name, "udp_src_max")) {
1045 len = num_arg(&user_buffer[i], 10, &value);
1046 if (len < 0)
1047 return len;
1049 i += len;
1050 if (value != pkt_dev->udp_src_max) {
1051 pkt_dev->udp_src_max = value;
1052 pkt_dev->cur_udp_src = value;
1054 sprintf(pg_result, "OK: udp_src_max=%u", pkt_dev->udp_src_max);
1055 return count;
1057 if (!strcmp(name, "udp_dst_max")) {
1058 len = num_arg(&user_buffer[i], 10, &value);
1059 if (len < 0)
1060 return len;
1062 i += len;
1063 if (value != pkt_dev->udp_dst_max) {
1064 pkt_dev->udp_dst_max = value;
1065 pkt_dev->cur_udp_dst = value;
1067 sprintf(pg_result, "OK: udp_dst_max=%u", pkt_dev->udp_dst_max);
1068 return count;
1070 if (!strcmp(name, "clone_skb")) {
1071 len = num_arg(&user_buffer[i], 10, &value);
1072 if (len < 0)
1073 return len;
1074 if ((value > 0) &&
1075 (!(pkt_dev->odev->priv_flags & IFF_TX_SKB_SHARING)))
1076 return -ENOTSUPP;
1077 i += len;
1078 pkt_dev->clone_skb = value;
1080 sprintf(pg_result, "OK: clone_skb=%d", pkt_dev->clone_skb);
1081 return count;
1083 if (!strcmp(name, "count")) {
1084 len = num_arg(&user_buffer[i], 10, &value);
1085 if (len < 0)
1086 return len;
1088 i += len;
1089 pkt_dev->count = value;
1090 sprintf(pg_result, "OK: count=%llu",
1091 (unsigned long long)pkt_dev->count);
1092 return count;
1094 if (!strcmp(name, "src_mac_count")) {
1095 len = num_arg(&user_buffer[i], 10, &value);
1096 if (len < 0)
1097 return len;
1099 i += len;
1100 if (pkt_dev->src_mac_count != value) {
1101 pkt_dev->src_mac_count = value;
1102 pkt_dev->cur_src_mac_offset = 0;
1104 sprintf(pg_result, "OK: src_mac_count=%d",
1105 pkt_dev->src_mac_count);
1106 return count;
1108 if (!strcmp(name, "dst_mac_count")) {
1109 len = num_arg(&user_buffer[i], 10, &value);
1110 if (len < 0)
1111 return len;
1113 i += len;
1114 if (pkt_dev->dst_mac_count != value) {
1115 pkt_dev->dst_mac_count = value;
1116 pkt_dev->cur_dst_mac_offset = 0;
1118 sprintf(pg_result, "OK: dst_mac_count=%d",
1119 pkt_dev->dst_mac_count);
1120 return count;
1122 if (!strcmp(name, "node")) {
1123 len = num_arg(&user_buffer[i], 10, &value);
1124 if (len < 0)
1125 return len;
1127 i += len;
1129 if (node_possible(value)) {
1130 pkt_dev->node = value;
1131 sprintf(pg_result, "OK: node=%d", pkt_dev->node);
1132 if (pkt_dev->page) {
1133 put_page(pkt_dev->page);
1134 pkt_dev->page = NULL;
1137 else
1138 sprintf(pg_result, "ERROR: node not possible");
1139 return count;
1141 if (!strcmp(name, "flag")) {
1142 char f[32];
1143 memset(f, 0, 32);
1144 len = strn_len(&user_buffer[i], sizeof(f) - 1);
1145 if (len < 0)
1146 return len;
1148 if (copy_from_user(f, &user_buffer[i], len))
1149 return -EFAULT;
1150 i += len;
1151 if (strcmp(f, "IPSRC_RND") == 0)
1152 pkt_dev->flags |= F_IPSRC_RND;
1154 else if (strcmp(f, "!IPSRC_RND") == 0)
1155 pkt_dev->flags &= ~F_IPSRC_RND;
1157 else if (strcmp(f, "TXSIZE_RND") == 0)
1158 pkt_dev->flags |= F_TXSIZE_RND;
1160 else if (strcmp(f, "!TXSIZE_RND") == 0)
1161 pkt_dev->flags &= ~F_TXSIZE_RND;
1163 else if (strcmp(f, "IPDST_RND") == 0)
1164 pkt_dev->flags |= F_IPDST_RND;
1166 else if (strcmp(f, "!IPDST_RND") == 0)
1167 pkt_dev->flags &= ~F_IPDST_RND;
1169 else if (strcmp(f, "UDPSRC_RND") == 0)
1170 pkt_dev->flags |= F_UDPSRC_RND;
1172 else if (strcmp(f, "!UDPSRC_RND") == 0)
1173 pkt_dev->flags &= ~F_UDPSRC_RND;
1175 else if (strcmp(f, "UDPDST_RND") == 0)
1176 pkt_dev->flags |= F_UDPDST_RND;
1178 else if (strcmp(f, "!UDPDST_RND") == 0)
1179 pkt_dev->flags &= ~F_UDPDST_RND;
1181 else if (strcmp(f, "MACSRC_RND") == 0)
1182 pkt_dev->flags |= F_MACSRC_RND;
1184 else if (strcmp(f, "!MACSRC_RND") == 0)
1185 pkt_dev->flags &= ~F_MACSRC_RND;
1187 else if (strcmp(f, "MACDST_RND") == 0)
1188 pkt_dev->flags |= F_MACDST_RND;
1190 else if (strcmp(f, "!MACDST_RND") == 0)
1191 pkt_dev->flags &= ~F_MACDST_RND;
1193 else if (strcmp(f, "MPLS_RND") == 0)
1194 pkt_dev->flags |= F_MPLS_RND;
1196 else if (strcmp(f, "!MPLS_RND") == 0)
1197 pkt_dev->flags &= ~F_MPLS_RND;
1199 else if (strcmp(f, "VID_RND") == 0)
1200 pkt_dev->flags |= F_VID_RND;
1202 else if (strcmp(f, "!VID_RND") == 0)
1203 pkt_dev->flags &= ~F_VID_RND;
1205 else if (strcmp(f, "SVID_RND") == 0)
1206 pkt_dev->flags |= F_SVID_RND;
1208 else if (strcmp(f, "!SVID_RND") == 0)
1209 pkt_dev->flags &= ~F_SVID_RND;
1211 else if (strcmp(f, "FLOW_SEQ") == 0)
1212 pkt_dev->flags |= F_FLOW_SEQ;
1214 else if (strcmp(f, "QUEUE_MAP_RND") == 0)
1215 pkt_dev->flags |= F_QUEUE_MAP_RND;
1217 else if (strcmp(f, "!QUEUE_MAP_RND") == 0)
1218 pkt_dev->flags &= ~F_QUEUE_MAP_RND;
1220 else if (strcmp(f, "QUEUE_MAP_CPU") == 0)
1221 pkt_dev->flags |= F_QUEUE_MAP_CPU;
1223 else if (strcmp(f, "!QUEUE_MAP_CPU") == 0)
1224 pkt_dev->flags &= ~F_QUEUE_MAP_CPU;
1225 #ifdef CONFIG_XFRM
1226 else if (strcmp(f, "IPSEC") == 0)
1227 pkt_dev->flags |= F_IPSEC_ON;
1228 #endif
1230 else if (strcmp(f, "!IPV6") == 0)
1231 pkt_dev->flags &= ~F_IPV6;
1233 else if (strcmp(f, "NODE_ALLOC") == 0)
1234 pkt_dev->flags |= F_NODE;
1236 else if (strcmp(f, "!NODE_ALLOC") == 0)
1237 pkt_dev->flags &= ~F_NODE;
1239 else if (strcmp(f, "UDPCSUM") == 0)
1240 pkt_dev->flags |= F_UDPCSUM;
1242 else if (strcmp(f, "!UDPCSUM") == 0)
1243 pkt_dev->flags &= ~F_UDPCSUM;
1245 else {
1246 sprintf(pg_result,
1247 "Flag -:%s:- unknown\nAvailable flags, (prepend ! to un-set flag):\n%s",
1249 "IPSRC_RND, IPDST_RND, UDPSRC_RND, UDPDST_RND, "
1250 "MACSRC_RND, MACDST_RND, TXSIZE_RND, IPV6, "
1251 "MPLS_RND, VID_RND, SVID_RND, FLOW_SEQ, "
1252 "QUEUE_MAP_RND, QUEUE_MAP_CPU, UDPCSUM, "
1253 #ifdef CONFIG_XFRM
1254 "IPSEC, "
1255 #endif
1256 "NODE_ALLOC\n");
1257 return count;
1259 sprintf(pg_result, "OK: flags=0x%x", pkt_dev->flags);
1260 return count;
1262 if (!strcmp(name, "dst_min") || !strcmp(name, "dst")) {
1263 len = strn_len(&user_buffer[i], sizeof(pkt_dev->dst_min) - 1);
1264 if (len < 0)
1265 return len;
1267 if (copy_from_user(buf, &user_buffer[i], len))
1268 return -EFAULT;
1269 buf[len] = 0;
1270 if (strcmp(buf, pkt_dev->dst_min) != 0) {
1271 memset(pkt_dev->dst_min, 0, sizeof(pkt_dev->dst_min));
1272 strncpy(pkt_dev->dst_min, buf, len);
1273 pkt_dev->daddr_min = in_aton(pkt_dev->dst_min);
1274 pkt_dev->cur_daddr = pkt_dev->daddr_min;
1276 if (debug)
1277 pr_debug("dst_min set to: %s\n", pkt_dev->dst_min);
1278 i += len;
1279 sprintf(pg_result, "OK: dst_min=%s", pkt_dev->dst_min);
1280 return count;
1282 if (!strcmp(name, "dst_max")) {
1283 len = strn_len(&user_buffer[i], sizeof(pkt_dev->dst_max) - 1);
1284 if (len < 0)
1285 return len;
1288 if (copy_from_user(buf, &user_buffer[i], len))
1289 return -EFAULT;
1291 buf[len] = 0;
1292 if (strcmp(buf, pkt_dev->dst_max) != 0) {
1293 memset(pkt_dev->dst_max, 0, sizeof(pkt_dev->dst_max));
1294 strncpy(pkt_dev->dst_max, buf, len);
1295 pkt_dev->daddr_max = in_aton(pkt_dev->dst_max);
1296 pkt_dev->cur_daddr = pkt_dev->daddr_max;
1298 if (debug)
1299 pr_debug("dst_max set to: %s\n", pkt_dev->dst_max);
1300 i += len;
1301 sprintf(pg_result, "OK: dst_max=%s", pkt_dev->dst_max);
1302 return count;
1304 if (!strcmp(name, "dst6")) {
1305 len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1306 if (len < 0)
1307 return len;
1309 pkt_dev->flags |= F_IPV6;
1311 if (copy_from_user(buf, &user_buffer[i], len))
1312 return -EFAULT;
1313 buf[len] = 0;
1315 in6_pton(buf, -1, pkt_dev->in6_daddr.s6_addr, -1, NULL);
1316 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->in6_daddr);
1318 pkt_dev->cur_in6_daddr = pkt_dev->in6_daddr;
1320 if (debug)
1321 pr_debug("dst6 set to: %s\n", buf);
1323 i += len;
1324 sprintf(pg_result, "OK: dst6=%s", buf);
1325 return count;
1327 if (!strcmp(name, "dst6_min")) {
1328 len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1329 if (len < 0)
1330 return len;
1332 pkt_dev->flags |= F_IPV6;
1334 if (copy_from_user(buf, &user_buffer[i], len))
1335 return -EFAULT;
1336 buf[len] = 0;
1338 in6_pton(buf, -1, pkt_dev->min_in6_daddr.s6_addr, -1, NULL);
1339 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->min_in6_daddr);
1341 pkt_dev->cur_in6_daddr = pkt_dev->min_in6_daddr;
1342 if (debug)
1343 pr_debug("dst6_min set to: %s\n", buf);
1345 i += len;
1346 sprintf(pg_result, "OK: dst6_min=%s", buf);
1347 return count;
1349 if (!strcmp(name, "dst6_max")) {
1350 len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1351 if (len < 0)
1352 return len;
1354 pkt_dev->flags |= F_IPV6;
1356 if (copy_from_user(buf, &user_buffer[i], len))
1357 return -EFAULT;
1358 buf[len] = 0;
1360 in6_pton(buf, -1, pkt_dev->max_in6_daddr.s6_addr, -1, NULL);
1361 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->max_in6_daddr);
1363 if (debug)
1364 pr_debug("dst6_max set to: %s\n", buf);
1366 i += len;
1367 sprintf(pg_result, "OK: dst6_max=%s", buf);
1368 return count;
1370 if (!strcmp(name, "src6")) {
1371 len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1372 if (len < 0)
1373 return len;
1375 pkt_dev->flags |= F_IPV6;
1377 if (copy_from_user(buf, &user_buffer[i], len))
1378 return -EFAULT;
1379 buf[len] = 0;
1381 in6_pton(buf, -1, pkt_dev->in6_saddr.s6_addr, -1, NULL);
1382 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->in6_saddr);
1384 pkt_dev->cur_in6_saddr = pkt_dev->in6_saddr;
1386 if (debug)
1387 pr_debug("src6 set to: %s\n", buf);
1389 i += len;
1390 sprintf(pg_result, "OK: src6=%s", buf);
1391 return count;
1393 if (!strcmp(name, "src_min")) {
1394 len = strn_len(&user_buffer[i], sizeof(pkt_dev->src_min) - 1);
1395 if (len < 0)
1396 return len;
1398 if (copy_from_user(buf, &user_buffer[i], len))
1399 return -EFAULT;
1400 buf[len] = 0;
1401 if (strcmp(buf, pkt_dev->src_min) != 0) {
1402 memset(pkt_dev->src_min, 0, sizeof(pkt_dev->src_min));
1403 strncpy(pkt_dev->src_min, buf, len);
1404 pkt_dev->saddr_min = in_aton(pkt_dev->src_min);
1405 pkt_dev->cur_saddr = pkt_dev->saddr_min;
1407 if (debug)
1408 pr_debug("src_min set to: %s\n", pkt_dev->src_min);
1409 i += len;
1410 sprintf(pg_result, "OK: src_min=%s", pkt_dev->src_min);
1411 return count;
1413 if (!strcmp(name, "src_max")) {
1414 len = strn_len(&user_buffer[i], sizeof(pkt_dev->src_max) - 1);
1415 if (len < 0)
1416 return len;
1418 if (copy_from_user(buf, &user_buffer[i], len))
1419 return -EFAULT;
1420 buf[len] = 0;
1421 if (strcmp(buf, pkt_dev->src_max) != 0) {
1422 memset(pkt_dev->src_max, 0, sizeof(pkt_dev->src_max));
1423 strncpy(pkt_dev->src_max, buf, len);
1424 pkt_dev->saddr_max = in_aton(pkt_dev->src_max);
1425 pkt_dev->cur_saddr = pkt_dev->saddr_max;
1427 if (debug)
1428 pr_debug("src_max set to: %s\n", pkt_dev->src_max);
1429 i += len;
1430 sprintf(pg_result, "OK: src_max=%s", pkt_dev->src_max);
1431 return count;
1433 if (!strcmp(name, "dst_mac")) {
1434 len = strn_len(&user_buffer[i], sizeof(valstr) - 1);
1435 if (len < 0)
1436 return len;
1438 memset(valstr, 0, sizeof(valstr));
1439 if (copy_from_user(valstr, &user_buffer[i], len))
1440 return -EFAULT;
1442 if (!mac_pton(valstr, pkt_dev->dst_mac))
1443 return -EINVAL;
1444 /* Set up Dest MAC */
1445 ether_addr_copy(&pkt_dev->hh[0], pkt_dev->dst_mac);
1447 sprintf(pg_result, "OK: dstmac %pM", pkt_dev->dst_mac);
1448 return count;
1450 if (!strcmp(name, "src_mac")) {
1451 len = strn_len(&user_buffer[i], sizeof(valstr) - 1);
1452 if (len < 0)
1453 return len;
1455 memset(valstr, 0, sizeof(valstr));
1456 if (copy_from_user(valstr, &user_buffer[i], len))
1457 return -EFAULT;
1459 if (!mac_pton(valstr, pkt_dev->src_mac))
1460 return -EINVAL;
1461 /* Set up Src MAC */
1462 ether_addr_copy(&pkt_dev->hh[6], pkt_dev->src_mac);
1464 sprintf(pg_result, "OK: srcmac %pM", pkt_dev->src_mac);
1465 return count;
1468 if (!strcmp(name, "clear_counters")) {
1469 pktgen_clear_counters(pkt_dev);
1470 sprintf(pg_result, "OK: Clearing counters.\n");
1471 return count;
1474 if (!strcmp(name, "flows")) {
1475 len = num_arg(&user_buffer[i], 10, &value);
1476 if (len < 0)
1477 return len;
1479 i += len;
1480 if (value > MAX_CFLOWS)
1481 value = MAX_CFLOWS;
1483 pkt_dev->cflows = value;
1484 sprintf(pg_result, "OK: flows=%u", pkt_dev->cflows);
1485 return count;
1487 #ifdef CONFIG_XFRM
1488 if (!strcmp(name, "spi")) {
1489 len = num_arg(&user_buffer[i], 10, &value);
1490 if (len < 0)
1491 return len;
1493 i += len;
1494 pkt_dev->spi = value;
1495 sprintf(pg_result, "OK: spi=%u", pkt_dev->spi);
1496 return count;
1498 #endif
1499 if (!strcmp(name, "flowlen")) {
1500 len = num_arg(&user_buffer[i], 10, &value);
1501 if (len < 0)
1502 return len;
1504 i += len;
1505 pkt_dev->lflow = value;
1506 sprintf(pg_result, "OK: flowlen=%u", pkt_dev->lflow);
1507 return count;
1510 if (!strcmp(name, "queue_map_min")) {
1511 len = num_arg(&user_buffer[i], 5, &value);
1512 if (len < 0)
1513 return len;
1515 i += len;
1516 pkt_dev->queue_map_min = value;
1517 sprintf(pg_result, "OK: queue_map_min=%u", pkt_dev->queue_map_min);
1518 return count;
1521 if (!strcmp(name, "queue_map_max")) {
1522 len = num_arg(&user_buffer[i], 5, &value);
1523 if (len < 0)
1524 return len;
1526 i += len;
1527 pkt_dev->queue_map_max = value;
1528 sprintf(pg_result, "OK: queue_map_max=%u", pkt_dev->queue_map_max);
1529 return count;
1532 if (!strcmp(name, "mpls")) {
1533 unsigned int n, cnt;
1535 len = get_labels(&user_buffer[i], pkt_dev);
1536 if (len < 0)
1537 return len;
1538 i += len;
1539 cnt = sprintf(pg_result, "OK: mpls=");
1540 for (n = 0; n < pkt_dev->nr_labels; n++)
1541 cnt += sprintf(pg_result + cnt,
1542 "%08x%s", ntohl(pkt_dev->labels[n]),
1543 n == pkt_dev->nr_labels-1 ? "" : ",");
1545 if (pkt_dev->nr_labels && pkt_dev->vlan_id != 0xffff) {
1546 pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1547 pkt_dev->svlan_id = 0xffff;
1549 if (debug)
1550 pr_debug("VLAN/SVLAN auto turned off\n");
1552 return count;
1555 if (!strcmp(name, "vlan_id")) {
1556 len = num_arg(&user_buffer[i], 4, &value);
1557 if (len < 0)
1558 return len;
1560 i += len;
1561 if (value <= 4095) {
1562 pkt_dev->vlan_id = value; /* turn on VLAN */
1564 if (debug)
1565 pr_debug("VLAN turned on\n");
1567 if (debug && pkt_dev->nr_labels)
1568 pr_debug("MPLS auto turned off\n");
1570 pkt_dev->nr_labels = 0; /* turn off MPLS */
1571 sprintf(pg_result, "OK: vlan_id=%u", pkt_dev->vlan_id);
1572 } else {
1573 pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1574 pkt_dev->svlan_id = 0xffff;
1576 if (debug)
1577 pr_debug("VLAN/SVLAN turned off\n");
1579 return count;
1582 if (!strcmp(name, "vlan_p")) {
1583 len = num_arg(&user_buffer[i], 1, &value);
1584 if (len < 0)
1585 return len;
1587 i += len;
1588 if ((value <= 7) && (pkt_dev->vlan_id != 0xffff)) {
1589 pkt_dev->vlan_p = value;
1590 sprintf(pg_result, "OK: vlan_p=%u", pkt_dev->vlan_p);
1591 } else {
1592 sprintf(pg_result, "ERROR: vlan_p must be 0-7");
1594 return count;
1597 if (!strcmp(name, "vlan_cfi")) {
1598 len = num_arg(&user_buffer[i], 1, &value);
1599 if (len < 0)
1600 return len;
1602 i += len;
1603 if ((value <= 1) && (pkt_dev->vlan_id != 0xffff)) {
1604 pkt_dev->vlan_cfi = value;
1605 sprintf(pg_result, "OK: vlan_cfi=%u", pkt_dev->vlan_cfi);
1606 } else {
1607 sprintf(pg_result, "ERROR: vlan_cfi must be 0-1");
1609 return count;
1612 if (!strcmp(name, "svlan_id")) {
1613 len = num_arg(&user_buffer[i], 4, &value);
1614 if (len < 0)
1615 return len;
1617 i += len;
1618 if ((value <= 4095) && ((pkt_dev->vlan_id != 0xffff))) {
1619 pkt_dev->svlan_id = value; /* turn on SVLAN */
1621 if (debug)
1622 pr_debug("SVLAN turned on\n");
1624 if (debug && pkt_dev->nr_labels)
1625 pr_debug("MPLS auto turned off\n");
1627 pkt_dev->nr_labels = 0; /* turn off MPLS */
1628 sprintf(pg_result, "OK: svlan_id=%u", pkt_dev->svlan_id);
1629 } else {
1630 pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1631 pkt_dev->svlan_id = 0xffff;
1633 if (debug)
1634 pr_debug("VLAN/SVLAN turned off\n");
1636 return count;
1639 if (!strcmp(name, "svlan_p")) {
1640 len = num_arg(&user_buffer[i], 1, &value);
1641 if (len < 0)
1642 return len;
1644 i += len;
1645 if ((value <= 7) && (pkt_dev->svlan_id != 0xffff)) {
1646 pkt_dev->svlan_p = value;
1647 sprintf(pg_result, "OK: svlan_p=%u", pkt_dev->svlan_p);
1648 } else {
1649 sprintf(pg_result, "ERROR: svlan_p must be 0-7");
1651 return count;
1654 if (!strcmp(name, "svlan_cfi")) {
1655 len = num_arg(&user_buffer[i], 1, &value);
1656 if (len < 0)
1657 return len;
1659 i += len;
1660 if ((value <= 1) && (pkt_dev->svlan_id != 0xffff)) {
1661 pkt_dev->svlan_cfi = value;
1662 sprintf(pg_result, "OK: svlan_cfi=%u", pkt_dev->svlan_cfi);
1663 } else {
1664 sprintf(pg_result, "ERROR: svlan_cfi must be 0-1");
1666 return count;
1669 if (!strcmp(name, "tos")) {
1670 __u32 tmp_value = 0;
1671 len = hex32_arg(&user_buffer[i], 2, &tmp_value);
1672 if (len < 0)
1673 return len;
1675 i += len;
1676 if (len == 2) {
1677 pkt_dev->tos = tmp_value;
1678 sprintf(pg_result, "OK: tos=0x%02x", pkt_dev->tos);
1679 } else {
1680 sprintf(pg_result, "ERROR: tos must be 00-ff");
1682 return count;
1685 if (!strcmp(name, "traffic_class")) {
1686 __u32 tmp_value = 0;
1687 len = hex32_arg(&user_buffer[i], 2, &tmp_value);
1688 if (len < 0)
1689 return len;
1691 i += len;
1692 if (len == 2) {
1693 pkt_dev->traffic_class = tmp_value;
1694 sprintf(pg_result, "OK: traffic_class=0x%02x", pkt_dev->traffic_class);
1695 } else {
1696 sprintf(pg_result, "ERROR: traffic_class must be 00-ff");
1698 return count;
1701 if (!strcmp(name, "skb_priority")) {
1702 len = num_arg(&user_buffer[i], 9, &value);
1703 if (len < 0)
1704 return len;
1706 i += len;
1707 pkt_dev->skb_priority = value;
1708 sprintf(pg_result, "OK: skb_priority=%i",
1709 pkt_dev->skb_priority);
1710 return count;
1713 sprintf(pkt_dev->result, "No such parameter \"%s\"", name);
1714 return -EINVAL;
1717 static int pktgen_if_open(struct inode *inode, struct file *file)
1719 return single_open(file, pktgen_if_show, PDE_DATA(inode));
1722 static const struct file_operations pktgen_if_fops = {
1723 .owner = THIS_MODULE,
1724 .open = pktgen_if_open,
1725 .read = seq_read,
1726 .llseek = seq_lseek,
1727 .write = pktgen_if_write,
1728 .release = single_release,
1731 static int pktgen_thread_show(struct seq_file *seq, void *v)
1733 struct pktgen_thread *t = seq->private;
1734 const struct pktgen_dev *pkt_dev;
1736 BUG_ON(!t);
1738 seq_printf(seq, "Running: ");
1740 if_lock(t);
1741 list_for_each_entry(pkt_dev, &t->if_list, list)
1742 if (pkt_dev->running)
1743 seq_printf(seq, "%s ", pkt_dev->odevname);
1745 seq_printf(seq, "\nStopped: ");
1747 list_for_each_entry(pkt_dev, &t->if_list, list)
1748 if (!pkt_dev->running)
1749 seq_printf(seq, "%s ", pkt_dev->odevname);
1751 if (t->result[0])
1752 seq_printf(seq, "\nResult: %s\n", t->result);
1753 else
1754 seq_printf(seq, "\nResult: NA\n");
1756 if_unlock(t);
1758 return 0;
1761 static ssize_t pktgen_thread_write(struct file *file,
1762 const char __user * user_buffer,
1763 size_t count, loff_t * offset)
1765 struct seq_file *seq = file->private_data;
1766 struct pktgen_thread *t = seq->private;
1767 int i, max, len, ret;
1768 char name[40];
1769 char *pg_result;
1771 if (count < 1) {
1772 // sprintf(pg_result, "Wrong command format");
1773 return -EINVAL;
1776 max = count;
1777 len = count_trail_chars(user_buffer, max);
1778 if (len < 0)
1779 return len;
1781 i = len;
1783 /* Read variable name */
1785 len = strn_len(&user_buffer[i], sizeof(name) - 1);
1786 if (len < 0)
1787 return len;
1789 memset(name, 0, sizeof(name));
1790 if (copy_from_user(name, &user_buffer[i], len))
1791 return -EFAULT;
1792 i += len;
1794 max = count - i;
1795 len = count_trail_chars(&user_buffer[i], max);
1796 if (len < 0)
1797 return len;
1799 i += len;
1801 if (debug)
1802 pr_debug("t=%s, count=%lu\n", name, (unsigned long)count);
1804 if (!t) {
1805 pr_err("ERROR: No thread\n");
1806 ret = -EINVAL;
1807 goto out;
1810 pg_result = &(t->result[0]);
1812 if (!strcmp(name, "add_device")) {
1813 char f[32];
1814 memset(f, 0, 32);
1815 len = strn_len(&user_buffer[i], sizeof(f) - 1);
1816 if (len < 0) {
1817 ret = len;
1818 goto out;
1820 if (copy_from_user(f, &user_buffer[i], len))
1821 return -EFAULT;
1822 i += len;
1823 mutex_lock(&pktgen_thread_lock);
1824 ret = pktgen_add_device(t, f);
1825 mutex_unlock(&pktgen_thread_lock);
1826 if (!ret) {
1827 ret = count;
1828 sprintf(pg_result, "OK: add_device=%s", f);
1829 } else
1830 sprintf(pg_result, "ERROR: can not add device %s", f);
1831 goto out;
1834 if (!strcmp(name, "rem_device_all")) {
1835 mutex_lock(&pktgen_thread_lock);
1836 t->control |= T_REMDEVALL;
1837 mutex_unlock(&pktgen_thread_lock);
1838 schedule_timeout_interruptible(msecs_to_jiffies(125)); /* Propagate thread->control */
1839 ret = count;
1840 sprintf(pg_result, "OK: rem_device_all");
1841 goto out;
1844 if (!strcmp(name, "max_before_softirq")) {
1845 sprintf(pg_result, "OK: Note! max_before_softirq is obsoleted -- Do not use");
1846 ret = count;
1847 goto out;
1850 ret = -EINVAL;
1851 out:
1852 return ret;
1855 static int pktgen_thread_open(struct inode *inode, struct file *file)
1857 return single_open(file, pktgen_thread_show, PDE_DATA(inode));
1860 static const struct file_operations pktgen_thread_fops = {
1861 .owner = THIS_MODULE,
1862 .open = pktgen_thread_open,
1863 .read = seq_read,
1864 .llseek = seq_lseek,
1865 .write = pktgen_thread_write,
1866 .release = single_release,
1869 /* Think find or remove for NN */
1870 static struct pktgen_dev *__pktgen_NN_threads(const struct pktgen_net *pn,
1871 const char *ifname, int remove)
1873 struct pktgen_thread *t;
1874 struct pktgen_dev *pkt_dev = NULL;
1875 bool exact = (remove == FIND);
1877 list_for_each_entry(t, &pn->pktgen_threads, th_list) {
1878 pkt_dev = pktgen_find_dev(t, ifname, exact);
1879 if (pkt_dev) {
1880 if (remove) {
1881 if_lock(t);
1882 pkt_dev->removal_mark = 1;
1883 t->control |= T_REMDEV;
1884 if_unlock(t);
1886 break;
1889 return pkt_dev;
1893 * mark a device for removal
1895 static void pktgen_mark_device(const struct pktgen_net *pn, const char *ifname)
1897 struct pktgen_dev *pkt_dev = NULL;
1898 const int max_tries = 10, msec_per_try = 125;
1899 int i = 0;
1901 mutex_lock(&pktgen_thread_lock);
1902 pr_debug("%s: marking %s for removal\n", __func__, ifname);
1904 while (1) {
1906 pkt_dev = __pktgen_NN_threads(pn, ifname, REMOVE);
1907 if (pkt_dev == NULL)
1908 break; /* success */
1910 mutex_unlock(&pktgen_thread_lock);
1911 pr_debug("%s: waiting for %s to disappear....\n",
1912 __func__, ifname);
1913 schedule_timeout_interruptible(msecs_to_jiffies(msec_per_try));
1914 mutex_lock(&pktgen_thread_lock);
1916 if (++i >= max_tries) {
1917 pr_err("%s: timed out after waiting %d msec for device %s to be removed\n",
1918 __func__, msec_per_try * i, ifname);
1919 break;
1924 mutex_unlock(&pktgen_thread_lock);
1927 static void pktgen_change_name(const struct pktgen_net *pn, struct net_device *dev)
1929 struct pktgen_thread *t;
1931 list_for_each_entry(t, &pn->pktgen_threads, th_list) {
1932 struct pktgen_dev *pkt_dev;
1934 list_for_each_entry(pkt_dev, &t->if_list, list) {
1935 if (pkt_dev->odev != dev)
1936 continue;
1938 proc_remove(pkt_dev->entry);
1940 pkt_dev->entry = proc_create_data(dev->name, 0600,
1941 pn->proc_dir,
1942 &pktgen_if_fops,
1943 pkt_dev);
1944 if (!pkt_dev->entry)
1945 pr_err("can't move proc entry for '%s'\n",
1946 dev->name);
1947 break;
1952 static int pktgen_device_event(struct notifier_block *unused,
1953 unsigned long event, void *ptr)
1955 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1956 struct pktgen_net *pn = net_generic(dev_net(dev), pg_net_id);
1958 if (pn->pktgen_exiting)
1959 return NOTIFY_DONE;
1961 /* It is OK that we do not hold the group lock right now,
1962 * as we run under the RTNL lock.
1965 switch (event) {
1966 case NETDEV_CHANGENAME:
1967 pktgen_change_name(pn, dev);
1968 break;
1970 case NETDEV_UNREGISTER:
1971 pktgen_mark_device(pn, dev->name);
1972 break;
1975 return NOTIFY_DONE;
1978 static struct net_device *pktgen_dev_get_by_name(const struct pktgen_net *pn,
1979 struct pktgen_dev *pkt_dev,
1980 const char *ifname)
1982 char b[IFNAMSIZ+5];
1983 int i;
1985 for (i = 0; ifname[i] != '@'; i++) {
1986 if (i == IFNAMSIZ)
1987 break;
1989 b[i] = ifname[i];
1991 b[i] = 0;
1993 return dev_get_by_name(pn->net, b);
1997 /* Associate pktgen_dev with a device. */
1999 static int pktgen_setup_dev(const struct pktgen_net *pn,
2000 struct pktgen_dev *pkt_dev, const char *ifname)
2002 struct net_device *odev;
2003 int err;
2005 /* Clean old setups */
2006 if (pkt_dev->odev) {
2007 dev_put(pkt_dev->odev);
2008 pkt_dev->odev = NULL;
2011 odev = pktgen_dev_get_by_name(pn, pkt_dev, ifname);
2012 if (!odev) {
2013 pr_err("no such netdevice: \"%s\"\n", ifname);
2014 return -ENODEV;
2017 if (odev->type != ARPHRD_ETHER) {
2018 pr_err("not an ethernet device: \"%s\"\n", ifname);
2019 err = -EINVAL;
2020 } else if (!netif_running(odev)) {
2021 pr_err("device is down: \"%s\"\n", ifname);
2022 err = -ENETDOWN;
2023 } else {
2024 pkt_dev->odev = odev;
2025 return 0;
2028 dev_put(odev);
2029 return err;
2032 /* Read pkt_dev from the interface and set up internal pktgen_dev
2033 * structure to have the right information to create/send packets
2035 static void pktgen_setup_inject(struct pktgen_dev *pkt_dev)
2037 int ntxq;
2039 if (!pkt_dev->odev) {
2040 pr_err("ERROR: pkt_dev->odev == NULL in setup_inject\n");
2041 sprintf(pkt_dev->result,
2042 "ERROR: pkt_dev->odev == NULL in setup_inject.\n");
2043 return;
2046 /* make sure that we don't pick a non-existing transmit queue */
2047 ntxq = pkt_dev->odev->real_num_tx_queues;
2049 if (ntxq <= pkt_dev->queue_map_min) {
2050 pr_warning("WARNING: Requested queue_map_min (zero-based) (%d) exceeds valid range [0 - %d] for (%d) queues on %s, resetting\n",
2051 pkt_dev->queue_map_min, (ntxq ?: 1) - 1, ntxq,
2052 pkt_dev->odevname);
2053 pkt_dev->queue_map_min = (ntxq ?: 1) - 1;
2055 if (pkt_dev->queue_map_max >= ntxq) {
2056 pr_warning("WARNING: Requested queue_map_max (zero-based) (%d) exceeds valid range [0 - %d] for (%d) queues on %s, resetting\n",
2057 pkt_dev->queue_map_max, (ntxq ?: 1) - 1, ntxq,
2058 pkt_dev->odevname);
2059 pkt_dev->queue_map_max = (ntxq ?: 1) - 1;
2062 /* Default to the interface's mac if not explicitly set. */
2064 if (is_zero_ether_addr(pkt_dev->src_mac))
2065 ether_addr_copy(&(pkt_dev->hh[6]), pkt_dev->odev->dev_addr);
2067 /* Set up Dest MAC */
2068 ether_addr_copy(&(pkt_dev->hh[0]), pkt_dev->dst_mac);
2070 if (pkt_dev->flags & F_IPV6) {
2071 int i, set = 0, err = 1;
2072 struct inet6_dev *idev;
2074 if (pkt_dev->min_pkt_size == 0) {
2075 pkt_dev->min_pkt_size = 14 + sizeof(struct ipv6hdr)
2076 + sizeof(struct udphdr)
2077 + sizeof(struct pktgen_hdr)
2078 + pkt_dev->pkt_overhead;
2081 for (i = 0; i < IN6_ADDR_HSIZE; i++)
2082 if (pkt_dev->cur_in6_saddr.s6_addr[i]) {
2083 set = 1;
2084 break;
2087 if (!set) {
2090 * Use linklevel address if unconfigured.
2092 * use ipv6_get_lladdr if/when it's get exported
2095 rcu_read_lock();
2096 idev = __in6_dev_get(pkt_dev->odev);
2097 if (idev) {
2098 struct inet6_ifaddr *ifp;
2100 read_lock_bh(&idev->lock);
2101 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2102 if ((ifp->scope & IFA_LINK) &&
2103 !(ifp->flags & IFA_F_TENTATIVE)) {
2104 pkt_dev->cur_in6_saddr = ifp->addr;
2105 err = 0;
2106 break;
2109 read_unlock_bh(&idev->lock);
2111 rcu_read_unlock();
2112 if (err)
2113 pr_err("ERROR: IPv6 link address not available\n");
2115 } else {
2116 if (pkt_dev->min_pkt_size == 0) {
2117 pkt_dev->min_pkt_size = 14 + sizeof(struct iphdr)
2118 + sizeof(struct udphdr)
2119 + sizeof(struct pktgen_hdr)
2120 + pkt_dev->pkt_overhead;
2123 pkt_dev->saddr_min = 0;
2124 pkt_dev->saddr_max = 0;
2125 if (strlen(pkt_dev->src_min) == 0) {
2127 struct in_device *in_dev;
2129 rcu_read_lock();
2130 in_dev = __in_dev_get_rcu(pkt_dev->odev);
2131 if (in_dev) {
2132 if (in_dev->ifa_list) {
2133 pkt_dev->saddr_min =
2134 in_dev->ifa_list->ifa_address;
2135 pkt_dev->saddr_max = pkt_dev->saddr_min;
2138 rcu_read_unlock();
2139 } else {
2140 pkt_dev->saddr_min = in_aton(pkt_dev->src_min);
2141 pkt_dev->saddr_max = in_aton(pkt_dev->src_max);
2144 pkt_dev->daddr_min = in_aton(pkt_dev->dst_min);
2145 pkt_dev->daddr_max = in_aton(pkt_dev->dst_max);
2147 /* Initialize current values. */
2148 pkt_dev->cur_pkt_size = pkt_dev->min_pkt_size;
2149 if (pkt_dev->min_pkt_size > pkt_dev->max_pkt_size)
2150 pkt_dev->max_pkt_size = pkt_dev->min_pkt_size;
2152 pkt_dev->cur_dst_mac_offset = 0;
2153 pkt_dev->cur_src_mac_offset = 0;
2154 pkt_dev->cur_saddr = pkt_dev->saddr_min;
2155 pkt_dev->cur_daddr = pkt_dev->daddr_min;
2156 pkt_dev->cur_udp_dst = pkt_dev->udp_dst_min;
2157 pkt_dev->cur_udp_src = pkt_dev->udp_src_min;
2158 pkt_dev->nflows = 0;
2162 static void spin(struct pktgen_dev *pkt_dev, ktime_t spin_until)
2164 ktime_t start_time, end_time;
2165 s64 remaining;
2166 struct hrtimer_sleeper t;
2168 hrtimer_init_on_stack(&t.timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS);
2169 hrtimer_set_expires(&t.timer, spin_until);
2171 remaining = ktime_to_ns(hrtimer_expires_remaining(&t.timer));
2172 if (remaining <= 0) {
2173 pkt_dev->next_tx = ktime_add_ns(spin_until, pkt_dev->delay);
2174 return;
2177 start_time = ktime_get();
2178 if (remaining < 100000) {
2179 /* for small delays (<100us), just loop until limit is reached */
2180 do {
2181 end_time = ktime_get();
2182 } while (ktime_compare(end_time, spin_until) < 0);
2183 } else {
2184 /* see do_nanosleep */
2185 hrtimer_init_sleeper(&t, current);
2186 do {
2187 set_current_state(TASK_INTERRUPTIBLE);
2188 hrtimer_start_expires(&t.timer, HRTIMER_MODE_ABS);
2189 if (!hrtimer_active(&t.timer))
2190 t.task = NULL;
2192 if (likely(t.task))
2193 schedule();
2195 hrtimer_cancel(&t.timer);
2196 } while (t.task && pkt_dev->running && !signal_pending(current));
2197 __set_current_state(TASK_RUNNING);
2198 end_time = ktime_get();
2201 pkt_dev->idle_acc += ktime_to_ns(ktime_sub(end_time, start_time));
2202 pkt_dev->next_tx = ktime_add_ns(spin_until, pkt_dev->delay);
2205 static inline void set_pkt_overhead(struct pktgen_dev *pkt_dev)
2207 pkt_dev->pkt_overhead = 0;
2208 pkt_dev->pkt_overhead += pkt_dev->nr_labels*sizeof(u32);
2209 pkt_dev->pkt_overhead += VLAN_TAG_SIZE(pkt_dev);
2210 pkt_dev->pkt_overhead += SVLAN_TAG_SIZE(pkt_dev);
2213 static inline int f_seen(const struct pktgen_dev *pkt_dev, int flow)
2215 return !!(pkt_dev->flows[flow].flags & F_INIT);
2218 static inline int f_pick(struct pktgen_dev *pkt_dev)
2220 int flow = pkt_dev->curfl;
2222 if (pkt_dev->flags & F_FLOW_SEQ) {
2223 if (pkt_dev->flows[flow].count >= pkt_dev->lflow) {
2224 /* reset time */
2225 pkt_dev->flows[flow].count = 0;
2226 pkt_dev->flows[flow].flags = 0;
2227 pkt_dev->curfl += 1;
2228 if (pkt_dev->curfl >= pkt_dev->cflows)
2229 pkt_dev->curfl = 0; /*reset */
2231 } else {
2232 flow = prandom_u32() % pkt_dev->cflows;
2233 pkt_dev->curfl = flow;
2235 if (pkt_dev->flows[flow].count > pkt_dev->lflow) {
2236 pkt_dev->flows[flow].count = 0;
2237 pkt_dev->flows[flow].flags = 0;
2241 return pkt_dev->curfl;
2245 #ifdef CONFIG_XFRM
2246 /* If there was already an IPSEC SA, we keep it as is, else
2247 * we go look for it ...
2249 #define DUMMY_MARK 0
2250 static void get_ipsec_sa(struct pktgen_dev *pkt_dev, int flow)
2252 struct xfrm_state *x = pkt_dev->flows[flow].x;
2253 struct pktgen_net *pn = net_generic(dev_net(pkt_dev->odev), pg_net_id);
2254 if (!x) {
2256 if (pkt_dev->spi) {
2257 /* We need as quick as possible to find the right SA
2258 * Searching with minimum criteria to archieve this.
2260 x = xfrm_state_lookup_byspi(pn->net, htonl(pkt_dev->spi), AF_INET);
2261 } else {
2262 /* slow path: we dont already have xfrm_state */
2263 x = xfrm_stateonly_find(pn->net, DUMMY_MARK,
2264 (xfrm_address_t *)&pkt_dev->cur_daddr,
2265 (xfrm_address_t *)&pkt_dev->cur_saddr,
2266 AF_INET,
2267 pkt_dev->ipsmode,
2268 pkt_dev->ipsproto, 0);
2270 if (x) {
2271 pkt_dev->flows[flow].x = x;
2272 set_pkt_overhead(pkt_dev);
2273 pkt_dev->pkt_overhead += x->props.header_len;
2278 #endif
2279 static void set_cur_queue_map(struct pktgen_dev *pkt_dev)
2282 if (pkt_dev->flags & F_QUEUE_MAP_CPU)
2283 pkt_dev->cur_queue_map = smp_processor_id();
2285 else if (pkt_dev->queue_map_min <= pkt_dev->queue_map_max) {
2286 __u16 t;
2287 if (pkt_dev->flags & F_QUEUE_MAP_RND) {
2288 t = prandom_u32() %
2289 (pkt_dev->queue_map_max -
2290 pkt_dev->queue_map_min + 1)
2291 + pkt_dev->queue_map_min;
2292 } else {
2293 t = pkt_dev->cur_queue_map + 1;
2294 if (t > pkt_dev->queue_map_max)
2295 t = pkt_dev->queue_map_min;
2297 pkt_dev->cur_queue_map = t;
2299 pkt_dev->cur_queue_map = pkt_dev->cur_queue_map % pkt_dev->odev->real_num_tx_queues;
2302 /* Increment/randomize headers according to flags and current values
2303 * for IP src/dest, UDP src/dst port, MAC-Addr src/dst
2305 static void mod_cur_headers(struct pktgen_dev *pkt_dev)
2307 __u32 imn;
2308 __u32 imx;
2309 int flow = 0;
2311 if (pkt_dev->cflows)
2312 flow = f_pick(pkt_dev);
2314 /* Deal with source MAC */
2315 if (pkt_dev->src_mac_count > 1) {
2316 __u32 mc;
2317 __u32 tmp;
2319 if (pkt_dev->flags & F_MACSRC_RND)
2320 mc = prandom_u32() % pkt_dev->src_mac_count;
2321 else {
2322 mc = pkt_dev->cur_src_mac_offset++;
2323 if (pkt_dev->cur_src_mac_offset >=
2324 pkt_dev->src_mac_count)
2325 pkt_dev->cur_src_mac_offset = 0;
2328 tmp = pkt_dev->src_mac[5] + (mc & 0xFF);
2329 pkt_dev->hh[11] = tmp;
2330 tmp = (pkt_dev->src_mac[4] + ((mc >> 8) & 0xFF) + (tmp >> 8));
2331 pkt_dev->hh[10] = tmp;
2332 tmp = (pkt_dev->src_mac[3] + ((mc >> 16) & 0xFF) + (tmp >> 8));
2333 pkt_dev->hh[9] = tmp;
2334 tmp = (pkt_dev->src_mac[2] + ((mc >> 24) & 0xFF) + (tmp >> 8));
2335 pkt_dev->hh[8] = tmp;
2336 tmp = (pkt_dev->src_mac[1] + (tmp >> 8));
2337 pkt_dev->hh[7] = tmp;
2340 /* Deal with Destination MAC */
2341 if (pkt_dev->dst_mac_count > 1) {
2342 __u32 mc;
2343 __u32 tmp;
2345 if (pkt_dev->flags & F_MACDST_RND)
2346 mc = prandom_u32() % pkt_dev->dst_mac_count;
2348 else {
2349 mc = pkt_dev->cur_dst_mac_offset++;
2350 if (pkt_dev->cur_dst_mac_offset >=
2351 pkt_dev->dst_mac_count) {
2352 pkt_dev->cur_dst_mac_offset = 0;
2356 tmp = pkt_dev->dst_mac[5] + (mc & 0xFF);
2357 pkt_dev->hh[5] = tmp;
2358 tmp = (pkt_dev->dst_mac[4] + ((mc >> 8) & 0xFF) + (tmp >> 8));
2359 pkt_dev->hh[4] = tmp;
2360 tmp = (pkt_dev->dst_mac[3] + ((mc >> 16) & 0xFF) + (tmp >> 8));
2361 pkt_dev->hh[3] = tmp;
2362 tmp = (pkt_dev->dst_mac[2] + ((mc >> 24) & 0xFF) + (tmp >> 8));
2363 pkt_dev->hh[2] = tmp;
2364 tmp = (pkt_dev->dst_mac[1] + (tmp >> 8));
2365 pkt_dev->hh[1] = tmp;
2368 if (pkt_dev->flags & F_MPLS_RND) {
2369 unsigned int i;
2370 for (i = 0; i < pkt_dev->nr_labels; i++)
2371 if (pkt_dev->labels[i] & MPLS_STACK_BOTTOM)
2372 pkt_dev->labels[i] = MPLS_STACK_BOTTOM |
2373 ((__force __be32)prandom_u32() &
2374 htonl(0x000fffff));
2377 if ((pkt_dev->flags & F_VID_RND) && (pkt_dev->vlan_id != 0xffff)) {
2378 pkt_dev->vlan_id = prandom_u32() & (4096 - 1);
2381 if ((pkt_dev->flags & F_SVID_RND) && (pkt_dev->svlan_id != 0xffff)) {
2382 pkt_dev->svlan_id = prandom_u32() & (4096 - 1);
2385 if (pkt_dev->udp_src_min < pkt_dev->udp_src_max) {
2386 if (pkt_dev->flags & F_UDPSRC_RND)
2387 pkt_dev->cur_udp_src = prandom_u32() %
2388 (pkt_dev->udp_src_max - pkt_dev->udp_src_min)
2389 + pkt_dev->udp_src_min;
2391 else {
2392 pkt_dev->cur_udp_src++;
2393 if (pkt_dev->cur_udp_src >= pkt_dev->udp_src_max)
2394 pkt_dev->cur_udp_src = pkt_dev->udp_src_min;
2398 if (pkt_dev->udp_dst_min < pkt_dev->udp_dst_max) {
2399 if (pkt_dev->flags & F_UDPDST_RND) {
2400 pkt_dev->cur_udp_dst = prandom_u32() %
2401 (pkt_dev->udp_dst_max - pkt_dev->udp_dst_min)
2402 + pkt_dev->udp_dst_min;
2403 } else {
2404 pkt_dev->cur_udp_dst++;
2405 if (pkt_dev->cur_udp_dst >= pkt_dev->udp_dst_max)
2406 pkt_dev->cur_udp_dst = pkt_dev->udp_dst_min;
2410 if (!(pkt_dev->flags & F_IPV6)) {
2412 imn = ntohl(pkt_dev->saddr_min);
2413 imx = ntohl(pkt_dev->saddr_max);
2414 if (imn < imx) {
2415 __u32 t;
2416 if (pkt_dev->flags & F_IPSRC_RND)
2417 t = prandom_u32() % (imx - imn) + imn;
2418 else {
2419 t = ntohl(pkt_dev->cur_saddr);
2420 t++;
2421 if (t > imx)
2422 t = imn;
2425 pkt_dev->cur_saddr = htonl(t);
2428 if (pkt_dev->cflows && f_seen(pkt_dev, flow)) {
2429 pkt_dev->cur_daddr = pkt_dev->flows[flow].cur_daddr;
2430 } else {
2431 imn = ntohl(pkt_dev->daddr_min);
2432 imx = ntohl(pkt_dev->daddr_max);
2433 if (imn < imx) {
2434 __u32 t;
2435 __be32 s;
2436 if (pkt_dev->flags & F_IPDST_RND) {
2438 do {
2439 t = prandom_u32() %
2440 (imx - imn) + imn;
2441 s = htonl(t);
2442 } while (ipv4_is_loopback(s) ||
2443 ipv4_is_multicast(s) ||
2444 ipv4_is_lbcast(s) ||
2445 ipv4_is_zeronet(s) ||
2446 ipv4_is_local_multicast(s));
2447 pkt_dev->cur_daddr = s;
2448 } else {
2449 t = ntohl(pkt_dev->cur_daddr);
2450 t++;
2451 if (t > imx) {
2452 t = imn;
2454 pkt_dev->cur_daddr = htonl(t);
2457 if (pkt_dev->cflows) {
2458 pkt_dev->flows[flow].flags |= F_INIT;
2459 pkt_dev->flows[flow].cur_daddr =
2460 pkt_dev->cur_daddr;
2461 #ifdef CONFIG_XFRM
2462 if (pkt_dev->flags & F_IPSEC_ON)
2463 get_ipsec_sa(pkt_dev, flow);
2464 #endif
2465 pkt_dev->nflows++;
2468 } else { /* IPV6 * */
2470 if (!ipv6_addr_any(&pkt_dev->min_in6_daddr)) {
2471 int i;
2473 /* Only random destinations yet */
2475 for (i = 0; i < 4; i++) {
2476 pkt_dev->cur_in6_daddr.s6_addr32[i] =
2477 (((__force __be32)prandom_u32() |
2478 pkt_dev->min_in6_daddr.s6_addr32[i]) &
2479 pkt_dev->max_in6_daddr.s6_addr32[i]);
2484 if (pkt_dev->min_pkt_size < pkt_dev->max_pkt_size) {
2485 __u32 t;
2486 if (pkt_dev->flags & F_TXSIZE_RND) {
2487 t = prandom_u32() %
2488 (pkt_dev->max_pkt_size - pkt_dev->min_pkt_size)
2489 + pkt_dev->min_pkt_size;
2490 } else {
2491 t = pkt_dev->cur_pkt_size + 1;
2492 if (t > pkt_dev->max_pkt_size)
2493 t = pkt_dev->min_pkt_size;
2495 pkt_dev->cur_pkt_size = t;
2498 set_cur_queue_map(pkt_dev);
2500 pkt_dev->flows[flow].count++;
2504 #ifdef CONFIG_XFRM
2505 static u32 pktgen_dst_metrics[RTAX_MAX + 1] = {
2507 [RTAX_HOPLIMIT] = 0x5, /* Set a static hoplimit */
2510 static int pktgen_output_ipsec(struct sk_buff *skb, struct pktgen_dev *pkt_dev)
2512 struct xfrm_state *x = pkt_dev->flows[pkt_dev->curfl].x;
2513 int err = 0;
2514 struct net *net = dev_net(pkt_dev->odev);
2516 if (!x)
2517 return 0;
2518 /* XXX: we dont support tunnel mode for now until
2519 * we resolve the dst issue */
2520 if ((x->props.mode != XFRM_MODE_TRANSPORT) && (pkt_dev->spi == 0))
2521 return 0;
2523 /* But when user specify an valid SPI, transformation
2524 * supports both transport/tunnel mode + ESP/AH type.
2526 if ((x->props.mode == XFRM_MODE_TUNNEL) && (pkt_dev->spi != 0))
2527 skb->_skb_refdst = (unsigned long)&pkt_dev->dst | SKB_DST_NOREF;
2529 rcu_read_lock_bh();
2530 err = x->outer_mode->output(x, skb);
2531 rcu_read_unlock_bh();
2532 if (err) {
2533 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATEMODEERROR);
2534 goto error;
2536 err = x->type->output(x, skb);
2537 if (err) {
2538 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATEPROTOERROR);
2539 goto error;
2541 spin_lock_bh(&x->lock);
2542 x->curlft.bytes += skb->len;
2543 x->curlft.packets++;
2544 spin_unlock_bh(&x->lock);
2545 error:
2546 return err;
2549 static void free_SAs(struct pktgen_dev *pkt_dev)
2551 if (pkt_dev->cflows) {
2552 /* let go of the SAs if we have them */
2553 int i;
2554 for (i = 0; i < pkt_dev->cflows; i++) {
2555 struct xfrm_state *x = pkt_dev->flows[i].x;
2556 if (x) {
2557 xfrm_state_put(x);
2558 pkt_dev->flows[i].x = NULL;
2564 static int process_ipsec(struct pktgen_dev *pkt_dev,
2565 struct sk_buff *skb, __be16 protocol)
2567 if (pkt_dev->flags & F_IPSEC_ON) {
2568 struct xfrm_state *x = pkt_dev->flows[pkt_dev->curfl].x;
2569 int nhead = 0;
2570 if (x) {
2571 int ret;
2572 __u8 *eth;
2573 struct iphdr *iph;
2575 nhead = x->props.header_len - skb_headroom(skb);
2576 if (nhead > 0) {
2577 ret = pskb_expand_head(skb, nhead, 0, GFP_ATOMIC);
2578 if (ret < 0) {
2579 pr_err("Error expanding ipsec packet %d\n",
2580 ret);
2581 goto err;
2585 /* ipsec is not expecting ll header */
2586 skb_pull(skb, ETH_HLEN);
2587 ret = pktgen_output_ipsec(skb, pkt_dev);
2588 if (ret) {
2589 pr_err("Error creating ipsec packet %d\n", ret);
2590 goto err;
2592 /* restore ll */
2593 eth = (__u8 *) skb_push(skb, ETH_HLEN);
2594 memcpy(eth, pkt_dev->hh, 12);
2595 *(u16 *) &eth[12] = protocol;
2597 /* Update IPv4 header len as well as checksum value */
2598 iph = ip_hdr(skb);
2599 iph->tot_len = htons(skb->len - ETH_HLEN);
2600 ip_send_check(iph);
2603 return 1;
2604 err:
2605 kfree_skb(skb);
2606 return 0;
2608 #endif
2610 static void mpls_push(__be32 *mpls, struct pktgen_dev *pkt_dev)
2612 unsigned int i;
2613 for (i = 0; i < pkt_dev->nr_labels; i++)
2614 *mpls++ = pkt_dev->labels[i] & ~MPLS_STACK_BOTTOM;
2616 mpls--;
2617 *mpls |= MPLS_STACK_BOTTOM;
2620 static inline __be16 build_tci(unsigned int id, unsigned int cfi,
2621 unsigned int prio)
2623 return htons(id | (cfi << 12) | (prio << 13));
2626 static void pktgen_finalize_skb(struct pktgen_dev *pkt_dev, struct sk_buff *skb,
2627 int datalen)
2629 struct timeval timestamp;
2630 struct pktgen_hdr *pgh;
2632 pgh = (struct pktgen_hdr *)skb_put(skb, sizeof(*pgh));
2633 datalen -= sizeof(*pgh);
2635 if (pkt_dev->nfrags <= 0) {
2636 memset(skb_put(skb, datalen), 0, datalen);
2637 } else {
2638 int frags = pkt_dev->nfrags;
2639 int i, len;
2640 int frag_len;
2643 if (frags > MAX_SKB_FRAGS)
2644 frags = MAX_SKB_FRAGS;
2645 len = datalen - frags * PAGE_SIZE;
2646 if (len > 0) {
2647 memset(skb_put(skb, len), 0, len);
2648 datalen = frags * PAGE_SIZE;
2651 i = 0;
2652 frag_len = (datalen/frags) < PAGE_SIZE ?
2653 (datalen/frags) : PAGE_SIZE;
2654 while (datalen > 0) {
2655 if (unlikely(!pkt_dev->page)) {
2656 int node = numa_node_id();
2658 if (pkt_dev->node >= 0 && (pkt_dev->flags & F_NODE))
2659 node = pkt_dev->node;
2660 pkt_dev->page = alloc_pages_node(node, GFP_KERNEL | __GFP_ZERO, 0);
2661 if (!pkt_dev->page)
2662 break;
2664 get_page(pkt_dev->page);
2665 skb_frag_set_page(skb, i, pkt_dev->page);
2666 skb_shinfo(skb)->frags[i].page_offset = 0;
2667 /*last fragment, fill rest of data*/
2668 if (i == (frags - 1))
2669 skb_frag_size_set(&skb_shinfo(skb)->frags[i],
2670 (datalen < PAGE_SIZE ? datalen : PAGE_SIZE));
2671 else
2672 skb_frag_size_set(&skb_shinfo(skb)->frags[i], frag_len);
2673 datalen -= skb_frag_size(&skb_shinfo(skb)->frags[i]);
2674 skb->len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2675 skb->data_len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2676 i++;
2677 skb_shinfo(skb)->nr_frags = i;
2681 /* Stamp the time, and sequence number,
2682 * convert them to network byte order
2684 pgh->pgh_magic = htonl(PKTGEN_MAGIC);
2685 pgh->seq_num = htonl(pkt_dev->seq_num);
2687 do_gettimeofday(&timestamp);
2688 pgh->tv_sec = htonl(timestamp.tv_sec);
2689 pgh->tv_usec = htonl(timestamp.tv_usec);
2692 static struct sk_buff *pktgen_alloc_skb(struct net_device *dev,
2693 struct pktgen_dev *pkt_dev,
2694 unsigned int extralen)
2696 struct sk_buff *skb = NULL;
2697 unsigned int size = pkt_dev->cur_pkt_size + 64 + extralen +
2698 pkt_dev->pkt_overhead;
2700 if (pkt_dev->flags & F_NODE) {
2701 int node = pkt_dev->node >= 0 ? pkt_dev->node : numa_node_id();
2703 skb = __alloc_skb(NET_SKB_PAD + size, GFP_NOWAIT, 0, node);
2704 if (likely(skb)) {
2705 skb_reserve(skb, NET_SKB_PAD);
2706 skb->dev = dev;
2708 } else {
2709 skb = __netdev_alloc_skb(dev, size, GFP_NOWAIT);
2712 return skb;
2715 static struct sk_buff *fill_packet_ipv4(struct net_device *odev,
2716 struct pktgen_dev *pkt_dev)
2718 struct sk_buff *skb = NULL;
2719 __u8 *eth;
2720 struct udphdr *udph;
2721 int datalen, iplen;
2722 struct iphdr *iph;
2723 __be16 protocol = htons(ETH_P_IP);
2724 __be32 *mpls;
2725 __be16 *vlan_tci = NULL; /* Encapsulates priority and VLAN ID */
2726 __be16 *vlan_encapsulated_proto = NULL; /* packet type ID field (or len) for VLAN tag */
2727 __be16 *svlan_tci = NULL; /* Encapsulates priority and SVLAN ID */
2728 __be16 *svlan_encapsulated_proto = NULL; /* packet type ID field (or len) for SVLAN tag */
2729 u16 queue_map;
2731 if (pkt_dev->nr_labels)
2732 protocol = htons(ETH_P_MPLS_UC);
2734 if (pkt_dev->vlan_id != 0xffff)
2735 protocol = htons(ETH_P_8021Q);
2737 /* Update any of the values, used when we're incrementing various
2738 * fields.
2740 mod_cur_headers(pkt_dev);
2741 queue_map = pkt_dev->cur_queue_map;
2743 datalen = (odev->hard_header_len + 16) & ~0xf;
2745 skb = pktgen_alloc_skb(odev, pkt_dev, datalen);
2746 if (!skb) {
2747 sprintf(pkt_dev->result, "No memory");
2748 return NULL;
2751 prefetchw(skb->data);
2752 skb_reserve(skb, datalen);
2754 /* Reserve for ethernet and IP header */
2755 eth = (__u8 *) skb_push(skb, 14);
2756 mpls = (__be32 *)skb_put(skb, pkt_dev->nr_labels*sizeof(__u32));
2757 if (pkt_dev->nr_labels)
2758 mpls_push(mpls, pkt_dev);
2760 if (pkt_dev->vlan_id != 0xffff) {
2761 if (pkt_dev->svlan_id != 0xffff) {
2762 svlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2763 *svlan_tci = build_tci(pkt_dev->svlan_id,
2764 pkt_dev->svlan_cfi,
2765 pkt_dev->svlan_p);
2766 svlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2767 *svlan_encapsulated_proto = htons(ETH_P_8021Q);
2769 vlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2770 *vlan_tci = build_tci(pkt_dev->vlan_id,
2771 pkt_dev->vlan_cfi,
2772 pkt_dev->vlan_p);
2773 vlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2774 *vlan_encapsulated_proto = htons(ETH_P_IP);
2777 skb_set_mac_header(skb, 0);
2778 skb_set_network_header(skb, skb->len);
2779 iph = (struct iphdr *) skb_put(skb, sizeof(struct iphdr));
2781 skb_set_transport_header(skb, skb->len);
2782 udph = (struct udphdr *) skb_put(skb, sizeof(struct udphdr));
2783 skb_set_queue_mapping(skb, queue_map);
2784 skb->priority = pkt_dev->skb_priority;
2786 memcpy(eth, pkt_dev->hh, 12);
2787 *(__be16 *) & eth[12] = protocol;
2789 /* Eth + IPh + UDPh + mpls */
2790 datalen = pkt_dev->cur_pkt_size - 14 - 20 - 8 -
2791 pkt_dev->pkt_overhead;
2792 if (datalen < 0 || datalen < sizeof(struct pktgen_hdr))
2793 datalen = sizeof(struct pktgen_hdr);
2795 udph->source = htons(pkt_dev->cur_udp_src);
2796 udph->dest = htons(pkt_dev->cur_udp_dst);
2797 udph->len = htons(datalen + 8); /* DATA + udphdr */
2798 udph->check = 0;
2800 iph->ihl = 5;
2801 iph->version = 4;
2802 iph->ttl = 32;
2803 iph->tos = pkt_dev->tos;
2804 iph->protocol = IPPROTO_UDP; /* UDP */
2805 iph->saddr = pkt_dev->cur_saddr;
2806 iph->daddr = pkt_dev->cur_daddr;
2807 iph->id = htons(pkt_dev->ip_id);
2808 pkt_dev->ip_id++;
2809 iph->frag_off = 0;
2810 iplen = 20 + 8 + datalen;
2811 iph->tot_len = htons(iplen);
2812 ip_send_check(iph);
2813 skb->protocol = protocol;
2814 skb->dev = odev;
2815 skb->pkt_type = PACKET_HOST;
2817 if (!(pkt_dev->flags & F_UDPCSUM)) {
2818 skb->ip_summed = CHECKSUM_NONE;
2819 } else if (odev->features & NETIF_F_V4_CSUM) {
2820 skb->ip_summed = CHECKSUM_PARTIAL;
2821 skb->csum = 0;
2822 udp4_hwcsum(skb, udph->source, udph->dest);
2823 } else {
2824 __wsum csum = udp_csum(skb);
2826 /* add protocol-dependent pseudo-header */
2827 udph->check = csum_tcpudp_magic(udph->source, udph->dest,
2828 datalen + 8, IPPROTO_UDP, csum);
2830 if (udph->check == 0)
2831 udph->check = CSUM_MANGLED_0;
2834 pktgen_finalize_skb(pkt_dev, skb, datalen);
2836 #ifdef CONFIG_XFRM
2837 if (!process_ipsec(pkt_dev, skb, protocol))
2838 return NULL;
2839 #endif
2841 return skb;
2844 static struct sk_buff *fill_packet_ipv6(struct net_device *odev,
2845 struct pktgen_dev *pkt_dev)
2847 struct sk_buff *skb = NULL;
2848 __u8 *eth;
2849 struct udphdr *udph;
2850 int datalen, udplen;
2851 struct ipv6hdr *iph;
2852 __be16 protocol = htons(ETH_P_IPV6);
2853 __be32 *mpls;
2854 __be16 *vlan_tci = NULL; /* Encapsulates priority and VLAN ID */
2855 __be16 *vlan_encapsulated_proto = NULL; /* packet type ID field (or len) for VLAN tag */
2856 __be16 *svlan_tci = NULL; /* Encapsulates priority and SVLAN ID */
2857 __be16 *svlan_encapsulated_proto = NULL; /* packet type ID field (or len) for SVLAN tag */
2858 u16 queue_map;
2860 if (pkt_dev->nr_labels)
2861 protocol = htons(ETH_P_MPLS_UC);
2863 if (pkt_dev->vlan_id != 0xffff)
2864 protocol = htons(ETH_P_8021Q);
2866 /* Update any of the values, used when we're incrementing various
2867 * fields.
2869 mod_cur_headers(pkt_dev);
2870 queue_map = pkt_dev->cur_queue_map;
2872 skb = pktgen_alloc_skb(odev, pkt_dev, 16);
2873 if (!skb) {
2874 sprintf(pkt_dev->result, "No memory");
2875 return NULL;
2878 prefetchw(skb->data);
2879 skb_reserve(skb, 16);
2881 /* Reserve for ethernet and IP header */
2882 eth = (__u8 *) skb_push(skb, 14);
2883 mpls = (__be32 *)skb_put(skb, pkt_dev->nr_labels*sizeof(__u32));
2884 if (pkt_dev->nr_labels)
2885 mpls_push(mpls, pkt_dev);
2887 if (pkt_dev->vlan_id != 0xffff) {
2888 if (pkt_dev->svlan_id != 0xffff) {
2889 svlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2890 *svlan_tci = build_tci(pkt_dev->svlan_id,
2891 pkt_dev->svlan_cfi,
2892 pkt_dev->svlan_p);
2893 svlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2894 *svlan_encapsulated_proto = htons(ETH_P_8021Q);
2896 vlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2897 *vlan_tci = build_tci(pkt_dev->vlan_id,
2898 pkt_dev->vlan_cfi,
2899 pkt_dev->vlan_p);
2900 vlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2901 *vlan_encapsulated_proto = htons(ETH_P_IPV6);
2904 skb_set_mac_header(skb, 0);
2905 skb_set_network_header(skb, skb->len);
2906 iph = (struct ipv6hdr *) skb_put(skb, sizeof(struct ipv6hdr));
2908 skb_set_transport_header(skb, skb->len);
2909 udph = (struct udphdr *) skb_put(skb, sizeof(struct udphdr));
2910 skb_set_queue_mapping(skb, queue_map);
2911 skb->priority = pkt_dev->skb_priority;
2913 memcpy(eth, pkt_dev->hh, 12);
2914 *(__be16 *) &eth[12] = protocol;
2916 /* Eth + IPh + UDPh + mpls */
2917 datalen = pkt_dev->cur_pkt_size - 14 -
2918 sizeof(struct ipv6hdr) - sizeof(struct udphdr) -
2919 pkt_dev->pkt_overhead;
2921 if (datalen < 0 || datalen < sizeof(struct pktgen_hdr)) {
2922 datalen = sizeof(struct pktgen_hdr);
2923 net_info_ratelimited("increased datalen to %d\n", datalen);
2926 udplen = datalen + sizeof(struct udphdr);
2927 udph->source = htons(pkt_dev->cur_udp_src);
2928 udph->dest = htons(pkt_dev->cur_udp_dst);
2929 udph->len = htons(udplen);
2930 udph->check = 0;
2932 *(__be32 *) iph = htonl(0x60000000); /* Version + flow */
2934 if (pkt_dev->traffic_class) {
2935 /* Version + traffic class + flow (0) */
2936 *(__be32 *)iph |= htonl(0x60000000 | (pkt_dev->traffic_class << 20));
2939 iph->hop_limit = 32;
2941 iph->payload_len = htons(udplen);
2942 iph->nexthdr = IPPROTO_UDP;
2944 iph->daddr = pkt_dev->cur_in6_daddr;
2945 iph->saddr = pkt_dev->cur_in6_saddr;
2947 skb->protocol = protocol;
2948 skb->dev = odev;
2949 skb->pkt_type = PACKET_HOST;
2951 if (!(pkt_dev->flags & F_UDPCSUM)) {
2952 skb->ip_summed = CHECKSUM_NONE;
2953 } else if (odev->features & NETIF_F_V6_CSUM) {
2954 skb->ip_summed = CHECKSUM_PARTIAL;
2955 skb->csum_start = skb_transport_header(skb) - skb->head;
2956 skb->csum_offset = offsetof(struct udphdr, check);
2957 udph->check = ~csum_ipv6_magic(&iph->saddr, &iph->daddr, udplen, IPPROTO_UDP, 0);
2958 } else {
2959 __wsum csum = udp_csum(skb);
2961 /* add protocol-dependent pseudo-header */
2962 udph->check = csum_ipv6_magic(&iph->saddr, &iph->daddr, udplen, IPPROTO_UDP, csum);
2964 if (udph->check == 0)
2965 udph->check = CSUM_MANGLED_0;
2968 pktgen_finalize_skb(pkt_dev, skb, datalen);
2970 return skb;
2973 static struct sk_buff *fill_packet(struct net_device *odev,
2974 struct pktgen_dev *pkt_dev)
2976 if (pkt_dev->flags & F_IPV6)
2977 return fill_packet_ipv6(odev, pkt_dev);
2978 else
2979 return fill_packet_ipv4(odev, pkt_dev);
2982 static void pktgen_clear_counters(struct pktgen_dev *pkt_dev)
2984 pkt_dev->seq_num = 1;
2985 pkt_dev->idle_acc = 0;
2986 pkt_dev->sofar = 0;
2987 pkt_dev->tx_bytes = 0;
2988 pkt_dev->errors = 0;
2991 /* Set up structure for sending pkts, clear counters */
2993 static void pktgen_run(struct pktgen_thread *t)
2995 struct pktgen_dev *pkt_dev;
2996 int started = 0;
2998 func_enter();
3000 if_lock(t);
3001 list_for_each_entry(pkt_dev, &t->if_list, list) {
3004 * setup odev and create initial packet.
3006 pktgen_setup_inject(pkt_dev);
3008 if (pkt_dev->odev) {
3009 pktgen_clear_counters(pkt_dev);
3010 pkt_dev->running = 1; /* Cranke yeself! */
3011 pkt_dev->skb = NULL;
3012 pkt_dev->started_at = pkt_dev->next_tx = ktime_get();
3014 set_pkt_overhead(pkt_dev);
3016 strcpy(pkt_dev->result, "Starting");
3017 started++;
3018 } else
3019 strcpy(pkt_dev->result, "Error starting");
3021 if_unlock(t);
3022 if (started)
3023 t->control &= ~(T_STOP);
3026 static void pktgen_stop_all_threads_ifs(struct pktgen_net *pn)
3028 struct pktgen_thread *t;
3030 func_enter();
3032 mutex_lock(&pktgen_thread_lock);
3034 list_for_each_entry(t, &pn->pktgen_threads, th_list)
3035 t->control |= T_STOP;
3037 mutex_unlock(&pktgen_thread_lock);
3040 static int thread_is_running(const struct pktgen_thread *t)
3042 const struct pktgen_dev *pkt_dev;
3044 list_for_each_entry(pkt_dev, &t->if_list, list)
3045 if (pkt_dev->running)
3046 return 1;
3047 return 0;
3050 static int pktgen_wait_thread_run(struct pktgen_thread *t)
3052 if_lock(t);
3054 while (thread_is_running(t)) {
3056 if_unlock(t);
3058 msleep_interruptible(100);
3060 if (signal_pending(current))
3061 goto signal;
3062 if_lock(t);
3064 if_unlock(t);
3065 return 1;
3066 signal:
3067 return 0;
3070 static int pktgen_wait_all_threads_run(struct pktgen_net *pn)
3072 struct pktgen_thread *t;
3073 int sig = 1;
3075 mutex_lock(&pktgen_thread_lock);
3077 list_for_each_entry(t, &pn->pktgen_threads, th_list) {
3078 sig = pktgen_wait_thread_run(t);
3079 if (sig == 0)
3080 break;
3083 if (sig == 0)
3084 list_for_each_entry(t, &pn->pktgen_threads, th_list)
3085 t->control |= (T_STOP);
3087 mutex_unlock(&pktgen_thread_lock);
3088 return sig;
3091 static void pktgen_run_all_threads(struct pktgen_net *pn)
3093 struct pktgen_thread *t;
3095 func_enter();
3097 mutex_lock(&pktgen_thread_lock);
3099 list_for_each_entry(t, &pn->pktgen_threads, th_list)
3100 t->control |= (T_RUN);
3102 mutex_unlock(&pktgen_thread_lock);
3104 /* Propagate thread->control */
3105 schedule_timeout_interruptible(msecs_to_jiffies(125));
3107 pktgen_wait_all_threads_run(pn);
3110 static void pktgen_reset_all_threads(struct pktgen_net *pn)
3112 struct pktgen_thread *t;
3114 func_enter();
3116 mutex_lock(&pktgen_thread_lock);
3118 list_for_each_entry(t, &pn->pktgen_threads, th_list)
3119 t->control |= (T_REMDEVALL);
3121 mutex_unlock(&pktgen_thread_lock);
3123 /* Propagate thread->control */
3124 schedule_timeout_interruptible(msecs_to_jiffies(125));
3126 pktgen_wait_all_threads_run(pn);
3129 static void show_results(struct pktgen_dev *pkt_dev, int nr_frags)
3131 __u64 bps, mbps, pps;
3132 char *p = pkt_dev->result;
3133 ktime_t elapsed = ktime_sub(pkt_dev->stopped_at,
3134 pkt_dev->started_at);
3135 ktime_t idle = ns_to_ktime(pkt_dev->idle_acc);
3137 p += sprintf(p, "OK: %llu(c%llu+d%llu) usec, %llu (%dbyte,%dfrags)\n",
3138 (unsigned long long)ktime_to_us(elapsed),
3139 (unsigned long long)ktime_to_us(ktime_sub(elapsed, idle)),
3140 (unsigned long long)ktime_to_us(idle),
3141 (unsigned long long)pkt_dev->sofar,
3142 pkt_dev->cur_pkt_size, nr_frags);
3144 pps = div64_u64(pkt_dev->sofar * NSEC_PER_SEC,
3145 ktime_to_ns(elapsed));
3147 bps = pps * 8 * pkt_dev->cur_pkt_size;
3149 mbps = bps;
3150 do_div(mbps, 1000000);
3151 p += sprintf(p, " %llupps %lluMb/sec (%llubps) errors: %llu",
3152 (unsigned long long)pps,
3153 (unsigned long long)mbps,
3154 (unsigned long long)bps,
3155 (unsigned long long)pkt_dev->errors);
3158 /* Set stopped-at timer, remove from running list, do counters & statistics */
3159 static int pktgen_stop_device(struct pktgen_dev *pkt_dev)
3161 int nr_frags = pkt_dev->skb ? skb_shinfo(pkt_dev->skb)->nr_frags : -1;
3163 if (!pkt_dev->running) {
3164 pr_warning("interface: %s is already stopped\n",
3165 pkt_dev->odevname);
3166 return -EINVAL;
3169 kfree_skb(pkt_dev->skb);
3170 pkt_dev->skb = NULL;
3171 pkt_dev->stopped_at = ktime_get();
3172 pkt_dev->running = 0;
3174 show_results(pkt_dev, nr_frags);
3176 return 0;
3179 static struct pktgen_dev *next_to_run(struct pktgen_thread *t)
3181 struct pktgen_dev *pkt_dev, *best = NULL;
3183 if_lock(t);
3185 list_for_each_entry(pkt_dev, &t->if_list, list) {
3186 if (!pkt_dev->running)
3187 continue;
3188 if (best == NULL)
3189 best = pkt_dev;
3190 else if (ktime_compare(pkt_dev->next_tx, best->next_tx) < 0)
3191 best = pkt_dev;
3193 if_unlock(t);
3194 return best;
3197 static void pktgen_stop(struct pktgen_thread *t)
3199 struct pktgen_dev *pkt_dev;
3201 func_enter();
3203 if_lock(t);
3205 list_for_each_entry(pkt_dev, &t->if_list, list) {
3206 pktgen_stop_device(pkt_dev);
3209 if_unlock(t);
3213 * one of our devices needs to be removed - find it
3214 * and remove it
3216 static void pktgen_rem_one_if(struct pktgen_thread *t)
3218 struct list_head *q, *n;
3219 struct pktgen_dev *cur;
3221 func_enter();
3223 if_lock(t);
3225 list_for_each_safe(q, n, &t->if_list) {
3226 cur = list_entry(q, struct pktgen_dev, list);
3228 if (!cur->removal_mark)
3229 continue;
3231 kfree_skb(cur->skb);
3232 cur->skb = NULL;
3234 pktgen_remove_device(t, cur);
3236 break;
3239 if_unlock(t);
3242 static void pktgen_rem_all_ifs(struct pktgen_thread *t)
3244 struct list_head *q, *n;
3245 struct pktgen_dev *cur;
3247 func_enter();
3249 /* Remove all devices, free mem */
3251 if_lock(t);
3253 list_for_each_safe(q, n, &t->if_list) {
3254 cur = list_entry(q, struct pktgen_dev, list);
3256 kfree_skb(cur->skb);
3257 cur->skb = NULL;
3259 pktgen_remove_device(t, cur);
3262 if_unlock(t);
3265 static void pktgen_rem_thread(struct pktgen_thread *t)
3267 /* Remove from the thread list */
3268 remove_proc_entry(t->tsk->comm, t->net->proc_dir);
3271 static void pktgen_resched(struct pktgen_dev *pkt_dev)
3273 ktime_t idle_start = ktime_get();
3274 schedule();
3275 pkt_dev->idle_acc += ktime_to_ns(ktime_sub(ktime_get(), idle_start));
3278 static void pktgen_wait_for_skb(struct pktgen_dev *pkt_dev)
3280 ktime_t idle_start = ktime_get();
3282 while (atomic_read(&(pkt_dev->skb->users)) != 1) {
3283 if (signal_pending(current))
3284 break;
3286 if (need_resched())
3287 pktgen_resched(pkt_dev);
3288 else
3289 cpu_relax();
3291 pkt_dev->idle_acc += ktime_to_ns(ktime_sub(ktime_get(), idle_start));
3294 static void pktgen_xmit(struct pktgen_dev *pkt_dev)
3296 struct net_device *odev = pkt_dev->odev;
3297 netdev_tx_t (*xmit)(struct sk_buff *, struct net_device *)
3298 = odev->netdev_ops->ndo_start_xmit;
3299 struct netdev_queue *txq;
3300 u16 queue_map;
3301 int ret;
3303 /* If device is offline, then don't send */
3304 if (unlikely(!netif_running(odev) || !netif_carrier_ok(odev))) {
3305 pktgen_stop_device(pkt_dev);
3306 return;
3309 /* This is max DELAY, this has special meaning of
3310 * "never transmit"
3312 if (unlikely(pkt_dev->delay == ULLONG_MAX)) {
3313 pkt_dev->next_tx = ktime_add_ns(ktime_get(), ULONG_MAX);
3314 return;
3317 /* If no skb or clone count exhausted then get new one */
3318 if (!pkt_dev->skb || (pkt_dev->last_ok &&
3319 ++pkt_dev->clone_count >= pkt_dev->clone_skb)) {
3320 /* build a new pkt */
3321 kfree_skb(pkt_dev->skb);
3323 pkt_dev->skb = fill_packet(odev, pkt_dev);
3324 if (pkt_dev->skb == NULL) {
3325 pr_err("ERROR: couldn't allocate skb in fill_packet\n");
3326 schedule();
3327 pkt_dev->clone_count--; /* back out increment, OOM */
3328 return;
3330 pkt_dev->last_pkt_size = pkt_dev->skb->len;
3331 pkt_dev->allocated_skbs++;
3332 pkt_dev->clone_count = 0; /* reset counter */
3335 if (pkt_dev->delay && pkt_dev->last_ok)
3336 spin(pkt_dev, pkt_dev->next_tx);
3338 queue_map = skb_get_queue_mapping(pkt_dev->skb);
3339 txq = netdev_get_tx_queue(odev, queue_map);
3341 __netif_tx_lock_bh(txq);
3343 if (unlikely(netif_xmit_frozen_or_stopped(txq))) {
3344 ret = NETDEV_TX_BUSY;
3345 pkt_dev->last_ok = 0;
3346 goto unlock;
3348 atomic_inc(&(pkt_dev->skb->users));
3349 ret = (*xmit)(pkt_dev->skb, odev);
3351 switch (ret) {
3352 case NETDEV_TX_OK:
3353 txq_trans_update(txq);
3354 pkt_dev->last_ok = 1;
3355 pkt_dev->sofar++;
3356 pkt_dev->seq_num++;
3357 pkt_dev->tx_bytes += pkt_dev->last_pkt_size;
3358 break;
3359 case NET_XMIT_DROP:
3360 case NET_XMIT_CN:
3361 case NET_XMIT_POLICED:
3362 /* skb has been consumed */
3363 pkt_dev->errors++;
3364 break;
3365 default: /* Drivers are not supposed to return other values! */
3366 net_info_ratelimited("%s xmit error: %d\n",
3367 pkt_dev->odevname, ret);
3368 pkt_dev->errors++;
3369 /* fallthru */
3370 case NETDEV_TX_LOCKED:
3371 case NETDEV_TX_BUSY:
3372 /* Retry it next time */
3373 atomic_dec(&(pkt_dev->skb->users));
3374 pkt_dev->last_ok = 0;
3376 unlock:
3377 __netif_tx_unlock_bh(txq);
3379 /* If pkt_dev->count is zero, then run forever */
3380 if ((pkt_dev->count != 0) && (pkt_dev->sofar >= pkt_dev->count)) {
3381 pktgen_wait_for_skb(pkt_dev);
3383 /* Done with this */
3384 pktgen_stop_device(pkt_dev);
3389 * Main loop of the thread goes here
3392 static int pktgen_thread_worker(void *arg)
3394 DEFINE_WAIT(wait);
3395 struct pktgen_thread *t = arg;
3396 struct pktgen_dev *pkt_dev = NULL;
3397 int cpu = t->cpu;
3399 BUG_ON(smp_processor_id() != cpu);
3401 init_waitqueue_head(&t->queue);
3402 complete(&t->start_done);
3404 pr_debug("starting pktgen/%d: pid=%d\n", cpu, task_pid_nr(current));
3406 set_current_state(TASK_INTERRUPTIBLE);
3408 set_freezable();
3410 while (!kthread_should_stop()) {
3411 pkt_dev = next_to_run(t);
3413 if (unlikely(!pkt_dev && t->control == 0)) {
3414 if (t->net->pktgen_exiting)
3415 break;
3416 wait_event_interruptible_timeout(t->queue,
3417 t->control != 0,
3418 HZ/10);
3419 try_to_freeze();
3420 continue;
3423 __set_current_state(TASK_RUNNING);
3425 if (likely(pkt_dev)) {
3426 pktgen_xmit(pkt_dev);
3428 if (need_resched())
3429 pktgen_resched(pkt_dev);
3430 else
3431 cpu_relax();
3434 if (t->control & T_STOP) {
3435 pktgen_stop(t);
3436 t->control &= ~(T_STOP);
3439 if (t->control & T_RUN) {
3440 pktgen_run(t);
3441 t->control &= ~(T_RUN);
3444 if (t->control & T_REMDEVALL) {
3445 pktgen_rem_all_ifs(t);
3446 t->control &= ~(T_REMDEVALL);
3449 if (t->control & T_REMDEV) {
3450 pktgen_rem_one_if(t);
3451 t->control &= ~(T_REMDEV);
3454 try_to_freeze();
3456 set_current_state(TASK_INTERRUPTIBLE);
3459 pr_debug("%s stopping all device\n", t->tsk->comm);
3460 pktgen_stop(t);
3462 pr_debug("%s removing all device\n", t->tsk->comm);
3463 pktgen_rem_all_ifs(t);
3465 pr_debug("%s removing thread\n", t->tsk->comm);
3466 pktgen_rem_thread(t);
3468 /* Wait for kthread_stop */
3469 while (!kthread_should_stop()) {
3470 set_current_state(TASK_INTERRUPTIBLE);
3471 schedule();
3473 __set_current_state(TASK_RUNNING);
3475 return 0;
3478 static struct pktgen_dev *pktgen_find_dev(struct pktgen_thread *t,
3479 const char *ifname, bool exact)
3481 struct pktgen_dev *p, *pkt_dev = NULL;
3482 size_t len = strlen(ifname);
3484 if_lock(t);
3485 list_for_each_entry(p, &t->if_list, list)
3486 if (strncmp(p->odevname, ifname, len) == 0) {
3487 if (p->odevname[len]) {
3488 if (exact || p->odevname[len] != '@')
3489 continue;
3491 pkt_dev = p;
3492 break;
3495 if_unlock(t);
3496 pr_debug("find_dev(%s) returning %p\n", ifname, pkt_dev);
3497 return pkt_dev;
3501 * Adds a dev at front of if_list.
3504 static int add_dev_to_thread(struct pktgen_thread *t,
3505 struct pktgen_dev *pkt_dev)
3507 int rv = 0;
3509 if_lock(t);
3511 if (pkt_dev->pg_thread) {
3512 pr_err("ERROR: already assigned to a thread\n");
3513 rv = -EBUSY;
3514 goto out;
3517 list_add(&pkt_dev->list, &t->if_list);
3518 pkt_dev->pg_thread = t;
3519 pkt_dev->running = 0;
3521 out:
3522 if_unlock(t);
3523 return rv;
3526 /* Called under thread lock */
3528 static int pktgen_add_device(struct pktgen_thread *t, const char *ifname)
3530 struct pktgen_dev *pkt_dev;
3531 int err;
3532 int node = cpu_to_node(t->cpu);
3534 /* We don't allow a device to be on several threads */
3536 pkt_dev = __pktgen_NN_threads(t->net, ifname, FIND);
3537 if (pkt_dev) {
3538 pr_err("ERROR: interface already used\n");
3539 return -EBUSY;
3542 pkt_dev = kzalloc_node(sizeof(struct pktgen_dev), GFP_KERNEL, node);
3543 if (!pkt_dev)
3544 return -ENOMEM;
3546 strcpy(pkt_dev->odevname, ifname);
3547 pkt_dev->flows = vzalloc_node(MAX_CFLOWS * sizeof(struct flow_state),
3548 node);
3549 if (pkt_dev->flows == NULL) {
3550 kfree(pkt_dev);
3551 return -ENOMEM;
3554 pkt_dev->removal_mark = 0;
3555 pkt_dev->nfrags = 0;
3556 pkt_dev->delay = pg_delay_d;
3557 pkt_dev->count = pg_count_d;
3558 pkt_dev->sofar = 0;
3559 pkt_dev->udp_src_min = 9; /* sink port */
3560 pkt_dev->udp_src_max = 9;
3561 pkt_dev->udp_dst_min = 9;
3562 pkt_dev->udp_dst_max = 9;
3563 pkt_dev->vlan_p = 0;
3564 pkt_dev->vlan_cfi = 0;
3565 pkt_dev->vlan_id = 0xffff;
3566 pkt_dev->svlan_p = 0;
3567 pkt_dev->svlan_cfi = 0;
3568 pkt_dev->svlan_id = 0xffff;
3569 pkt_dev->node = -1;
3571 err = pktgen_setup_dev(t->net, pkt_dev, ifname);
3572 if (err)
3573 goto out1;
3574 if (pkt_dev->odev->priv_flags & IFF_TX_SKB_SHARING)
3575 pkt_dev->clone_skb = pg_clone_skb_d;
3577 pkt_dev->entry = proc_create_data(ifname, 0600, t->net->proc_dir,
3578 &pktgen_if_fops, pkt_dev);
3579 if (!pkt_dev->entry) {
3580 pr_err("cannot create %s/%s procfs entry\n",
3581 PG_PROC_DIR, ifname);
3582 err = -EINVAL;
3583 goto out2;
3585 #ifdef CONFIG_XFRM
3586 pkt_dev->ipsmode = XFRM_MODE_TRANSPORT;
3587 pkt_dev->ipsproto = IPPROTO_ESP;
3589 /* xfrm tunnel mode needs additional dst to extract outter
3590 * ip header protocol/ttl/id field, here creat a phony one.
3591 * instead of looking for a valid rt, which definitely hurting
3592 * performance under such circumstance.
3594 pkt_dev->dstops.family = AF_INET;
3595 pkt_dev->dst.dev = pkt_dev->odev;
3596 dst_init_metrics(&pkt_dev->dst, pktgen_dst_metrics, false);
3597 pkt_dev->dst.child = &pkt_dev->dst;
3598 pkt_dev->dst.ops = &pkt_dev->dstops;
3599 #endif
3601 return add_dev_to_thread(t, pkt_dev);
3602 out2:
3603 dev_put(pkt_dev->odev);
3604 out1:
3605 #ifdef CONFIG_XFRM
3606 free_SAs(pkt_dev);
3607 #endif
3608 vfree(pkt_dev->flows);
3609 kfree(pkt_dev);
3610 return err;
3613 static int __net_init pktgen_create_thread(int cpu, struct pktgen_net *pn)
3615 struct pktgen_thread *t;
3616 struct proc_dir_entry *pe;
3617 struct task_struct *p;
3619 t = kzalloc_node(sizeof(struct pktgen_thread), GFP_KERNEL,
3620 cpu_to_node(cpu));
3621 if (!t) {
3622 pr_err("ERROR: out of memory, can't create new thread\n");
3623 return -ENOMEM;
3626 spin_lock_init(&t->if_lock);
3627 t->cpu = cpu;
3629 INIT_LIST_HEAD(&t->if_list);
3631 list_add_tail(&t->th_list, &pn->pktgen_threads);
3632 init_completion(&t->start_done);
3634 p = kthread_create_on_node(pktgen_thread_worker,
3636 cpu_to_node(cpu),
3637 "kpktgend_%d", cpu);
3638 if (IS_ERR(p)) {
3639 pr_err("kernel_thread() failed for cpu %d\n", t->cpu);
3640 list_del(&t->th_list);
3641 kfree(t);
3642 return PTR_ERR(p);
3644 kthread_bind(p, cpu);
3645 t->tsk = p;
3647 pe = proc_create_data(t->tsk->comm, 0600, pn->proc_dir,
3648 &pktgen_thread_fops, t);
3649 if (!pe) {
3650 pr_err("cannot create %s/%s procfs entry\n",
3651 PG_PROC_DIR, t->tsk->comm);
3652 kthread_stop(p);
3653 list_del(&t->th_list);
3654 kfree(t);
3655 return -EINVAL;
3658 t->net = pn;
3659 wake_up_process(p);
3660 wait_for_completion(&t->start_done);
3662 return 0;
3666 * Removes a device from the thread if_list.
3668 static void _rem_dev_from_if_list(struct pktgen_thread *t,
3669 struct pktgen_dev *pkt_dev)
3671 struct list_head *q, *n;
3672 struct pktgen_dev *p;
3674 list_for_each_safe(q, n, &t->if_list) {
3675 p = list_entry(q, struct pktgen_dev, list);
3676 if (p == pkt_dev)
3677 list_del(&p->list);
3681 static int pktgen_remove_device(struct pktgen_thread *t,
3682 struct pktgen_dev *pkt_dev)
3684 pr_debug("remove_device pkt_dev=%p\n", pkt_dev);
3686 if (pkt_dev->running) {
3687 pr_warning("WARNING: trying to remove a running interface, stopping it now\n");
3688 pktgen_stop_device(pkt_dev);
3691 /* Dis-associate from the interface */
3693 if (pkt_dev->odev) {
3694 dev_put(pkt_dev->odev);
3695 pkt_dev->odev = NULL;
3698 /* And update the thread if_list */
3700 _rem_dev_from_if_list(t, pkt_dev);
3702 if (pkt_dev->entry)
3703 proc_remove(pkt_dev->entry);
3705 #ifdef CONFIG_XFRM
3706 free_SAs(pkt_dev);
3707 #endif
3708 vfree(pkt_dev->flows);
3709 if (pkt_dev->page)
3710 put_page(pkt_dev->page);
3711 kfree(pkt_dev);
3712 return 0;
3715 static int __net_init pg_net_init(struct net *net)
3717 struct pktgen_net *pn = net_generic(net, pg_net_id);
3718 struct proc_dir_entry *pe;
3719 int cpu, ret = 0;
3721 pn->net = net;
3722 INIT_LIST_HEAD(&pn->pktgen_threads);
3723 pn->pktgen_exiting = false;
3724 pn->proc_dir = proc_mkdir(PG_PROC_DIR, pn->net->proc_net);
3725 if (!pn->proc_dir) {
3726 pr_warn("cannot create /proc/net/%s\n", PG_PROC_DIR);
3727 return -ENODEV;
3729 pe = proc_create(PGCTRL, 0600, pn->proc_dir, &pktgen_fops);
3730 if (pe == NULL) {
3731 pr_err("cannot create %s procfs entry\n", PGCTRL);
3732 ret = -EINVAL;
3733 goto remove;
3736 for_each_online_cpu(cpu) {
3737 int err;
3739 err = pktgen_create_thread(cpu, pn);
3740 if (err)
3741 pr_warn("Cannot create thread for cpu %d (%d)\n",
3742 cpu, err);
3745 if (list_empty(&pn->pktgen_threads)) {
3746 pr_err("Initialization failed for all threads\n");
3747 ret = -ENODEV;
3748 goto remove_entry;
3751 return 0;
3753 remove_entry:
3754 remove_proc_entry(PGCTRL, pn->proc_dir);
3755 remove:
3756 remove_proc_entry(PG_PROC_DIR, pn->net->proc_net);
3757 return ret;
3760 static void __net_exit pg_net_exit(struct net *net)
3762 struct pktgen_net *pn = net_generic(net, pg_net_id);
3763 struct pktgen_thread *t;
3764 struct list_head *q, *n;
3765 LIST_HEAD(list);
3767 /* Stop all interfaces & threads */
3768 pn->pktgen_exiting = true;
3770 mutex_lock(&pktgen_thread_lock);
3771 list_splice_init(&pn->pktgen_threads, &list);
3772 mutex_unlock(&pktgen_thread_lock);
3774 list_for_each_safe(q, n, &list) {
3775 t = list_entry(q, struct pktgen_thread, th_list);
3776 list_del(&t->th_list);
3777 kthread_stop(t->tsk);
3778 kfree(t);
3781 remove_proc_entry(PGCTRL, pn->proc_dir);
3782 remove_proc_entry(PG_PROC_DIR, pn->net->proc_net);
3785 static struct pernet_operations pg_net_ops = {
3786 .init = pg_net_init,
3787 .exit = pg_net_exit,
3788 .id = &pg_net_id,
3789 .size = sizeof(struct pktgen_net),
3792 static int __init pg_init(void)
3794 int ret = 0;
3796 pr_info("%s", version);
3797 ret = register_pernet_subsys(&pg_net_ops);
3798 if (ret)
3799 return ret;
3800 ret = register_netdevice_notifier(&pktgen_notifier_block);
3801 if (ret)
3802 unregister_pernet_subsys(&pg_net_ops);
3804 return ret;
3807 static void __exit pg_cleanup(void)
3809 unregister_netdevice_notifier(&pktgen_notifier_block);
3810 unregister_pernet_subsys(&pg_net_ops);
3813 module_init(pg_init);
3814 module_exit(pg_cleanup);
3816 MODULE_AUTHOR("Robert Olsson <robert.olsson@its.uu.se>");
3817 MODULE_DESCRIPTION("Packet Generator tool");
3818 MODULE_LICENSE("GPL");
3819 MODULE_VERSION(VERSION);
3820 module_param(pg_count_d, int, 0);
3821 MODULE_PARM_DESC(pg_count_d, "Default number of packets to inject");
3822 module_param(pg_delay_d, int, 0);
3823 MODULE_PARM_DESC(pg_delay_d, "Default delay between packets (nanoseconds)");
3824 module_param(pg_clone_skb_d, int, 0);
3825 MODULE_PARM_DESC(pg_clone_skb_d, "Default number of copies of the same packet");
3826 module_param(debug, int, 0);
3827 MODULE_PARM_DESC(debug, "Enable debugging of pktgen module");