arm64: mm: ensure that the zero page is visible to the page table walker
[linux/fpc-iii.git] / net / core / pktgen.c
blob21a23d97e99c10ba41ac3af05127b43aca751946
1 /*
2 * Authors:
3 * Copyright 2001, 2002 by Robert Olsson <robert.olsson@its.uu.se>
4 * Uppsala University and
5 * Swedish University of Agricultural Sciences
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
8 * Ben Greear <greearb@candelatech.com>
9 * Jens Låås <jens.laas@data.slu.se>
11 * This program is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU General Public License
13 * as published by the Free Software Foundation; either version
14 * 2 of the License, or (at your option) any later version.
17 * A tool for loading the network with preconfigurated packets.
18 * The tool is implemented as a linux module. Parameters are output
19 * device, delay (to hard_xmit), number of packets, and whether
20 * to use multiple SKBs or just the same one.
21 * pktgen uses the installed interface's output routine.
23 * Additional hacking by:
25 * Jens.Laas@data.slu.se
26 * Improved by ANK. 010120.
27 * Improved by ANK even more. 010212.
28 * MAC address typo fixed. 010417 --ro
29 * Integrated. 020301 --DaveM
30 * Added multiskb option 020301 --DaveM
31 * Scaling of results. 020417--sigurdur@linpro.no
32 * Significant re-work of the module:
33 * * Convert to threaded model to more efficiently be able to transmit
34 * and receive on multiple interfaces at once.
35 * * Converted many counters to __u64 to allow longer runs.
36 * * Allow configuration of ranges, like min/max IP address, MACs,
37 * and UDP-ports, for both source and destination, and can
38 * set to use a random distribution or sequentially walk the range.
39 * * Can now change most values after starting.
40 * * Place 12-byte packet in UDP payload with magic number,
41 * sequence number, and timestamp.
42 * * Add receiver code that detects dropped pkts, re-ordered pkts, and
43 * latencies (with micro-second) precision.
44 * * Add IOCTL interface to easily get counters & configuration.
45 * --Ben Greear <greearb@candelatech.com>
47 * Renamed multiskb to clone_skb and cleaned up sending core for two distinct
48 * skb modes. A clone_skb=0 mode for Ben "ranges" work and a clone_skb != 0
49 * as a "fastpath" with a configurable number of clones after alloc's.
50 * clone_skb=0 means all packets are allocated this also means ranges time
51 * stamps etc can be used. clone_skb=100 means 1 malloc is followed by 100
52 * clones.
54 * Also moved to /proc/net/pktgen/
55 * --ro
57 * Sept 10: Fixed threading/locking. Lots of bone-headed and more clever
58 * mistakes. Also merged in DaveM's patch in the -pre6 patch.
59 * --Ben Greear <greearb@candelatech.com>
61 * Integrated to 2.5.x 021029 --Lucio Maciel (luciomaciel@zipmail.com.br)
64 * 021124 Finished major redesign and rewrite for new functionality.
65 * See Documentation/networking/pktgen.txt for how to use this.
67 * The new operation:
68 * For each CPU one thread/process is created at start. This process checks
69 * for running devices in the if_list and sends packets until count is 0 it
70 * also the thread checks the thread->control which is used for inter-process
71 * communication. controlling process "posts" operations to the threads this
72 * way. The if_lock should be possible to remove when add/rem_device is merged
73 * into this too.
75 * By design there should only be *one* "controlling" process. In practice
76 * multiple write accesses gives unpredictable result. Understood by "write"
77 * to /proc gives result code thats should be read be the "writer".
78 * For practical use this should be no problem.
80 * Note when adding devices to a specific CPU there good idea to also assign
81 * /proc/irq/XX/smp_affinity so TX-interrupts gets bound to the same CPU.
82 * --ro
84 * Fix refcount off by one if first packet fails, potential null deref,
85 * memleak 030710- KJP
87 * First "ranges" functionality for ipv6 030726 --ro
89 * Included flow support. 030802 ANK.
91 * Fixed unaligned access on IA-64 Grant Grundler <grundler@parisc-linux.org>
93 * Remove if fix from added Harald Welte <laforge@netfilter.org> 040419
94 * ia64 compilation fix from Aron Griffis <aron@hp.com> 040604
96 * New xmit() return, do_div and misc clean up by Stephen Hemminger
97 * <shemminger@osdl.org> 040923
99 * Randy Dunlap fixed u64 printk compiler waring
101 * Remove FCS from BW calculation. Lennert Buytenhek <buytenh@wantstofly.org>
102 * New time handling. Lennert Buytenhek <buytenh@wantstofly.org> 041213
104 * Corrections from Nikolai Malykh (nmalykh@bilim.com)
105 * Removed unused flags F_SET_SRCMAC & F_SET_SRCIP 041230
107 * interruptible_sleep_on_timeout() replaced Nishanth Aravamudan <nacc@us.ibm.com>
108 * 050103
110 * MPLS support by Steven Whitehouse <steve@chygwyn.com>
112 * 802.1Q/Q-in-Q support by Francesco Fondelli (FF) <francesco.fondelli@gmail.com>
114 * Fixed src_mac command to set source mac of packet to value specified in
115 * command by Adit Ranadive <adit.262@gmail.com>
119 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
121 #include <linux/sys.h>
122 #include <linux/types.h>
123 #include <linux/module.h>
124 #include <linux/moduleparam.h>
125 #include <linux/kernel.h>
126 #include <linux/mutex.h>
127 #include <linux/sched.h>
128 #include <linux/slab.h>
129 #include <linux/vmalloc.h>
130 #include <linux/unistd.h>
131 #include <linux/string.h>
132 #include <linux/ptrace.h>
133 #include <linux/errno.h>
134 #include <linux/ioport.h>
135 #include <linux/interrupt.h>
136 #include <linux/capability.h>
137 #include <linux/hrtimer.h>
138 #include <linux/freezer.h>
139 #include <linux/delay.h>
140 #include <linux/timer.h>
141 #include <linux/list.h>
142 #include <linux/init.h>
143 #include <linux/skbuff.h>
144 #include <linux/netdevice.h>
145 #include <linux/inet.h>
146 #include <linux/inetdevice.h>
147 #include <linux/rtnetlink.h>
148 #include <linux/if_arp.h>
149 #include <linux/if_vlan.h>
150 #include <linux/in.h>
151 #include <linux/ip.h>
152 #include <linux/ipv6.h>
153 #include <linux/udp.h>
154 #include <linux/proc_fs.h>
155 #include <linux/seq_file.h>
156 #include <linux/wait.h>
157 #include <linux/etherdevice.h>
158 #include <linux/kthread.h>
159 #include <linux/prefetch.h>
160 #include <net/net_namespace.h>
161 #include <net/checksum.h>
162 #include <net/ipv6.h>
163 #include <net/addrconf.h>
164 #ifdef CONFIG_XFRM
165 #include <net/xfrm.h>
166 #endif
167 #include <net/netns/generic.h>
168 #include <asm/byteorder.h>
169 #include <linux/rcupdate.h>
170 #include <linux/bitops.h>
171 #include <linux/io.h>
172 #include <linux/timex.h>
173 #include <linux/uaccess.h>
174 #include <asm/dma.h>
175 #include <asm/div64.h> /* do_div */
177 #define VERSION "2.74"
178 #define IP_NAME_SZ 32
179 #define MAX_MPLS_LABELS 16 /* This is the max label stack depth */
180 #define MPLS_STACK_BOTTOM htonl(0x00000100)
182 #define func_enter() pr_debug("entering %s\n", __func__);
184 /* Device flag bits */
185 #define F_IPSRC_RND (1<<0) /* IP-Src Random */
186 #define F_IPDST_RND (1<<1) /* IP-Dst Random */
187 #define F_UDPSRC_RND (1<<2) /* UDP-Src Random */
188 #define F_UDPDST_RND (1<<3) /* UDP-Dst Random */
189 #define F_MACSRC_RND (1<<4) /* MAC-Src Random */
190 #define F_MACDST_RND (1<<5) /* MAC-Dst Random */
191 #define F_TXSIZE_RND (1<<6) /* Transmit size is random */
192 #define F_IPV6 (1<<7) /* Interface in IPV6 Mode */
193 #define F_MPLS_RND (1<<8) /* Random MPLS labels */
194 #define F_VID_RND (1<<9) /* Random VLAN ID */
195 #define F_SVID_RND (1<<10) /* Random SVLAN ID */
196 #define F_FLOW_SEQ (1<<11) /* Sequential flows */
197 #define F_IPSEC_ON (1<<12) /* ipsec on for flows */
198 #define F_QUEUE_MAP_RND (1<<13) /* queue map Random */
199 #define F_QUEUE_MAP_CPU (1<<14) /* queue map mirrors smp_processor_id() */
200 #define F_NODE (1<<15) /* Node memory alloc*/
202 /* Thread control flag bits */
203 #define T_STOP (1<<0) /* Stop run */
204 #define T_RUN (1<<1) /* Start run */
205 #define T_REMDEVALL (1<<2) /* Remove all devs */
206 #define T_REMDEV (1<<3) /* Remove one dev */
208 /* If lock -- can be removed after some work */
209 #define if_lock(t) spin_lock(&(t->if_lock));
210 #define if_unlock(t) spin_unlock(&(t->if_lock));
212 /* Used to help with determining the pkts on receive */
213 #define PKTGEN_MAGIC 0xbe9be955
214 #define PG_PROC_DIR "pktgen"
215 #define PGCTRL "pgctrl"
217 #define MAX_CFLOWS 65536
219 #define VLAN_TAG_SIZE(x) ((x)->vlan_id == 0xffff ? 0 : 4)
220 #define SVLAN_TAG_SIZE(x) ((x)->svlan_id == 0xffff ? 0 : 4)
222 struct flow_state {
223 __be32 cur_daddr;
224 int count;
225 #ifdef CONFIG_XFRM
226 struct xfrm_state *x;
227 #endif
228 __u32 flags;
231 /* flow flag bits */
232 #define F_INIT (1<<0) /* flow has been initialized */
234 struct pktgen_dev {
236 * Try to keep frequent/infrequent used vars. separated.
238 struct proc_dir_entry *entry; /* proc file */
239 struct pktgen_thread *pg_thread;/* the owner */
240 struct list_head list; /* chaining in the thread's run-queue */
242 int running; /* if false, the test will stop */
244 /* If min != max, then we will either do a linear iteration, or
245 * we will do a random selection from within the range.
247 __u32 flags;
248 int removal_mark; /* non-zero => the device is marked for
249 * removal by worker thread */
251 int min_pkt_size;
252 int max_pkt_size;
253 int pkt_overhead; /* overhead for MPLS, VLANs, IPSEC etc */
254 int nfrags;
255 struct page *page;
256 u64 delay; /* nano-seconds */
258 __u64 count; /* Default No packets to send */
259 __u64 sofar; /* How many pkts we've sent so far */
260 __u64 tx_bytes; /* How many bytes we've transmitted */
261 __u64 errors; /* Errors when trying to transmit, */
263 /* runtime counters relating to clone_skb */
265 __u64 allocated_skbs;
266 __u32 clone_count;
267 int last_ok; /* Was last skb sent?
268 * Or a failed transmit of some sort?
269 * This will keep sequence numbers in order
271 ktime_t next_tx;
272 ktime_t started_at;
273 ktime_t stopped_at;
274 u64 idle_acc; /* nano-seconds */
276 __u32 seq_num;
278 int clone_skb; /*
279 * Use multiple SKBs during packet gen.
280 * If this number is greater than 1, then
281 * that many copies of the same packet will be
282 * sent before a new packet is allocated.
283 * If you want to send 1024 identical packets
284 * before creating a new packet,
285 * set clone_skb to 1024.
288 char dst_min[IP_NAME_SZ]; /* IP, ie 1.2.3.4 */
289 char dst_max[IP_NAME_SZ]; /* IP, ie 1.2.3.4 */
290 char src_min[IP_NAME_SZ]; /* IP, ie 1.2.3.4 */
291 char src_max[IP_NAME_SZ]; /* IP, ie 1.2.3.4 */
293 struct in6_addr in6_saddr;
294 struct in6_addr in6_daddr;
295 struct in6_addr cur_in6_daddr;
296 struct in6_addr cur_in6_saddr;
297 /* For ranges */
298 struct in6_addr min_in6_daddr;
299 struct in6_addr max_in6_daddr;
300 struct in6_addr min_in6_saddr;
301 struct in6_addr max_in6_saddr;
303 /* If we're doing ranges, random or incremental, then this
304 * defines the min/max for those ranges.
306 __be32 saddr_min; /* inclusive, source IP address */
307 __be32 saddr_max; /* exclusive, source IP address */
308 __be32 daddr_min; /* inclusive, dest IP address */
309 __be32 daddr_max; /* exclusive, dest IP address */
311 __u16 udp_src_min; /* inclusive, source UDP port */
312 __u16 udp_src_max; /* exclusive, source UDP port */
313 __u16 udp_dst_min; /* inclusive, dest UDP port */
314 __u16 udp_dst_max; /* exclusive, dest UDP port */
316 /* DSCP + ECN */
317 __u8 tos; /* six MSB of (former) IPv4 TOS
318 are for dscp codepoint */
319 __u8 traffic_class; /* ditto for the (former) Traffic Class in IPv6
320 (see RFC 3260, sec. 4) */
322 /* MPLS */
323 unsigned int nr_labels; /* Depth of stack, 0 = no MPLS */
324 __be32 labels[MAX_MPLS_LABELS];
326 /* VLAN/SVLAN (802.1Q/Q-in-Q) */
327 __u8 vlan_p;
328 __u8 vlan_cfi;
329 __u16 vlan_id; /* 0xffff means no vlan tag */
331 __u8 svlan_p;
332 __u8 svlan_cfi;
333 __u16 svlan_id; /* 0xffff means no svlan tag */
335 __u32 src_mac_count; /* How many MACs to iterate through */
336 __u32 dst_mac_count; /* How many MACs to iterate through */
338 unsigned char dst_mac[ETH_ALEN];
339 unsigned char src_mac[ETH_ALEN];
341 __u32 cur_dst_mac_offset;
342 __u32 cur_src_mac_offset;
343 __be32 cur_saddr;
344 __be32 cur_daddr;
345 __u16 ip_id;
346 __u16 cur_udp_dst;
347 __u16 cur_udp_src;
348 __u16 cur_queue_map;
349 __u32 cur_pkt_size;
350 __u32 last_pkt_size;
352 __u8 hh[14];
353 /* = {
354 0x00, 0x80, 0xC8, 0x79, 0xB3, 0xCB,
356 We fill in SRC address later
357 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
358 0x08, 0x00
361 __u16 pad; /* pad out the hh struct to an even 16 bytes */
363 struct sk_buff *skb; /* skb we are to transmit next, used for when we
364 * are transmitting the same one multiple times
366 struct net_device *odev; /* The out-going device.
367 * Note that the device should have it's
368 * pg_info pointer pointing back to this
369 * device.
370 * Set when the user specifies the out-going
371 * device name (not when the inject is
372 * started as it used to do.)
374 char odevname[32];
375 struct flow_state *flows;
376 unsigned int cflows; /* Concurrent flows (config) */
377 unsigned int lflow; /* Flow length (config) */
378 unsigned int nflows; /* accumulated flows (stats) */
379 unsigned int curfl; /* current sequenced flow (state)*/
381 u16 queue_map_min;
382 u16 queue_map_max;
383 __u32 skb_priority; /* skb priority field */
384 int node; /* Memory node */
386 #ifdef CONFIG_XFRM
387 __u8 ipsmode; /* IPSEC mode (config) */
388 __u8 ipsproto; /* IPSEC type (config) */
389 #endif
390 char result[512];
393 struct pktgen_hdr {
394 __be32 pgh_magic;
395 __be32 seq_num;
396 __be32 tv_sec;
397 __be32 tv_usec;
401 static int pg_net_id __read_mostly;
403 struct pktgen_net {
404 struct net *net;
405 struct proc_dir_entry *proc_dir;
406 struct list_head pktgen_threads;
407 bool pktgen_exiting;
410 struct pktgen_thread {
411 spinlock_t if_lock; /* for list of devices */
412 struct list_head if_list; /* All device here */
413 struct list_head th_list;
414 struct task_struct *tsk;
415 char result[512];
417 /* Field for thread to receive "posted" events terminate,
418 stop ifs etc. */
420 u32 control;
421 int cpu;
423 wait_queue_head_t queue;
424 struct completion start_done;
425 struct pktgen_net *net;
428 #define REMOVE 1
429 #define FIND 0
431 static const char version[] =
432 "Packet Generator for packet performance testing. "
433 "Version: " VERSION "\n";
435 static int pktgen_remove_device(struct pktgen_thread *t, struct pktgen_dev *i);
436 static int pktgen_add_device(struct pktgen_thread *t, const char *ifname);
437 static struct pktgen_dev *pktgen_find_dev(struct pktgen_thread *t,
438 const char *ifname, bool exact);
439 static int pktgen_device_event(struct notifier_block *, unsigned long, void *);
440 static void pktgen_run_all_threads(struct pktgen_net *pn);
441 static void pktgen_reset_all_threads(struct pktgen_net *pn);
442 static void pktgen_stop_all_threads_ifs(struct pktgen_net *pn);
444 static void pktgen_stop(struct pktgen_thread *t);
445 static void pktgen_clear_counters(struct pktgen_dev *pkt_dev);
447 /* Module parameters, defaults. */
448 static int pg_count_d __read_mostly = 1000;
449 static int pg_delay_d __read_mostly;
450 static int pg_clone_skb_d __read_mostly;
451 static int debug __read_mostly;
453 static DEFINE_MUTEX(pktgen_thread_lock);
455 static struct notifier_block pktgen_notifier_block = {
456 .notifier_call = pktgen_device_event,
460 * /proc handling functions
464 static int pgctrl_show(struct seq_file *seq, void *v)
466 seq_puts(seq, version);
467 return 0;
470 static ssize_t pgctrl_write(struct file *file, const char __user *buf,
471 size_t count, loff_t *ppos)
473 int err = 0;
474 char data[128];
475 struct pktgen_net *pn = net_generic(current->nsproxy->net_ns, pg_net_id);
477 if (!capable(CAP_NET_ADMIN)) {
478 err = -EPERM;
479 goto out;
482 if (count > sizeof(data))
483 count = sizeof(data);
485 if (copy_from_user(data, buf, count)) {
486 err = -EFAULT;
487 goto out;
489 data[count - 1] = 0; /* Make string */
491 if (!strcmp(data, "stop"))
492 pktgen_stop_all_threads_ifs(pn);
494 else if (!strcmp(data, "start"))
495 pktgen_run_all_threads(pn);
497 else if (!strcmp(data, "reset"))
498 pktgen_reset_all_threads(pn);
500 else
501 pr_warning("Unknown command: %s\n", data);
503 err = count;
505 out:
506 return err;
509 static int pgctrl_open(struct inode *inode, struct file *file)
511 return single_open(file, pgctrl_show, PDE_DATA(inode));
514 static const struct file_operations pktgen_fops = {
515 .owner = THIS_MODULE,
516 .open = pgctrl_open,
517 .read = seq_read,
518 .llseek = seq_lseek,
519 .write = pgctrl_write,
520 .release = single_release,
523 static int pktgen_if_show(struct seq_file *seq, void *v)
525 const struct pktgen_dev *pkt_dev = seq->private;
526 ktime_t stopped;
527 u64 idle;
529 seq_printf(seq,
530 "Params: count %llu min_pkt_size: %u max_pkt_size: %u\n",
531 (unsigned long long)pkt_dev->count, pkt_dev->min_pkt_size,
532 pkt_dev->max_pkt_size);
534 seq_printf(seq,
535 " frags: %d delay: %llu clone_skb: %d ifname: %s\n",
536 pkt_dev->nfrags, (unsigned long long) pkt_dev->delay,
537 pkt_dev->clone_skb, pkt_dev->odevname);
539 seq_printf(seq, " flows: %u flowlen: %u\n", pkt_dev->cflows,
540 pkt_dev->lflow);
542 seq_printf(seq,
543 " queue_map_min: %u queue_map_max: %u\n",
544 pkt_dev->queue_map_min,
545 pkt_dev->queue_map_max);
547 if (pkt_dev->skb_priority)
548 seq_printf(seq, " skb_priority: %u\n",
549 pkt_dev->skb_priority);
551 if (pkt_dev->flags & F_IPV6) {
552 seq_printf(seq,
553 " saddr: %pI6c min_saddr: %pI6c max_saddr: %pI6c\n"
554 " daddr: %pI6c min_daddr: %pI6c max_daddr: %pI6c\n",
555 &pkt_dev->in6_saddr,
556 &pkt_dev->min_in6_saddr, &pkt_dev->max_in6_saddr,
557 &pkt_dev->in6_daddr,
558 &pkt_dev->min_in6_daddr, &pkt_dev->max_in6_daddr);
559 } else {
560 seq_printf(seq,
561 " dst_min: %s dst_max: %s\n",
562 pkt_dev->dst_min, pkt_dev->dst_max);
563 seq_printf(seq,
564 " src_min: %s src_max: %s\n",
565 pkt_dev->src_min, pkt_dev->src_max);
568 seq_puts(seq, " src_mac: ");
570 seq_printf(seq, "%pM ",
571 is_zero_ether_addr(pkt_dev->src_mac) ?
572 pkt_dev->odev->dev_addr : pkt_dev->src_mac);
574 seq_printf(seq, "dst_mac: ");
575 seq_printf(seq, "%pM\n", pkt_dev->dst_mac);
577 seq_printf(seq,
578 " udp_src_min: %d udp_src_max: %d"
579 " udp_dst_min: %d udp_dst_max: %d\n",
580 pkt_dev->udp_src_min, pkt_dev->udp_src_max,
581 pkt_dev->udp_dst_min, pkt_dev->udp_dst_max);
583 seq_printf(seq,
584 " src_mac_count: %d dst_mac_count: %d\n",
585 pkt_dev->src_mac_count, pkt_dev->dst_mac_count);
587 if (pkt_dev->nr_labels) {
588 unsigned int i;
589 seq_printf(seq, " mpls: ");
590 for (i = 0; i < pkt_dev->nr_labels; i++)
591 seq_printf(seq, "%08x%s", ntohl(pkt_dev->labels[i]),
592 i == pkt_dev->nr_labels-1 ? "\n" : ", ");
595 if (pkt_dev->vlan_id != 0xffff)
596 seq_printf(seq, " vlan_id: %u vlan_p: %u vlan_cfi: %u\n",
597 pkt_dev->vlan_id, pkt_dev->vlan_p,
598 pkt_dev->vlan_cfi);
600 if (pkt_dev->svlan_id != 0xffff)
601 seq_printf(seq, " svlan_id: %u vlan_p: %u vlan_cfi: %u\n",
602 pkt_dev->svlan_id, pkt_dev->svlan_p,
603 pkt_dev->svlan_cfi);
605 if (pkt_dev->tos)
606 seq_printf(seq, " tos: 0x%02x\n", pkt_dev->tos);
608 if (pkt_dev->traffic_class)
609 seq_printf(seq, " traffic_class: 0x%02x\n", pkt_dev->traffic_class);
611 if (pkt_dev->node >= 0)
612 seq_printf(seq, " node: %d\n", pkt_dev->node);
614 seq_printf(seq, " Flags: ");
616 if (pkt_dev->flags & F_IPV6)
617 seq_printf(seq, "IPV6 ");
619 if (pkt_dev->flags & F_IPSRC_RND)
620 seq_printf(seq, "IPSRC_RND ");
622 if (pkt_dev->flags & F_IPDST_RND)
623 seq_printf(seq, "IPDST_RND ");
625 if (pkt_dev->flags & F_TXSIZE_RND)
626 seq_printf(seq, "TXSIZE_RND ");
628 if (pkt_dev->flags & F_UDPSRC_RND)
629 seq_printf(seq, "UDPSRC_RND ");
631 if (pkt_dev->flags & F_UDPDST_RND)
632 seq_printf(seq, "UDPDST_RND ");
634 if (pkt_dev->flags & F_MPLS_RND)
635 seq_printf(seq, "MPLS_RND ");
637 if (pkt_dev->flags & F_QUEUE_MAP_RND)
638 seq_printf(seq, "QUEUE_MAP_RND ");
640 if (pkt_dev->flags & F_QUEUE_MAP_CPU)
641 seq_printf(seq, "QUEUE_MAP_CPU ");
643 if (pkt_dev->cflows) {
644 if (pkt_dev->flags & F_FLOW_SEQ)
645 seq_printf(seq, "FLOW_SEQ "); /*in sequence flows*/
646 else
647 seq_printf(seq, "FLOW_RND ");
650 #ifdef CONFIG_XFRM
651 if (pkt_dev->flags & F_IPSEC_ON)
652 seq_printf(seq, "IPSEC ");
653 #endif
655 if (pkt_dev->flags & F_MACSRC_RND)
656 seq_printf(seq, "MACSRC_RND ");
658 if (pkt_dev->flags & F_MACDST_RND)
659 seq_printf(seq, "MACDST_RND ");
661 if (pkt_dev->flags & F_VID_RND)
662 seq_printf(seq, "VID_RND ");
664 if (pkt_dev->flags & F_SVID_RND)
665 seq_printf(seq, "SVID_RND ");
667 if (pkt_dev->flags & F_NODE)
668 seq_printf(seq, "NODE_ALLOC ");
670 seq_puts(seq, "\n");
672 /* not really stopped, more like last-running-at */
673 stopped = pkt_dev->running ? ktime_get() : pkt_dev->stopped_at;
674 idle = pkt_dev->idle_acc;
675 do_div(idle, NSEC_PER_USEC);
677 seq_printf(seq,
678 "Current:\n pkts-sofar: %llu errors: %llu\n",
679 (unsigned long long)pkt_dev->sofar,
680 (unsigned long long)pkt_dev->errors);
682 seq_printf(seq,
683 " started: %lluus stopped: %lluus idle: %lluus\n",
684 (unsigned long long) ktime_to_us(pkt_dev->started_at),
685 (unsigned long long) ktime_to_us(stopped),
686 (unsigned long long) idle);
688 seq_printf(seq,
689 " seq_num: %d cur_dst_mac_offset: %d cur_src_mac_offset: %d\n",
690 pkt_dev->seq_num, pkt_dev->cur_dst_mac_offset,
691 pkt_dev->cur_src_mac_offset);
693 if (pkt_dev->flags & F_IPV6) {
694 seq_printf(seq, " cur_saddr: %pI6c cur_daddr: %pI6c\n",
695 &pkt_dev->cur_in6_saddr,
696 &pkt_dev->cur_in6_daddr);
697 } else
698 seq_printf(seq, " cur_saddr: %pI4 cur_daddr: %pI4\n",
699 &pkt_dev->cur_saddr, &pkt_dev->cur_daddr);
701 seq_printf(seq, " cur_udp_dst: %d cur_udp_src: %d\n",
702 pkt_dev->cur_udp_dst, pkt_dev->cur_udp_src);
704 seq_printf(seq, " cur_queue_map: %u\n", pkt_dev->cur_queue_map);
706 seq_printf(seq, " flows: %u\n", pkt_dev->nflows);
708 if (pkt_dev->result[0])
709 seq_printf(seq, "Result: %s\n", pkt_dev->result);
710 else
711 seq_printf(seq, "Result: Idle\n");
713 return 0;
717 static int hex32_arg(const char __user *user_buffer, unsigned long maxlen,
718 __u32 *num)
720 int i = 0;
721 *num = 0;
723 for (; i < maxlen; i++) {
724 int value;
725 char c;
726 *num <<= 4;
727 if (get_user(c, &user_buffer[i]))
728 return -EFAULT;
729 value = hex_to_bin(c);
730 if (value >= 0)
731 *num |= value;
732 else
733 break;
735 return i;
738 static int count_trail_chars(const char __user * user_buffer,
739 unsigned int maxlen)
741 int i;
743 for (i = 0; i < maxlen; i++) {
744 char c;
745 if (get_user(c, &user_buffer[i]))
746 return -EFAULT;
747 switch (c) {
748 case '\"':
749 case '\n':
750 case '\r':
751 case '\t':
752 case ' ':
753 case '=':
754 break;
755 default:
756 goto done;
759 done:
760 return i;
763 static long num_arg(const char __user *user_buffer, unsigned long maxlen,
764 unsigned long *num)
766 int i;
767 *num = 0;
769 for (i = 0; i < maxlen; i++) {
770 char c;
771 if (get_user(c, &user_buffer[i]))
772 return -EFAULT;
773 if ((c >= '0') && (c <= '9')) {
774 *num *= 10;
775 *num += c - '0';
776 } else
777 break;
779 return i;
782 static int strn_len(const char __user * user_buffer, unsigned int maxlen)
784 int i;
786 for (i = 0; i < maxlen; i++) {
787 char c;
788 if (get_user(c, &user_buffer[i]))
789 return -EFAULT;
790 switch (c) {
791 case '\"':
792 case '\n':
793 case '\r':
794 case '\t':
795 case ' ':
796 goto done_str;
797 break;
798 default:
799 break;
802 done_str:
803 return i;
806 static ssize_t get_labels(const char __user *buffer, struct pktgen_dev *pkt_dev)
808 unsigned int n = 0;
809 char c;
810 ssize_t i = 0;
811 int len;
813 pkt_dev->nr_labels = 0;
814 do {
815 __u32 tmp;
816 len = hex32_arg(&buffer[i], 8, &tmp);
817 if (len <= 0)
818 return len;
819 pkt_dev->labels[n] = htonl(tmp);
820 if (pkt_dev->labels[n] & MPLS_STACK_BOTTOM)
821 pkt_dev->flags |= F_MPLS_RND;
822 i += len;
823 if (get_user(c, &buffer[i]))
824 return -EFAULT;
825 i++;
826 n++;
827 if (n >= MAX_MPLS_LABELS)
828 return -E2BIG;
829 } while (c == ',');
831 pkt_dev->nr_labels = n;
832 return i;
835 static ssize_t pktgen_if_write(struct file *file,
836 const char __user * user_buffer, size_t count,
837 loff_t * offset)
839 struct seq_file *seq = file->private_data;
840 struct pktgen_dev *pkt_dev = seq->private;
841 int i, max, len;
842 char name[16], valstr[32];
843 unsigned long value = 0;
844 char *pg_result = NULL;
845 int tmp = 0;
846 char buf[128];
848 pg_result = &(pkt_dev->result[0]);
850 if (count < 1) {
851 pr_warning("wrong command format\n");
852 return -EINVAL;
855 max = count;
856 tmp = count_trail_chars(user_buffer, max);
857 if (tmp < 0) {
858 pr_warning("illegal format\n");
859 return tmp;
861 i = tmp;
863 /* Read variable name */
865 len = strn_len(&user_buffer[i], sizeof(name) - 1);
866 if (len < 0)
867 return len;
869 memset(name, 0, sizeof(name));
870 if (copy_from_user(name, &user_buffer[i], len))
871 return -EFAULT;
872 i += len;
874 max = count - i;
875 len = count_trail_chars(&user_buffer[i], max);
876 if (len < 0)
877 return len;
879 i += len;
881 if (debug) {
882 size_t copy = min_t(size_t, count, 1023);
883 char tb[copy + 1];
884 if (copy_from_user(tb, user_buffer, copy))
885 return -EFAULT;
886 tb[copy] = 0;
887 pr_debug("%s,%lu buffer -:%s:-\n",
888 name, (unsigned long)count, tb);
891 if (!strcmp(name, "min_pkt_size")) {
892 len = num_arg(&user_buffer[i], 10, &value);
893 if (len < 0)
894 return len;
896 i += len;
897 if (value < 14 + 20 + 8)
898 value = 14 + 20 + 8;
899 if (value != pkt_dev->min_pkt_size) {
900 pkt_dev->min_pkt_size = value;
901 pkt_dev->cur_pkt_size = value;
903 sprintf(pg_result, "OK: min_pkt_size=%u",
904 pkt_dev->min_pkt_size);
905 return count;
908 if (!strcmp(name, "max_pkt_size")) {
909 len = num_arg(&user_buffer[i], 10, &value);
910 if (len < 0)
911 return len;
913 i += len;
914 if (value < 14 + 20 + 8)
915 value = 14 + 20 + 8;
916 if (value != pkt_dev->max_pkt_size) {
917 pkt_dev->max_pkt_size = value;
918 pkt_dev->cur_pkt_size = value;
920 sprintf(pg_result, "OK: max_pkt_size=%u",
921 pkt_dev->max_pkt_size);
922 return count;
925 /* Shortcut for min = max */
927 if (!strcmp(name, "pkt_size")) {
928 len = num_arg(&user_buffer[i], 10, &value);
929 if (len < 0)
930 return len;
932 i += len;
933 if (value < 14 + 20 + 8)
934 value = 14 + 20 + 8;
935 if (value != pkt_dev->min_pkt_size) {
936 pkt_dev->min_pkt_size = value;
937 pkt_dev->max_pkt_size = value;
938 pkt_dev->cur_pkt_size = value;
940 sprintf(pg_result, "OK: pkt_size=%u", pkt_dev->min_pkt_size);
941 return count;
944 if (!strcmp(name, "debug")) {
945 len = num_arg(&user_buffer[i], 10, &value);
946 if (len < 0)
947 return len;
949 i += len;
950 debug = value;
951 sprintf(pg_result, "OK: debug=%u", debug);
952 return count;
955 if (!strcmp(name, "frags")) {
956 len = num_arg(&user_buffer[i], 10, &value);
957 if (len < 0)
958 return len;
960 i += len;
961 pkt_dev->nfrags = value;
962 sprintf(pg_result, "OK: frags=%u", pkt_dev->nfrags);
963 return count;
965 if (!strcmp(name, "delay")) {
966 len = num_arg(&user_buffer[i], 10, &value);
967 if (len < 0)
968 return len;
970 i += len;
971 if (value == 0x7FFFFFFF)
972 pkt_dev->delay = ULLONG_MAX;
973 else
974 pkt_dev->delay = (u64)value;
976 sprintf(pg_result, "OK: delay=%llu",
977 (unsigned long long) pkt_dev->delay);
978 return count;
980 if (!strcmp(name, "rate")) {
981 len = num_arg(&user_buffer[i], 10, &value);
982 if (len < 0)
983 return len;
985 i += len;
986 if (!value)
987 return len;
988 pkt_dev->delay = pkt_dev->min_pkt_size*8*NSEC_PER_USEC/value;
989 if (debug)
990 pr_info("Delay set at: %llu ns\n", pkt_dev->delay);
992 sprintf(pg_result, "OK: rate=%lu", value);
993 return count;
995 if (!strcmp(name, "ratep")) {
996 len = num_arg(&user_buffer[i], 10, &value);
997 if (len < 0)
998 return len;
1000 i += len;
1001 if (!value)
1002 return len;
1003 pkt_dev->delay = NSEC_PER_SEC/value;
1004 if (debug)
1005 pr_info("Delay set at: %llu ns\n", pkt_dev->delay);
1007 sprintf(pg_result, "OK: rate=%lu", value);
1008 return count;
1010 if (!strcmp(name, "udp_src_min")) {
1011 len = num_arg(&user_buffer[i], 10, &value);
1012 if (len < 0)
1013 return len;
1015 i += len;
1016 if (value != pkt_dev->udp_src_min) {
1017 pkt_dev->udp_src_min = value;
1018 pkt_dev->cur_udp_src = value;
1020 sprintf(pg_result, "OK: udp_src_min=%u", pkt_dev->udp_src_min);
1021 return count;
1023 if (!strcmp(name, "udp_dst_min")) {
1024 len = num_arg(&user_buffer[i], 10, &value);
1025 if (len < 0)
1026 return len;
1028 i += len;
1029 if (value != pkt_dev->udp_dst_min) {
1030 pkt_dev->udp_dst_min = value;
1031 pkt_dev->cur_udp_dst = value;
1033 sprintf(pg_result, "OK: udp_dst_min=%u", pkt_dev->udp_dst_min);
1034 return count;
1036 if (!strcmp(name, "udp_src_max")) {
1037 len = num_arg(&user_buffer[i], 10, &value);
1038 if (len < 0)
1039 return len;
1041 i += len;
1042 if (value != pkt_dev->udp_src_max) {
1043 pkt_dev->udp_src_max = value;
1044 pkt_dev->cur_udp_src = value;
1046 sprintf(pg_result, "OK: udp_src_max=%u", pkt_dev->udp_src_max);
1047 return count;
1049 if (!strcmp(name, "udp_dst_max")) {
1050 len = num_arg(&user_buffer[i], 10, &value);
1051 if (len < 0)
1052 return len;
1054 i += len;
1055 if (value != pkt_dev->udp_dst_max) {
1056 pkt_dev->udp_dst_max = value;
1057 pkt_dev->cur_udp_dst = value;
1059 sprintf(pg_result, "OK: udp_dst_max=%u", pkt_dev->udp_dst_max);
1060 return count;
1062 if (!strcmp(name, "clone_skb")) {
1063 len = num_arg(&user_buffer[i], 10, &value);
1064 if (len < 0)
1065 return len;
1066 if ((value > 0) &&
1067 (!(pkt_dev->odev->priv_flags & IFF_TX_SKB_SHARING)))
1068 return -ENOTSUPP;
1069 i += len;
1070 pkt_dev->clone_skb = value;
1072 sprintf(pg_result, "OK: clone_skb=%d", pkt_dev->clone_skb);
1073 return count;
1075 if (!strcmp(name, "count")) {
1076 len = num_arg(&user_buffer[i], 10, &value);
1077 if (len < 0)
1078 return len;
1080 i += len;
1081 pkt_dev->count = value;
1082 sprintf(pg_result, "OK: count=%llu",
1083 (unsigned long long)pkt_dev->count);
1084 return count;
1086 if (!strcmp(name, "src_mac_count")) {
1087 len = num_arg(&user_buffer[i], 10, &value);
1088 if (len < 0)
1089 return len;
1091 i += len;
1092 if (pkt_dev->src_mac_count != value) {
1093 pkt_dev->src_mac_count = value;
1094 pkt_dev->cur_src_mac_offset = 0;
1096 sprintf(pg_result, "OK: src_mac_count=%d",
1097 pkt_dev->src_mac_count);
1098 return count;
1100 if (!strcmp(name, "dst_mac_count")) {
1101 len = num_arg(&user_buffer[i], 10, &value);
1102 if (len < 0)
1103 return len;
1105 i += len;
1106 if (pkt_dev->dst_mac_count != value) {
1107 pkt_dev->dst_mac_count = value;
1108 pkt_dev->cur_dst_mac_offset = 0;
1110 sprintf(pg_result, "OK: dst_mac_count=%d",
1111 pkt_dev->dst_mac_count);
1112 return count;
1114 if (!strcmp(name, "node")) {
1115 len = num_arg(&user_buffer[i], 10, &value);
1116 if (len < 0)
1117 return len;
1119 i += len;
1121 if (node_possible(value)) {
1122 pkt_dev->node = value;
1123 sprintf(pg_result, "OK: node=%d", pkt_dev->node);
1124 if (pkt_dev->page) {
1125 put_page(pkt_dev->page);
1126 pkt_dev->page = NULL;
1129 else
1130 sprintf(pg_result, "ERROR: node not possible");
1131 return count;
1133 if (!strcmp(name, "flag")) {
1134 char f[32];
1135 memset(f, 0, 32);
1136 len = strn_len(&user_buffer[i], sizeof(f) - 1);
1137 if (len < 0)
1138 return len;
1140 if (copy_from_user(f, &user_buffer[i], len))
1141 return -EFAULT;
1142 i += len;
1143 if (strcmp(f, "IPSRC_RND") == 0)
1144 pkt_dev->flags |= F_IPSRC_RND;
1146 else if (strcmp(f, "!IPSRC_RND") == 0)
1147 pkt_dev->flags &= ~F_IPSRC_RND;
1149 else if (strcmp(f, "TXSIZE_RND") == 0)
1150 pkt_dev->flags |= F_TXSIZE_RND;
1152 else if (strcmp(f, "!TXSIZE_RND") == 0)
1153 pkt_dev->flags &= ~F_TXSIZE_RND;
1155 else if (strcmp(f, "IPDST_RND") == 0)
1156 pkt_dev->flags |= F_IPDST_RND;
1158 else if (strcmp(f, "!IPDST_RND") == 0)
1159 pkt_dev->flags &= ~F_IPDST_RND;
1161 else if (strcmp(f, "UDPSRC_RND") == 0)
1162 pkt_dev->flags |= F_UDPSRC_RND;
1164 else if (strcmp(f, "!UDPSRC_RND") == 0)
1165 pkt_dev->flags &= ~F_UDPSRC_RND;
1167 else if (strcmp(f, "UDPDST_RND") == 0)
1168 pkt_dev->flags |= F_UDPDST_RND;
1170 else if (strcmp(f, "!UDPDST_RND") == 0)
1171 pkt_dev->flags &= ~F_UDPDST_RND;
1173 else if (strcmp(f, "MACSRC_RND") == 0)
1174 pkt_dev->flags |= F_MACSRC_RND;
1176 else if (strcmp(f, "!MACSRC_RND") == 0)
1177 pkt_dev->flags &= ~F_MACSRC_RND;
1179 else if (strcmp(f, "MACDST_RND") == 0)
1180 pkt_dev->flags |= F_MACDST_RND;
1182 else if (strcmp(f, "!MACDST_RND") == 0)
1183 pkt_dev->flags &= ~F_MACDST_RND;
1185 else if (strcmp(f, "MPLS_RND") == 0)
1186 pkt_dev->flags |= F_MPLS_RND;
1188 else if (strcmp(f, "!MPLS_RND") == 0)
1189 pkt_dev->flags &= ~F_MPLS_RND;
1191 else if (strcmp(f, "VID_RND") == 0)
1192 pkt_dev->flags |= F_VID_RND;
1194 else if (strcmp(f, "!VID_RND") == 0)
1195 pkt_dev->flags &= ~F_VID_RND;
1197 else if (strcmp(f, "SVID_RND") == 0)
1198 pkt_dev->flags |= F_SVID_RND;
1200 else if (strcmp(f, "!SVID_RND") == 0)
1201 pkt_dev->flags &= ~F_SVID_RND;
1203 else if (strcmp(f, "FLOW_SEQ") == 0)
1204 pkt_dev->flags |= F_FLOW_SEQ;
1206 else if (strcmp(f, "QUEUE_MAP_RND") == 0)
1207 pkt_dev->flags |= F_QUEUE_MAP_RND;
1209 else if (strcmp(f, "!QUEUE_MAP_RND") == 0)
1210 pkt_dev->flags &= ~F_QUEUE_MAP_RND;
1212 else if (strcmp(f, "QUEUE_MAP_CPU") == 0)
1213 pkt_dev->flags |= F_QUEUE_MAP_CPU;
1215 else if (strcmp(f, "!QUEUE_MAP_CPU") == 0)
1216 pkt_dev->flags &= ~F_QUEUE_MAP_CPU;
1217 #ifdef CONFIG_XFRM
1218 else if (strcmp(f, "IPSEC") == 0)
1219 pkt_dev->flags |= F_IPSEC_ON;
1220 #endif
1222 else if (strcmp(f, "!IPV6") == 0)
1223 pkt_dev->flags &= ~F_IPV6;
1225 else if (strcmp(f, "NODE_ALLOC") == 0)
1226 pkt_dev->flags |= F_NODE;
1228 else if (strcmp(f, "!NODE_ALLOC") == 0)
1229 pkt_dev->flags &= ~F_NODE;
1231 else {
1232 sprintf(pg_result,
1233 "Flag -:%s:- unknown\nAvailable flags, (prepend ! to un-set flag):\n%s",
1235 "IPSRC_RND, IPDST_RND, UDPSRC_RND, UDPDST_RND, "
1236 "MACSRC_RND, MACDST_RND, TXSIZE_RND, IPV6, MPLS_RND, VID_RND, SVID_RND, FLOW_SEQ, IPSEC, NODE_ALLOC\n");
1237 return count;
1239 sprintf(pg_result, "OK: flags=0x%x", pkt_dev->flags);
1240 return count;
1242 if (!strcmp(name, "dst_min") || !strcmp(name, "dst")) {
1243 len = strn_len(&user_buffer[i], sizeof(pkt_dev->dst_min) - 1);
1244 if (len < 0)
1245 return len;
1247 if (copy_from_user(buf, &user_buffer[i], len))
1248 return -EFAULT;
1249 buf[len] = 0;
1250 if (strcmp(buf, pkt_dev->dst_min) != 0) {
1251 memset(pkt_dev->dst_min, 0, sizeof(pkt_dev->dst_min));
1252 strncpy(pkt_dev->dst_min, buf, len);
1253 pkt_dev->daddr_min = in_aton(pkt_dev->dst_min);
1254 pkt_dev->cur_daddr = pkt_dev->daddr_min;
1256 if (debug)
1257 pr_debug("dst_min set to: %s\n", pkt_dev->dst_min);
1258 i += len;
1259 sprintf(pg_result, "OK: dst_min=%s", pkt_dev->dst_min);
1260 return count;
1262 if (!strcmp(name, "dst_max")) {
1263 len = strn_len(&user_buffer[i], sizeof(pkt_dev->dst_max) - 1);
1264 if (len < 0)
1265 return len;
1268 if (copy_from_user(buf, &user_buffer[i], len))
1269 return -EFAULT;
1271 buf[len] = 0;
1272 if (strcmp(buf, pkt_dev->dst_max) != 0) {
1273 memset(pkt_dev->dst_max, 0, sizeof(pkt_dev->dst_max));
1274 strncpy(pkt_dev->dst_max, buf, len);
1275 pkt_dev->daddr_max = in_aton(pkt_dev->dst_max);
1276 pkt_dev->cur_daddr = pkt_dev->daddr_max;
1278 if (debug)
1279 pr_debug("dst_max set to: %s\n", pkt_dev->dst_max);
1280 i += len;
1281 sprintf(pg_result, "OK: dst_max=%s", pkt_dev->dst_max);
1282 return count;
1284 if (!strcmp(name, "dst6")) {
1285 len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1286 if (len < 0)
1287 return len;
1289 pkt_dev->flags |= F_IPV6;
1291 if (copy_from_user(buf, &user_buffer[i], len))
1292 return -EFAULT;
1293 buf[len] = 0;
1295 in6_pton(buf, -1, pkt_dev->in6_daddr.s6_addr, -1, NULL);
1296 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->in6_daddr);
1298 pkt_dev->cur_in6_daddr = pkt_dev->in6_daddr;
1300 if (debug)
1301 pr_debug("dst6 set to: %s\n", buf);
1303 i += len;
1304 sprintf(pg_result, "OK: dst6=%s", buf);
1305 return count;
1307 if (!strcmp(name, "dst6_min")) {
1308 len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1309 if (len < 0)
1310 return len;
1312 pkt_dev->flags |= F_IPV6;
1314 if (copy_from_user(buf, &user_buffer[i], len))
1315 return -EFAULT;
1316 buf[len] = 0;
1318 in6_pton(buf, -1, pkt_dev->min_in6_daddr.s6_addr, -1, NULL);
1319 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->min_in6_daddr);
1321 pkt_dev->cur_in6_daddr = pkt_dev->min_in6_daddr;
1322 if (debug)
1323 pr_debug("dst6_min set to: %s\n", buf);
1325 i += len;
1326 sprintf(pg_result, "OK: dst6_min=%s", buf);
1327 return count;
1329 if (!strcmp(name, "dst6_max")) {
1330 len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1331 if (len < 0)
1332 return len;
1334 pkt_dev->flags |= F_IPV6;
1336 if (copy_from_user(buf, &user_buffer[i], len))
1337 return -EFAULT;
1338 buf[len] = 0;
1340 in6_pton(buf, -1, pkt_dev->max_in6_daddr.s6_addr, -1, NULL);
1341 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->max_in6_daddr);
1343 if (debug)
1344 pr_debug("dst6_max set to: %s\n", buf);
1346 i += len;
1347 sprintf(pg_result, "OK: dst6_max=%s", buf);
1348 return count;
1350 if (!strcmp(name, "src6")) {
1351 len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1352 if (len < 0)
1353 return len;
1355 pkt_dev->flags |= F_IPV6;
1357 if (copy_from_user(buf, &user_buffer[i], len))
1358 return -EFAULT;
1359 buf[len] = 0;
1361 in6_pton(buf, -1, pkt_dev->in6_saddr.s6_addr, -1, NULL);
1362 snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->in6_saddr);
1364 pkt_dev->cur_in6_saddr = pkt_dev->in6_saddr;
1366 if (debug)
1367 pr_debug("src6 set to: %s\n", buf);
1369 i += len;
1370 sprintf(pg_result, "OK: src6=%s", buf);
1371 return count;
1373 if (!strcmp(name, "src_min")) {
1374 len = strn_len(&user_buffer[i], sizeof(pkt_dev->src_min) - 1);
1375 if (len < 0)
1376 return len;
1378 if (copy_from_user(buf, &user_buffer[i], len))
1379 return -EFAULT;
1380 buf[len] = 0;
1381 if (strcmp(buf, pkt_dev->src_min) != 0) {
1382 memset(pkt_dev->src_min, 0, sizeof(pkt_dev->src_min));
1383 strncpy(pkt_dev->src_min, buf, len);
1384 pkt_dev->saddr_min = in_aton(pkt_dev->src_min);
1385 pkt_dev->cur_saddr = pkt_dev->saddr_min;
1387 if (debug)
1388 pr_debug("src_min set to: %s\n", pkt_dev->src_min);
1389 i += len;
1390 sprintf(pg_result, "OK: src_min=%s", pkt_dev->src_min);
1391 return count;
1393 if (!strcmp(name, "src_max")) {
1394 len = strn_len(&user_buffer[i], sizeof(pkt_dev->src_max) - 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_max) != 0) {
1402 memset(pkt_dev->src_max, 0, sizeof(pkt_dev->src_max));
1403 strncpy(pkt_dev->src_max, buf, len);
1404 pkt_dev->saddr_max = in_aton(pkt_dev->src_max);
1405 pkt_dev->cur_saddr = pkt_dev->saddr_max;
1407 if (debug)
1408 pr_debug("src_max set to: %s\n", pkt_dev->src_max);
1409 i += len;
1410 sprintf(pg_result, "OK: src_max=%s", pkt_dev->src_max);
1411 return count;
1413 if (!strcmp(name, "dst_mac")) {
1414 len = strn_len(&user_buffer[i], sizeof(valstr) - 1);
1415 if (len < 0)
1416 return len;
1418 memset(valstr, 0, sizeof(valstr));
1419 if (copy_from_user(valstr, &user_buffer[i], len))
1420 return -EFAULT;
1422 if (!mac_pton(valstr, pkt_dev->dst_mac))
1423 return -EINVAL;
1424 /* Set up Dest MAC */
1425 memcpy(&pkt_dev->hh[0], pkt_dev->dst_mac, ETH_ALEN);
1427 sprintf(pg_result, "OK: dstmac %pM", pkt_dev->dst_mac);
1428 return count;
1430 if (!strcmp(name, "src_mac")) {
1431 len = strn_len(&user_buffer[i], sizeof(valstr) - 1);
1432 if (len < 0)
1433 return len;
1435 memset(valstr, 0, sizeof(valstr));
1436 if (copy_from_user(valstr, &user_buffer[i], len))
1437 return -EFAULT;
1439 if (!mac_pton(valstr, pkt_dev->src_mac))
1440 return -EINVAL;
1441 /* Set up Src MAC */
1442 memcpy(&pkt_dev->hh[6], pkt_dev->src_mac, ETH_ALEN);
1444 sprintf(pg_result, "OK: srcmac %pM", pkt_dev->src_mac);
1445 return count;
1448 if (!strcmp(name, "clear_counters")) {
1449 pktgen_clear_counters(pkt_dev);
1450 sprintf(pg_result, "OK: Clearing counters.\n");
1451 return count;
1454 if (!strcmp(name, "flows")) {
1455 len = num_arg(&user_buffer[i], 10, &value);
1456 if (len < 0)
1457 return len;
1459 i += len;
1460 if (value > MAX_CFLOWS)
1461 value = MAX_CFLOWS;
1463 pkt_dev->cflows = value;
1464 sprintf(pg_result, "OK: flows=%u", pkt_dev->cflows);
1465 return count;
1468 if (!strcmp(name, "flowlen")) {
1469 len = num_arg(&user_buffer[i], 10, &value);
1470 if (len < 0)
1471 return len;
1473 i += len;
1474 pkt_dev->lflow = value;
1475 sprintf(pg_result, "OK: flowlen=%u", pkt_dev->lflow);
1476 return count;
1479 if (!strcmp(name, "queue_map_min")) {
1480 len = num_arg(&user_buffer[i], 5, &value);
1481 if (len < 0)
1482 return len;
1484 i += len;
1485 pkt_dev->queue_map_min = value;
1486 sprintf(pg_result, "OK: queue_map_min=%u", pkt_dev->queue_map_min);
1487 return count;
1490 if (!strcmp(name, "queue_map_max")) {
1491 len = num_arg(&user_buffer[i], 5, &value);
1492 if (len < 0)
1493 return len;
1495 i += len;
1496 pkt_dev->queue_map_max = value;
1497 sprintf(pg_result, "OK: queue_map_max=%u", pkt_dev->queue_map_max);
1498 return count;
1501 if (!strcmp(name, "mpls")) {
1502 unsigned int n, cnt;
1504 len = get_labels(&user_buffer[i], pkt_dev);
1505 if (len < 0)
1506 return len;
1507 i += len;
1508 cnt = sprintf(pg_result, "OK: mpls=");
1509 for (n = 0; n < pkt_dev->nr_labels; n++)
1510 cnt += sprintf(pg_result + cnt,
1511 "%08x%s", ntohl(pkt_dev->labels[n]),
1512 n == pkt_dev->nr_labels-1 ? "" : ",");
1514 if (pkt_dev->nr_labels && pkt_dev->vlan_id != 0xffff) {
1515 pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1516 pkt_dev->svlan_id = 0xffff;
1518 if (debug)
1519 pr_debug("VLAN/SVLAN auto turned off\n");
1521 return count;
1524 if (!strcmp(name, "vlan_id")) {
1525 len = num_arg(&user_buffer[i], 4, &value);
1526 if (len < 0)
1527 return len;
1529 i += len;
1530 if (value <= 4095) {
1531 pkt_dev->vlan_id = value; /* turn on VLAN */
1533 if (debug)
1534 pr_debug("VLAN turned on\n");
1536 if (debug && pkt_dev->nr_labels)
1537 pr_debug("MPLS auto turned off\n");
1539 pkt_dev->nr_labels = 0; /* turn off MPLS */
1540 sprintf(pg_result, "OK: vlan_id=%u", pkt_dev->vlan_id);
1541 } else {
1542 pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1543 pkt_dev->svlan_id = 0xffff;
1545 if (debug)
1546 pr_debug("VLAN/SVLAN turned off\n");
1548 return count;
1551 if (!strcmp(name, "vlan_p")) {
1552 len = num_arg(&user_buffer[i], 1, &value);
1553 if (len < 0)
1554 return len;
1556 i += len;
1557 if ((value <= 7) && (pkt_dev->vlan_id != 0xffff)) {
1558 pkt_dev->vlan_p = value;
1559 sprintf(pg_result, "OK: vlan_p=%u", pkt_dev->vlan_p);
1560 } else {
1561 sprintf(pg_result, "ERROR: vlan_p must be 0-7");
1563 return count;
1566 if (!strcmp(name, "vlan_cfi")) {
1567 len = num_arg(&user_buffer[i], 1, &value);
1568 if (len < 0)
1569 return len;
1571 i += len;
1572 if ((value <= 1) && (pkt_dev->vlan_id != 0xffff)) {
1573 pkt_dev->vlan_cfi = value;
1574 sprintf(pg_result, "OK: vlan_cfi=%u", pkt_dev->vlan_cfi);
1575 } else {
1576 sprintf(pg_result, "ERROR: vlan_cfi must be 0-1");
1578 return count;
1581 if (!strcmp(name, "svlan_id")) {
1582 len = num_arg(&user_buffer[i], 4, &value);
1583 if (len < 0)
1584 return len;
1586 i += len;
1587 if ((value <= 4095) && ((pkt_dev->vlan_id != 0xffff))) {
1588 pkt_dev->svlan_id = value; /* turn on SVLAN */
1590 if (debug)
1591 pr_debug("SVLAN turned on\n");
1593 if (debug && pkt_dev->nr_labels)
1594 pr_debug("MPLS auto turned off\n");
1596 pkt_dev->nr_labels = 0; /* turn off MPLS */
1597 sprintf(pg_result, "OK: svlan_id=%u", pkt_dev->svlan_id);
1598 } else {
1599 pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1600 pkt_dev->svlan_id = 0xffff;
1602 if (debug)
1603 pr_debug("VLAN/SVLAN turned off\n");
1605 return count;
1608 if (!strcmp(name, "svlan_p")) {
1609 len = num_arg(&user_buffer[i], 1, &value);
1610 if (len < 0)
1611 return len;
1613 i += len;
1614 if ((value <= 7) && (pkt_dev->svlan_id != 0xffff)) {
1615 pkt_dev->svlan_p = value;
1616 sprintf(pg_result, "OK: svlan_p=%u", pkt_dev->svlan_p);
1617 } else {
1618 sprintf(pg_result, "ERROR: svlan_p must be 0-7");
1620 return count;
1623 if (!strcmp(name, "svlan_cfi")) {
1624 len = num_arg(&user_buffer[i], 1, &value);
1625 if (len < 0)
1626 return len;
1628 i += len;
1629 if ((value <= 1) && (pkt_dev->svlan_id != 0xffff)) {
1630 pkt_dev->svlan_cfi = value;
1631 sprintf(pg_result, "OK: svlan_cfi=%u", pkt_dev->svlan_cfi);
1632 } else {
1633 sprintf(pg_result, "ERROR: svlan_cfi must be 0-1");
1635 return count;
1638 if (!strcmp(name, "tos")) {
1639 __u32 tmp_value = 0;
1640 len = hex32_arg(&user_buffer[i], 2, &tmp_value);
1641 if (len < 0)
1642 return len;
1644 i += len;
1645 if (len == 2) {
1646 pkt_dev->tos = tmp_value;
1647 sprintf(pg_result, "OK: tos=0x%02x", pkt_dev->tos);
1648 } else {
1649 sprintf(pg_result, "ERROR: tos must be 00-ff");
1651 return count;
1654 if (!strcmp(name, "traffic_class")) {
1655 __u32 tmp_value = 0;
1656 len = hex32_arg(&user_buffer[i], 2, &tmp_value);
1657 if (len < 0)
1658 return len;
1660 i += len;
1661 if (len == 2) {
1662 pkt_dev->traffic_class = tmp_value;
1663 sprintf(pg_result, "OK: traffic_class=0x%02x", pkt_dev->traffic_class);
1664 } else {
1665 sprintf(pg_result, "ERROR: traffic_class must be 00-ff");
1667 return count;
1670 if (!strcmp(name, "skb_priority")) {
1671 len = num_arg(&user_buffer[i], 9, &value);
1672 if (len < 0)
1673 return len;
1675 i += len;
1676 pkt_dev->skb_priority = value;
1677 sprintf(pg_result, "OK: skb_priority=%i",
1678 pkt_dev->skb_priority);
1679 return count;
1682 sprintf(pkt_dev->result, "No such parameter \"%s\"", name);
1683 return -EINVAL;
1686 static int pktgen_if_open(struct inode *inode, struct file *file)
1688 return single_open(file, pktgen_if_show, PDE_DATA(inode));
1691 static const struct file_operations pktgen_if_fops = {
1692 .owner = THIS_MODULE,
1693 .open = pktgen_if_open,
1694 .read = seq_read,
1695 .llseek = seq_lseek,
1696 .write = pktgen_if_write,
1697 .release = single_release,
1700 static int pktgen_thread_show(struct seq_file *seq, void *v)
1702 struct pktgen_thread *t = seq->private;
1703 const struct pktgen_dev *pkt_dev;
1705 BUG_ON(!t);
1707 seq_printf(seq, "Running: ");
1709 if_lock(t);
1710 list_for_each_entry(pkt_dev, &t->if_list, list)
1711 if (pkt_dev->running)
1712 seq_printf(seq, "%s ", pkt_dev->odevname);
1714 seq_printf(seq, "\nStopped: ");
1716 list_for_each_entry(pkt_dev, &t->if_list, list)
1717 if (!pkt_dev->running)
1718 seq_printf(seq, "%s ", pkt_dev->odevname);
1720 if (t->result[0])
1721 seq_printf(seq, "\nResult: %s\n", t->result);
1722 else
1723 seq_printf(seq, "\nResult: NA\n");
1725 if_unlock(t);
1727 return 0;
1730 static ssize_t pktgen_thread_write(struct file *file,
1731 const char __user * user_buffer,
1732 size_t count, loff_t * offset)
1734 struct seq_file *seq = file->private_data;
1735 struct pktgen_thread *t = seq->private;
1736 int i, max, len, ret;
1737 char name[40];
1738 char *pg_result;
1740 if (count < 1) {
1741 // sprintf(pg_result, "Wrong command format");
1742 return -EINVAL;
1745 max = count;
1746 len = count_trail_chars(user_buffer, max);
1747 if (len < 0)
1748 return len;
1750 i = len;
1752 /* Read variable name */
1754 len = strn_len(&user_buffer[i], sizeof(name) - 1);
1755 if (len < 0)
1756 return len;
1758 memset(name, 0, sizeof(name));
1759 if (copy_from_user(name, &user_buffer[i], len))
1760 return -EFAULT;
1761 i += len;
1763 max = count - i;
1764 len = count_trail_chars(&user_buffer[i], max);
1765 if (len < 0)
1766 return len;
1768 i += len;
1770 if (debug)
1771 pr_debug("t=%s, count=%lu\n", name, (unsigned long)count);
1773 if (!t) {
1774 pr_err("ERROR: No thread\n");
1775 ret = -EINVAL;
1776 goto out;
1779 pg_result = &(t->result[0]);
1781 if (!strcmp(name, "add_device")) {
1782 char f[32];
1783 memset(f, 0, 32);
1784 len = strn_len(&user_buffer[i], sizeof(f) - 1);
1785 if (len < 0) {
1786 ret = len;
1787 goto out;
1789 if (copy_from_user(f, &user_buffer[i], len))
1790 return -EFAULT;
1791 i += len;
1792 mutex_lock(&pktgen_thread_lock);
1793 ret = pktgen_add_device(t, f);
1794 mutex_unlock(&pktgen_thread_lock);
1795 if (!ret) {
1796 ret = count;
1797 sprintf(pg_result, "OK: add_device=%s", f);
1798 } else
1799 sprintf(pg_result, "ERROR: can not add device %s", f);
1800 goto out;
1803 if (!strcmp(name, "rem_device_all")) {
1804 mutex_lock(&pktgen_thread_lock);
1805 t->control |= T_REMDEVALL;
1806 mutex_unlock(&pktgen_thread_lock);
1807 schedule_timeout_interruptible(msecs_to_jiffies(125)); /* Propagate thread->control */
1808 ret = count;
1809 sprintf(pg_result, "OK: rem_device_all");
1810 goto out;
1813 if (!strcmp(name, "max_before_softirq")) {
1814 sprintf(pg_result, "OK: Note! max_before_softirq is obsoleted -- Do not use");
1815 ret = count;
1816 goto out;
1819 ret = -EINVAL;
1820 out:
1821 return ret;
1824 static int pktgen_thread_open(struct inode *inode, struct file *file)
1826 return single_open(file, pktgen_thread_show, PDE_DATA(inode));
1829 static const struct file_operations pktgen_thread_fops = {
1830 .owner = THIS_MODULE,
1831 .open = pktgen_thread_open,
1832 .read = seq_read,
1833 .llseek = seq_lseek,
1834 .write = pktgen_thread_write,
1835 .release = single_release,
1838 /* Think find or remove for NN */
1839 static struct pktgen_dev *__pktgen_NN_threads(const struct pktgen_net *pn,
1840 const char *ifname, int remove)
1842 struct pktgen_thread *t;
1843 struct pktgen_dev *pkt_dev = NULL;
1844 bool exact = (remove == FIND);
1846 list_for_each_entry(t, &pn->pktgen_threads, th_list) {
1847 pkt_dev = pktgen_find_dev(t, ifname, exact);
1848 if (pkt_dev) {
1849 if (remove) {
1850 if_lock(t);
1851 pkt_dev->removal_mark = 1;
1852 t->control |= T_REMDEV;
1853 if_unlock(t);
1855 break;
1858 return pkt_dev;
1862 * mark a device for removal
1864 static void pktgen_mark_device(const struct pktgen_net *pn, const char *ifname)
1866 struct pktgen_dev *pkt_dev = NULL;
1867 const int max_tries = 10, msec_per_try = 125;
1868 int i = 0;
1870 mutex_lock(&pktgen_thread_lock);
1871 pr_debug("%s: marking %s for removal\n", __func__, ifname);
1873 while (1) {
1875 pkt_dev = __pktgen_NN_threads(pn, ifname, REMOVE);
1876 if (pkt_dev == NULL)
1877 break; /* success */
1879 mutex_unlock(&pktgen_thread_lock);
1880 pr_debug("%s: waiting for %s to disappear....\n",
1881 __func__, ifname);
1882 schedule_timeout_interruptible(msecs_to_jiffies(msec_per_try));
1883 mutex_lock(&pktgen_thread_lock);
1885 if (++i >= max_tries) {
1886 pr_err("%s: timed out after waiting %d msec for device %s to be removed\n",
1887 __func__, msec_per_try * i, ifname);
1888 break;
1893 mutex_unlock(&pktgen_thread_lock);
1896 static void pktgen_change_name(const struct pktgen_net *pn, struct net_device *dev)
1898 struct pktgen_thread *t;
1900 list_for_each_entry(t, &pn->pktgen_threads, th_list) {
1901 struct pktgen_dev *pkt_dev;
1903 list_for_each_entry(pkt_dev, &t->if_list, list) {
1904 if (pkt_dev->odev != dev)
1905 continue;
1907 proc_remove(pkt_dev->entry);
1909 pkt_dev->entry = proc_create_data(dev->name, 0600,
1910 pn->proc_dir,
1911 &pktgen_if_fops,
1912 pkt_dev);
1913 if (!pkt_dev->entry)
1914 pr_err("can't move proc entry for '%s'\n",
1915 dev->name);
1916 break;
1921 static int pktgen_device_event(struct notifier_block *unused,
1922 unsigned long event, void *ptr)
1924 struct net_device *dev = ptr;
1925 struct pktgen_net *pn = net_generic(dev_net(dev), pg_net_id);
1927 if (pn->pktgen_exiting)
1928 return NOTIFY_DONE;
1930 /* It is OK that we do not hold the group lock right now,
1931 * as we run under the RTNL lock.
1934 switch (event) {
1935 case NETDEV_CHANGENAME:
1936 pktgen_change_name(pn, dev);
1937 break;
1939 case NETDEV_UNREGISTER:
1940 pktgen_mark_device(pn, dev->name);
1941 break;
1944 return NOTIFY_DONE;
1947 static struct net_device *pktgen_dev_get_by_name(const struct pktgen_net *pn,
1948 struct pktgen_dev *pkt_dev,
1949 const char *ifname)
1951 char b[IFNAMSIZ+5];
1952 int i;
1954 for (i = 0; ifname[i] != '@'; i++) {
1955 if (i == IFNAMSIZ)
1956 break;
1958 b[i] = ifname[i];
1960 b[i] = 0;
1962 return dev_get_by_name(pn->net, b);
1966 /* Associate pktgen_dev with a device. */
1968 static int pktgen_setup_dev(const struct pktgen_net *pn,
1969 struct pktgen_dev *pkt_dev, const char *ifname)
1971 struct net_device *odev;
1972 int err;
1974 /* Clean old setups */
1975 if (pkt_dev->odev) {
1976 dev_put(pkt_dev->odev);
1977 pkt_dev->odev = NULL;
1980 odev = pktgen_dev_get_by_name(pn, pkt_dev, ifname);
1981 if (!odev) {
1982 pr_err("no such netdevice: \"%s\"\n", ifname);
1983 return -ENODEV;
1986 if (odev->type != ARPHRD_ETHER) {
1987 pr_err("not an ethernet device: \"%s\"\n", ifname);
1988 err = -EINVAL;
1989 } else if (!netif_running(odev)) {
1990 pr_err("device is down: \"%s\"\n", ifname);
1991 err = -ENETDOWN;
1992 } else {
1993 pkt_dev->odev = odev;
1994 return 0;
1997 dev_put(odev);
1998 return err;
2001 /* Read pkt_dev from the interface and set up internal pktgen_dev
2002 * structure to have the right information to create/send packets
2004 static void pktgen_setup_inject(struct pktgen_dev *pkt_dev)
2006 int ntxq;
2008 if (!pkt_dev->odev) {
2009 pr_err("ERROR: pkt_dev->odev == NULL in setup_inject\n");
2010 sprintf(pkt_dev->result,
2011 "ERROR: pkt_dev->odev == NULL in setup_inject.\n");
2012 return;
2015 /* make sure that we don't pick a non-existing transmit queue */
2016 ntxq = pkt_dev->odev->real_num_tx_queues;
2018 if (ntxq <= pkt_dev->queue_map_min) {
2019 pr_warning("WARNING: Requested queue_map_min (zero-based) (%d) exceeds valid range [0 - %d] for (%d) queues on %s, resetting\n",
2020 pkt_dev->queue_map_min, (ntxq ?: 1) - 1, ntxq,
2021 pkt_dev->odevname);
2022 pkt_dev->queue_map_min = (ntxq ?: 1) - 1;
2024 if (pkt_dev->queue_map_max >= ntxq) {
2025 pr_warning("WARNING: Requested queue_map_max (zero-based) (%d) exceeds valid range [0 - %d] for (%d) queues on %s, resetting\n",
2026 pkt_dev->queue_map_max, (ntxq ?: 1) - 1, ntxq,
2027 pkt_dev->odevname);
2028 pkt_dev->queue_map_max = (ntxq ?: 1) - 1;
2031 /* Default to the interface's mac if not explicitly set. */
2033 if (is_zero_ether_addr(pkt_dev->src_mac))
2034 memcpy(&(pkt_dev->hh[6]), pkt_dev->odev->dev_addr, ETH_ALEN);
2036 /* Set up Dest MAC */
2037 memcpy(&(pkt_dev->hh[0]), pkt_dev->dst_mac, ETH_ALEN);
2039 if (pkt_dev->flags & F_IPV6) {
2040 int i, set = 0, err = 1;
2041 struct inet6_dev *idev;
2043 if (pkt_dev->min_pkt_size == 0) {
2044 pkt_dev->min_pkt_size = 14 + sizeof(struct ipv6hdr)
2045 + sizeof(struct udphdr)
2046 + sizeof(struct pktgen_hdr)
2047 + pkt_dev->pkt_overhead;
2050 for (i = 0; i < IN6_ADDR_HSIZE; i++)
2051 if (pkt_dev->cur_in6_saddr.s6_addr[i]) {
2052 set = 1;
2053 break;
2056 if (!set) {
2059 * Use linklevel address if unconfigured.
2061 * use ipv6_get_lladdr if/when it's get exported
2064 rcu_read_lock();
2065 idev = __in6_dev_get(pkt_dev->odev);
2066 if (idev) {
2067 struct inet6_ifaddr *ifp;
2069 read_lock_bh(&idev->lock);
2070 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2071 if ((ifp->scope & IFA_LINK) &&
2072 !(ifp->flags & IFA_F_TENTATIVE)) {
2073 pkt_dev->cur_in6_saddr = ifp->addr;
2074 err = 0;
2075 break;
2078 read_unlock_bh(&idev->lock);
2080 rcu_read_unlock();
2081 if (err)
2082 pr_err("ERROR: IPv6 link address not available\n");
2084 } else {
2085 if (pkt_dev->min_pkt_size == 0) {
2086 pkt_dev->min_pkt_size = 14 + sizeof(struct iphdr)
2087 + sizeof(struct udphdr)
2088 + sizeof(struct pktgen_hdr)
2089 + pkt_dev->pkt_overhead;
2092 pkt_dev->saddr_min = 0;
2093 pkt_dev->saddr_max = 0;
2094 if (strlen(pkt_dev->src_min) == 0) {
2096 struct in_device *in_dev;
2098 rcu_read_lock();
2099 in_dev = __in_dev_get_rcu(pkt_dev->odev);
2100 if (in_dev) {
2101 if (in_dev->ifa_list) {
2102 pkt_dev->saddr_min =
2103 in_dev->ifa_list->ifa_address;
2104 pkt_dev->saddr_max = pkt_dev->saddr_min;
2107 rcu_read_unlock();
2108 } else {
2109 pkt_dev->saddr_min = in_aton(pkt_dev->src_min);
2110 pkt_dev->saddr_max = in_aton(pkt_dev->src_max);
2113 pkt_dev->daddr_min = in_aton(pkt_dev->dst_min);
2114 pkt_dev->daddr_max = in_aton(pkt_dev->dst_max);
2116 /* Initialize current values. */
2117 pkt_dev->cur_pkt_size = pkt_dev->min_pkt_size;
2118 if (pkt_dev->min_pkt_size > pkt_dev->max_pkt_size)
2119 pkt_dev->max_pkt_size = pkt_dev->min_pkt_size;
2121 pkt_dev->cur_dst_mac_offset = 0;
2122 pkt_dev->cur_src_mac_offset = 0;
2123 pkt_dev->cur_saddr = pkt_dev->saddr_min;
2124 pkt_dev->cur_daddr = pkt_dev->daddr_min;
2125 pkt_dev->cur_udp_dst = pkt_dev->udp_dst_min;
2126 pkt_dev->cur_udp_src = pkt_dev->udp_src_min;
2127 pkt_dev->nflows = 0;
2131 static void spin(struct pktgen_dev *pkt_dev, ktime_t spin_until)
2133 ktime_t start_time, end_time;
2134 s64 remaining;
2135 struct hrtimer_sleeper t;
2137 hrtimer_init_on_stack(&t.timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS);
2138 hrtimer_set_expires(&t.timer, spin_until);
2140 remaining = ktime_to_ns(hrtimer_expires_remaining(&t.timer));
2141 if (remaining <= 0) {
2142 pkt_dev->next_tx = ktime_add_ns(spin_until, pkt_dev->delay);
2143 return;
2146 start_time = ktime_get();
2147 if (remaining < 100000) {
2148 /* for small delays (<100us), just loop until limit is reached */
2149 do {
2150 end_time = ktime_get();
2151 } while (ktime_compare(end_time, spin_until) < 0);
2152 } else {
2153 /* see do_nanosleep */
2154 hrtimer_init_sleeper(&t, current);
2155 do {
2156 set_current_state(TASK_INTERRUPTIBLE);
2157 hrtimer_start_expires(&t.timer, HRTIMER_MODE_ABS);
2158 if (!hrtimer_active(&t.timer))
2159 t.task = NULL;
2161 if (likely(t.task))
2162 schedule();
2164 hrtimer_cancel(&t.timer);
2165 } while (t.task && pkt_dev->running && !signal_pending(current));
2166 __set_current_state(TASK_RUNNING);
2167 end_time = ktime_get();
2170 pkt_dev->idle_acc += ktime_to_ns(ktime_sub(end_time, start_time));
2171 pkt_dev->next_tx = ktime_add_ns(spin_until, pkt_dev->delay);
2174 static inline void set_pkt_overhead(struct pktgen_dev *pkt_dev)
2176 pkt_dev->pkt_overhead = 0;
2177 pkt_dev->pkt_overhead += pkt_dev->nr_labels*sizeof(u32);
2178 pkt_dev->pkt_overhead += VLAN_TAG_SIZE(pkt_dev);
2179 pkt_dev->pkt_overhead += SVLAN_TAG_SIZE(pkt_dev);
2182 static inline int f_seen(const struct pktgen_dev *pkt_dev, int flow)
2184 return !!(pkt_dev->flows[flow].flags & F_INIT);
2187 static inline int f_pick(struct pktgen_dev *pkt_dev)
2189 int flow = pkt_dev->curfl;
2191 if (pkt_dev->flags & F_FLOW_SEQ) {
2192 if (pkt_dev->flows[flow].count >= pkt_dev->lflow) {
2193 /* reset time */
2194 pkt_dev->flows[flow].count = 0;
2195 pkt_dev->flows[flow].flags = 0;
2196 pkt_dev->curfl += 1;
2197 if (pkt_dev->curfl >= pkt_dev->cflows)
2198 pkt_dev->curfl = 0; /*reset */
2200 } else {
2201 flow = prandom_u32() % pkt_dev->cflows;
2202 pkt_dev->curfl = flow;
2204 if (pkt_dev->flows[flow].count > pkt_dev->lflow) {
2205 pkt_dev->flows[flow].count = 0;
2206 pkt_dev->flows[flow].flags = 0;
2210 return pkt_dev->curfl;
2214 #ifdef CONFIG_XFRM
2215 /* If there was already an IPSEC SA, we keep it as is, else
2216 * we go look for it ...
2218 #define DUMMY_MARK 0
2219 static void get_ipsec_sa(struct pktgen_dev *pkt_dev, int flow)
2221 struct xfrm_state *x = pkt_dev->flows[flow].x;
2222 struct pktgen_net *pn = net_generic(dev_net(pkt_dev->odev), pg_net_id);
2223 if (!x) {
2224 /*slow path: we dont already have xfrm_state*/
2225 x = xfrm_stateonly_find(pn->net, DUMMY_MARK,
2226 (xfrm_address_t *)&pkt_dev->cur_daddr,
2227 (xfrm_address_t *)&pkt_dev->cur_saddr,
2228 AF_INET,
2229 pkt_dev->ipsmode,
2230 pkt_dev->ipsproto, 0);
2231 if (x) {
2232 pkt_dev->flows[flow].x = x;
2233 set_pkt_overhead(pkt_dev);
2234 pkt_dev->pkt_overhead += x->props.header_len;
2239 #endif
2240 static void set_cur_queue_map(struct pktgen_dev *pkt_dev)
2243 if (pkt_dev->flags & F_QUEUE_MAP_CPU)
2244 pkt_dev->cur_queue_map = smp_processor_id();
2246 else if (pkt_dev->queue_map_min <= pkt_dev->queue_map_max) {
2247 __u16 t;
2248 if (pkt_dev->flags & F_QUEUE_MAP_RND) {
2249 t = prandom_u32() %
2250 (pkt_dev->queue_map_max -
2251 pkt_dev->queue_map_min + 1)
2252 + pkt_dev->queue_map_min;
2253 } else {
2254 t = pkt_dev->cur_queue_map + 1;
2255 if (t > pkt_dev->queue_map_max)
2256 t = pkt_dev->queue_map_min;
2258 pkt_dev->cur_queue_map = t;
2260 pkt_dev->cur_queue_map = pkt_dev->cur_queue_map % pkt_dev->odev->real_num_tx_queues;
2263 /* Increment/randomize headers according to flags and current values
2264 * for IP src/dest, UDP src/dst port, MAC-Addr src/dst
2266 static void mod_cur_headers(struct pktgen_dev *pkt_dev)
2268 __u32 imn;
2269 __u32 imx;
2270 int flow = 0;
2272 if (pkt_dev->cflows)
2273 flow = f_pick(pkt_dev);
2275 /* Deal with source MAC */
2276 if (pkt_dev->src_mac_count > 1) {
2277 __u32 mc;
2278 __u32 tmp;
2280 if (pkt_dev->flags & F_MACSRC_RND)
2281 mc = prandom_u32() % pkt_dev->src_mac_count;
2282 else {
2283 mc = pkt_dev->cur_src_mac_offset++;
2284 if (pkt_dev->cur_src_mac_offset >=
2285 pkt_dev->src_mac_count)
2286 pkt_dev->cur_src_mac_offset = 0;
2289 tmp = pkt_dev->src_mac[5] + (mc & 0xFF);
2290 pkt_dev->hh[11] = tmp;
2291 tmp = (pkt_dev->src_mac[4] + ((mc >> 8) & 0xFF) + (tmp >> 8));
2292 pkt_dev->hh[10] = tmp;
2293 tmp = (pkt_dev->src_mac[3] + ((mc >> 16) & 0xFF) + (tmp >> 8));
2294 pkt_dev->hh[9] = tmp;
2295 tmp = (pkt_dev->src_mac[2] + ((mc >> 24) & 0xFF) + (tmp >> 8));
2296 pkt_dev->hh[8] = tmp;
2297 tmp = (pkt_dev->src_mac[1] + (tmp >> 8));
2298 pkt_dev->hh[7] = tmp;
2301 /* Deal with Destination MAC */
2302 if (pkt_dev->dst_mac_count > 1) {
2303 __u32 mc;
2304 __u32 tmp;
2306 if (pkt_dev->flags & F_MACDST_RND)
2307 mc = prandom_u32() % pkt_dev->dst_mac_count;
2309 else {
2310 mc = pkt_dev->cur_dst_mac_offset++;
2311 if (pkt_dev->cur_dst_mac_offset >=
2312 pkt_dev->dst_mac_count) {
2313 pkt_dev->cur_dst_mac_offset = 0;
2317 tmp = pkt_dev->dst_mac[5] + (mc & 0xFF);
2318 pkt_dev->hh[5] = tmp;
2319 tmp = (pkt_dev->dst_mac[4] + ((mc >> 8) & 0xFF) + (tmp >> 8));
2320 pkt_dev->hh[4] = tmp;
2321 tmp = (pkt_dev->dst_mac[3] + ((mc >> 16) & 0xFF) + (tmp >> 8));
2322 pkt_dev->hh[3] = tmp;
2323 tmp = (pkt_dev->dst_mac[2] + ((mc >> 24) & 0xFF) + (tmp >> 8));
2324 pkt_dev->hh[2] = tmp;
2325 tmp = (pkt_dev->dst_mac[1] + (tmp >> 8));
2326 pkt_dev->hh[1] = tmp;
2329 if (pkt_dev->flags & F_MPLS_RND) {
2330 unsigned int i;
2331 for (i = 0; i < pkt_dev->nr_labels; i++)
2332 if (pkt_dev->labels[i] & MPLS_STACK_BOTTOM)
2333 pkt_dev->labels[i] = MPLS_STACK_BOTTOM |
2334 ((__force __be32)prandom_u32() &
2335 htonl(0x000fffff));
2338 if ((pkt_dev->flags & F_VID_RND) && (pkt_dev->vlan_id != 0xffff)) {
2339 pkt_dev->vlan_id = prandom_u32() & (4096 - 1);
2342 if ((pkt_dev->flags & F_SVID_RND) && (pkt_dev->svlan_id != 0xffff)) {
2343 pkt_dev->svlan_id = prandom_u32() & (4096 - 1);
2346 if (pkt_dev->udp_src_min < pkt_dev->udp_src_max) {
2347 if (pkt_dev->flags & F_UDPSRC_RND)
2348 pkt_dev->cur_udp_src = prandom_u32() %
2349 (pkt_dev->udp_src_max - pkt_dev->udp_src_min)
2350 + pkt_dev->udp_src_min;
2352 else {
2353 pkt_dev->cur_udp_src++;
2354 if (pkt_dev->cur_udp_src >= pkt_dev->udp_src_max)
2355 pkt_dev->cur_udp_src = pkt_dev->udp_src_min;
2359 if (pkt_dev->udp_dst_min < pkt_dev->udp_dst_max) {
2360 if (pkt_dev->flags & F_UDPDST_RND) {
2361 pkt_dev->cur_udp_dst = prandom_u32() %
2362 (pkt_dev->udp_dst_max - pkt_dev->udp_dst_min)
2363 + pkt_dev->udp_dst_min;
2364 } else {
2365 pkt_dev->cur_udp_dst++;
2366 if (pkt_dev->cur_udp_dst >= pkt_dev->udp_dst_max)
2367 pkt_dev->cur_udp_dst = pkt_dev->udp_dst_min;
2371 if (!(pkt_dev->flags & F_IPV6)) {
2373 imn = ntohl(pkt_dev->saddr_min);
2374 imx = ntohl(pkt_dev->saddr_max);
2375 if (imn < imx) {
2376 __u32 t;
2377 if (pkt_dev->flags & F_IPSRC_RND)
2378 t = prandom_u32() % (imx - imn) + imn;
2379 else {
2380 t = ntohl(pkt_dev->cur_saddr);
2381 t++;
2382 if (t > imx)
2383 t = imn;
2386 pkt_dev->cur_saddr = htonl(t);
2389 if (pkt_dev->cflows && f_seen(pkt_dev, flow)) {
2390 pkt_dev->cur_daddr = pkt_dev->flows[flow].cur_daddr;
2391 } else {
2392 imn = ntohl(pkt_dev->daddr_min);
2393 imx = ntohl(pkt_dev->daddr_max);
2394 if (imn < imx) {
2395 __u32 t;
2396 __be32 s;
2397 if (pkt_dev->flags & F_IPDST_RND) {
2399 do {
2400 t = prandom_u32() %
2401 (imx - imn) + imn;
2402 s = htonl(t);
2403 } while (ipv4_is_loopback(s) ||
2404 ipv4_is_multicast(s) ||
2405 ipv4_is_lbcast(s) ||
2406 ipv4_is_zeronet(s) ||
2407 ipv4_is_local_multicast(s));
2408 pkt_dev->cur_daddr = s;
2409 } else {
2410 t = ntohl(pkt_dev->cur_daddr);
2411 t++;
2412 if (t > imx) {
2413 t = imn;
2415 pkt_dev->cur_daddr = htonl(t);
2418 if (pkt_dev->cflows) {
2419 pkt_dev->flows[flow].flags |= F_INIT;
2420 pkt_dev->flows[flow].cur_daddr =
2421 pkt_dev->cur_daddr;
2422 #ifdef CONFIG_XFRM
2423 if (pkt_dev->flags & F_IPSEC_ON)
2424 get_ipsec_sa(pkt_dev, flow);
2425 #endif
2426 pkt_dev->nflows++;
2429 } else { /* IPV6 * */
2431 if (!ipv6_addr_any(&pkt_dev->min_in6_daddr)) {
2432 int i;
2434 /* Only random destinations yet */
2436 for (i = 0; i < 4; i++) {
2437 pkt_dev->cur_in6_daddr.s6_addr32[i] =
2438 (((__force __be32)prandom_u32() |
2439 pkt_dev->min_in6_daddr.s6_addr32[i]) &
2440 pkt_dev->max_in6_daddr.s6_addr32[i]);
2445 if (pkt_dev->min_pkt_size < pkt_dev->max_pkt_size) {
2446 __u32 t;
2447 if (pkt_dev->flags & F_TXSIZE_RND) {
2448 t = prandom_u32() %
2449 (pkt_dev->max_pkt_size - pkt_dev->min_pkt_size)
2450 + pkt_dev->min_pkt_size;
2451 } else {
2452 t = pkt_dev->cur_pkt_size + 1;
2453 if (t > pkt_dev->max_pkt_size)
2454 t = pkt_dev->min_pkt_size;
2456 pkt_dev->cur_pkt_size = t;
2459 set_cur_queue_map(pkt_dev);
2461 pkt_dev->flows[flow].count++;
2465 #ifdef CONFIG_XFRM
2466 static int pktgen_output_ipsec(struct sk_buff *skb, struct pktgen_dev *pkt_dev)
2468 struct xfrm_state *x = pkt_dev->flows[pkt_dev->curfl].x;
2469 int err = 0;
2471 if (!x)
2472 return 0;
2473 /* XXX: we dont support tunnel mode for now until
2474 * we resolve the dst issue */
2475 if (x->props.mode != XFRM_MODE_TRANSPORT)
2476 return 0;
2478 spin_lock(&x->lock);
2480 err = x->outer_mode->output(x, skb);
2481 if (err)
2482 goto error;
2483 err = x->type->output(x, skb);
2484 if (err)
2485 goto error;
2487 x->curlft.bytes += skb->len;
2488 x->curlft.packets++;
2489 error:
2490 spin_unlock(&x->lock);
2491 return err;
2494 static void free_SAs(struct pktgen_dev *pkt_dev)
2496 if (pkt_dev->cflows) {
2497 /* let go of the SAs if we have them */
2498 int i;
2499 for (i = 0; i < pkt_dev->cflows; i++) {
2500 struct xfrm_state *x = pkt_dev->flows[i].x;
2501 if (x) {
2502 xfrm_state_put(x);
2503 pkt_dev->flows[i].x = NULL;
2509 static int process_ipsec(struct pktgen_dev *pkt_dev,
2510 struct sk_buff *skb, __be16 protocol)
2512 if (pkt_dev->flags & F_IPSEC_ON) {
2513 struct xfrm_state *x = pkt_dev->flows[pkt_dev->curfl].x;
2514 int nhead = 0;
2515 if (x) {
2516 int ret;
2517 __u8 *eth;
2518 struct iphdr *iph;
2520 nhead = x->props.header_len - skb_headroom(skb);
2521 if (nhead > 0) {
2522 ret = pskb_expand_head(skb, nhead, 0, GFP_ATOMIC);
2523 if (ret < 0) {
2524 pr_err("Error expanding ipsec packet %d\n",
2525 ret);
2526 goto err;
2530 /* ipsec is not expecting ll header */
2531 skb_pull(skb, ETH_HLEN);
2532 ret = pktgen_output_ipsec(skb, pkt_dev);
2533 if (ret) {
2534 pr_err("Error creating ipsec packet %d\n", ret);
2535 goto err;
2537 /* restore ll */
2538 eth = (__u8 *) skb_push(skb, ETH_HLEN);
2539 memcpy(eth, pkt_dev->hh, 12);
2540 *(u16 *) &eth[12] = protocol;
2542 /* Update IPv4 header len as well as checksum value */
2543 iph = ip_hdr(skb);
2544 iph->tot_len = htons(skb->len - ETH_HLEN);
2545 ip_send_check(iph);
2548 return 1;
2549 err:
2550 kfree_skb(skb);
2551 return 0;
2553 #endif
2555 static void mpls_push(__be32 *mpls, struct pktgen_dev *pkt_dev)
2557 unsigned int i;
2558 for (i = 0; i < pkt_dev->nr_labels; i++)
2559 *mpls++ = pkt_dev->labels[i] & ~MPLS_STACK_BOTTOM;
2561 mpls--;
2562 *mpls |= MPLS_STACK_BOTTOM;
2565 static inline __be16 build_tci(unsigned int id, unsigned int cfi,
2566 unsigned int prio)
2568 return htons(id | (cfi << 12) | (prio << 13));
2571 static void pktgen_finalize_skb(struct pktgen_dev *pkt_dev, struct sk_buff *skb,
2572 int datalen)
2574 struct timeval timestamp;
2575 struct pktgen_hdr *pgh;
2577 pgh = (struct pktgen_hdr *)skb_put(skb, sizeof(*pgh));
2578 datalen -= sizeof(*pgh);
2580 if (pkt_dev->nfrags <= 0) {
2581 memset(skb_put(skb, datalen), 0, datalen);
2582 } else {
2583 int frags = pkt_dev->nfrags;
2584 int i, len;
2585 int frag_len;
2588 if (frags > MAX_SKB_FRAGS)
2589 frags = MAX_SKB_FRAGS;
2590 len = datalen - frags * PAGE_SIZE;
2591 if (len > 0) {
2592 memset(skb_put(skb, len), 0, len);
2593 datalen = frags * PAGE_SIZE;
2596 i = 0;
2597 frag_len = (datalen/frags) < PAGE_SIZE ?
2598 (datalen/frags) : PAGE_SIZE;
2599 while (datalen > 0) {
2600 if (unlikely(!pkt_dev->page)) {
2601 int node = numa_node_id();
2603 if (pkt_dev->node >= 0 && (pkt_dev->flags & F_NODE))
2604 node = pkt_dev->node;
2605 pkt_dev->page = alloc_pages_node(node, GFP_KERNEL | __GFP_ZERO, 0);
2606 if (!pkt_dev->page)
2607 break;
2609 get_page(pkt_dev->page);
2610 skb_frag_set_page(skb, i, pkt_dev->page);
2611 skb_shinfo(skb)->frags[i].page_offset = 0;
2612 /*last fragment, fill rest of data*/
2613 if (i == (frags - 1))
2614 skb_frag_size_set(&skb_shinfo(skb)->frags[i],
2615 (datalen < PAGE_SIZE ? datalen : PAGE_SIZE));
2616 else
2617 skb_frag_size_set(&skb_shinfo(skb)->frags[i], frag_len);
2618 datalen -= skb_frag_size(&skb_shinfo(skb)->frags[i]);
2619 skb->len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2620 skb->data_len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2621 i++;
2622 skb_shinfo(skb)->nr_frags = i;
2626 /* Stamp the time, and sequence number,
2627 * convert them to network byte order
2629 pgh->pgh_magic = htonl(PKTGEN_MAGIC);
2630 pgh->seq_num = htonl(pkt_dev->seq_num);
2632 do_gettimeofday(&timestamp);
2633 pgh->tv_sec = htonl(timestamp.tv_sec);
2634 pgh->tv_usec = htonl(timestamp.tv_usec);
2637 static struct sk_buff *fill_packet_ipv4(struct net_device *odev,
2638 struct pktgen_dev *pkt_dev)
2640 struct sk_buff *skb = NULL;
2641 __u8 *eth;
2642 struct udphdr *udph;
2643 int datalen, iplen;
2644 struct iphdr *iph;
2645 __be16 protocol = htons(ETH_P_IP);
2646 __be32 *mpls;
2647 __be16 *vlan_tci = NULL; /* Encapsulates priority and VLAN ID */
2648 __be16 *vlan_encapsulated_proto = NULL; /* packet type ID field (or len) for VLAN tag */
2649 __be16 *svlan_tci = NULL; /* Encapsulates priority and SVLAN ID */
2650 __be16 *svlan_encapsulated_proto = NULL; /* packet type ID field (or len) for SVLAN tag */
2651 u16 queue_map;
2653 if (pkt_dev->nr_labels)
2654 protocol = htons(ETH_P_MPLS_UC);
2656 if (pkt_dev->vlan_id != 0xffff)
2657 protocol = htons(ETH_P_8021Q);
2659 /* Update any of the values, used when we're incrementing various
2660 * fields.
2662 mod_cur_headers(pkt_dev);
2663 queue_map = pkt_dev->cur_queue_map;
2665 datalen = (odev->hard_header_len + 16) & ~0xf;
2667 if (pkt_dev->flags & F_NODE) {
2668 int node;
2670 if (pkt_dev->node >= 0)
2671 node = pkt_dev->node;
2672 else
2673 node = numa_node_id();
2675 skb = __alloc_skb(NET_SKB_PAD + pkt_dev->cur_pkt_size + 64
2676 + datalen + pkt_dev->pkt_overhead, GFP_NOWAIT, 0, node);
2677 if (likely(skb)) {
2678 skb_reserve(skb, NET_SKB_PAD);
2679 skb->dev = odev;
2682 else
2683 skb = __netdev_alloc_skb(odev,
2684 pkt_dev->cur_pkt_size + 64
2685 + datalen + pkt_dev->pkt_overhead, GFP_NOWAIT);
2687 if (!skb) {
2688 sprintf(pkt_dev->result, "No memory");
2689 return NULL;
2691 prefetchw(skb->data);
2693 skb_reserve(skb, datalen);
2695 /* Reserve for ethernet and IP header */
2696 eth = (__u8 *) skb_push(skb, 14);
2697 mpls = (__be32 *)skb_put(skb, pkt_dev->nr_labels*sizeof(__u32));
2698 if (pkt_dev->nr_labels)
2699 mpls_push(mpls, pkt_dev);
2701 if (pkt_dev->vlan_id != 0xffff) {
2702 if (pkt_dev->svlan_id != 0xffff) {
2703 svlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2704 *svlan_tci = build_tci(pkt_dev->svlan_id,
2705 pkt_dev->svlan_cfi,
2706 pkt_dev->svlan_p);
2707 svlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2708 *svlan_encapsulated_proto = htons(ETH_P_8021Q);
2710 vlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2711 *vlan_tci = build_tci(pkt_dev->vlan_id,
2712 pkt_dev->vlan_cfi,
2713 pkt_dev->vlan_p);
2714 vlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2715 *vlan_encapsulated_proto = htons(ETH_P_IP);
2718 skb->network_header = skb->tail;
2719 skb->transport_header = skb->network_header + sizeof(struct iphdr);
2720 skb_put(skb, sizeof(struct iphdr) + sizeof(struct udphdr));
2721 skb_set_queue_mapping(skb, queue_map);
2722 skb->priority = pkt_dev->skb_priority;
2724 iph = ip_hdr(skb);
2725 udph = udp_hdr(skb);
2727 memcpy(eth, pkt_dev->hh, 12);
2728 *(__be16 *) & eth[12] = protocol;
2730 /* Eth + IPh + UDPh + mpls */
2731 datalen = pkt_dev->cur_pkt_size - 14 - 20 - 8 -
2732 pkt_dev->pkt_overhead;
2733 if (datalen < 0 || datalen < sizeof(struct pktgen_hdr))
2734 datalen = sizeof(struct pktgen_hdr);
2736 udph->source = htons(pkt_dev->cur_udp_src);
2737 udph->dest = htons(pkt_dev->cur_udp_dst);
2738 udph->len = htons(datalen + 8); /* DATA + udphdr */
2739 udph->check = 0; /* No checksum */
2741 iph->ihl = 5;
2742 iph->version = 4;
2743 iph->ttl = 32;
2744 iph->tos = pkt_dev->tos;
2745 iph->protocol = IPPROTO_UDP; /* UDP */
2746 iph->saddr = pkt_dev->cur_saddr;
2747 iph->daddr = pkt_dev->cur_daddr;
2748 iph->id = htons(pkt_dev->ip_id);
2749 pkt_dev->ip_id++;
2750 iph->frag_off = 0;
2751 iplen = 20 + 8 + datalen;
2752 iph->tot_len = htons(iplen);
2753 iph->check = 0;
2754 iph->check = ip_fast_csum((void *)iph, iph->ihl);
2755 skb->protocol = protocol;
2756 skb->mac_header = (skb->network_header - ETH_HLEN -
2757 pkt_dev->pkt_overhead);
2758 skb->dev = odev;
2759 skb->pkt_type = PACKET_HOST;
2760 pktgen_finalize_skb(pkt_dev, skb, datalen);
2762 #ifdef CONFIG_XFRM
2763 if (!process_ipsec(pkt_dev, skb, protocol))
2764 return NULL;
2765 #endif
2767 return skb;
2770 static struct sk_buff *fill_packet_ipv6(struct net_device *odev,
2771 struct pktgen_dev *pkt_dev)
2773 struct sk_buff *skb = NULL;
2774 __u8 *eth;
2775 struct udphdr *udph;
2776 int datalen;
2777 struct ipv6hdr *iph;
2778 __be16 protocol = htons(ETH_P_IPV6);
2779 __be32 *mpls;
2780 __be16 *vlan_tci = NULL; /* Encapsulates priority and VLAN ID */
2781 __be16 *vlan_encapsulated_proto = NULL; /* packet type ID field (or len) for VLAN tag */
2782 __be16 *svlan_tci = NULL; /* Encapsulates priority and SVLAN ID */
2783 __be16 *svlan_encapsulated_proto = NULL; /* packet type ID field (or len) for SVLAN tag */
2784 u16 queue_map;
2786 if (pkt_dev->nr_labels)
2787 protocol = htons(ETH_P_MPLS_UC);
2789 if (pkt_dev->vlan_id != 0xffff)
2790 protocol = htons(ETH_P_8021Q);
2792 /* Update any of the values, used when we're incrementing various
2793 * fields.
2795 mod_cur_headers(pkt_dev);
2796 queue_map = pkt_dev->cur_queue_map;
2798 skb = __netdev_alloc_skb(odev,
2799 pkt_dev->cur_pkt_size + 64
2800 + 16 + pkt_dev->pkt_overhead, GFP_NOWAIT);
2801 if (!skb) {
2802 sprintf(pkt_dev->result, "No memory");
2803 return NULL;
2805 prefetchw(skb->data);
2807 skb_reserve(skb, 16);
2809 /* Reserve for ethernet and IP header */
2810 eth = (__u8 *) skb_push(skb, 14);
2811 mpls = (__be32 *)skb_put(skb, pkt_dev->nr_labels*sizeof(__u32));
2812 if (pkt_dev->nr_labels)
2813 mpls_push(mpls, pkt_dev);
2815 if (pkt_dev->vlan_id != 0xffff) {
2816 if (pkt_dev->svlan_id != 0xffff) {
2817 svlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2818 *svlan_tci = build_tci(pkt_dev->svlan_id,
2819 pkt_dev->svlan_cfi,
2820 pkt_dev->svlan_p);
2821 svlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2822 *svlan_encapsulated_proto = htons(ETH_P_8021Q);
2824 vlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2825 *vlan_tci = build_tci(pkt_dev->vlan_id,
2826 pkt_dev->vlan_cfi,
2827 pkt_dev->vlan_p);
2828 vlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2829 *vlan_encapsulated_proto = htons(ETH_P_IPV6);
2832 skb->network_header = skb->tail;
2833 skb->transport_header = skb->network_header + sizeof(struct ipv6hdr);
2834 skb_put(skb, sizeof(struct ipv6hdr) + sizeof(struct udphdr));
2835 skb_set_queue_mapping(skb, queue_map);
2836 skb->priority = pkt_dev->skb_priority;
2837 iph = ipv6_hdr(skb);
2838 udph = udp_hdr(skb);
2840 memcpy(eth, pkt_dev->hh, 12);
2841 *(__be16 *) &eth[12] = protocol;
2843 /* Eth + IPh + UDPh + mpls */
2844 datalen = pkt_dev->cur_pkt_size - 14 -
2845 sizeof(struct ipv6hdr) - sizeof(struct udphdr) -
2846 pkt_dev->pkt_overhead;
2848 if (datalen < 0 || datalen < sizeof(struct pktgen_hdr)) {
2849 datalen = sizeof(struct pktgen_hdr);
2850 net_info_ratelimited("increased datalen to %d\n", datalen);
2853 udph->source = htons(pkt_dev->cur_udp_src);
2854 udph->dest = htons(pkt_dev->cur_udp_dst);
2855 udph->len = htons(datalen + sizeof(struct udphdr));
2856 udph->check = 0; /* No checksum */
2858 *(__be32 *) iph = htonl(0x60000000); /* Version + flow */
2860 if (pkt_dev->traffic_class) {
2861 /* Version + traffic class + flow (0) */
2862 *(__be32 *)iph |= htonl(0x60000000 | (pkt_dev->traffic_class << 20));
2865 iph->hop_limit = 32;
2867 iph->payload_len = htons(sizeof(struct udphdr) + datalen);
2868 iph->nexthdr = IPPROTO_UDP;
2870 iph->daddr = pkt_dev->cur_in6_daddr;
2871 iph->saddr = pkt_dev->cur_in6_saddr;
2873 skb->mac_header = (skb->network_header - ETH_HLEN -
2874 pkt_dev->pkt_overhead);
2875 skb->protocol = protocol;
2876 skb->dev = odev;
2877 skb->pkt_type = PACKET_HOST;
2879 pktgen_finalize_skb(pkt_dev, skb, datalen);
2881 return skb;
2884 static struct sk_buff *fill_packet(struct net_device *odev,
2885 struct pktgen_dev *pkt_dev)
2887 if (pkt_dev->flags & F_IPV6)
2888 return fill_packet_ipv6(odev, pkt_dev);
2889 else
2890 return fill_packet_ipv4(odev, pkt_dev);
2893 static void pktgen_clear_counters(struct pktgen_dev *pkt_dev)
2895 pkt_dev->seq_num = 1;
2896 pkt_dev->idle_acc = 0;
2897 pkt_dev->sofar = 0;
2898 pkt_dev->tx_bytes = 0;
2899 pkt_dev->errors = 0;
2902 /* Set up structure for sending pkts, clear counters */
2904 static void pktgen_run(struct pktgen_thread *t)
2906 struct pktgen_dev *pkt_dev;
2907 int started = 0;
2909 func_enter();
2911 if_lock(t);
2912 list_for_each_entry(pkt_dev, &t->if_list, list) {
2915 * setup odev and create initial packet.
2917 pktgen_setup_inject(pkt_dev);
2919 if (pkt_dev->odev) {
2920 pktgen_clear_counters(pkt_dev);
2921 pkt_dev->running = 1; /* Cranke yeself! */
2922 pkt_dev->skb = NULL;
2923 pkt_dev->started_at = pkt_dev->next_tx = ktime_get();
2925 set_pkt_overhead(pkt_dev);
2927 strcpy(pkt_dev->result, "Starting");
2928 started++;
2929 } else
2930 strcpy(pkt_dev->result, "Error starting");
2932 if_unlock(t);
2933 if (started)
2934 t->control &= ~(T_STOP);
2937 static void pktgen_stop_all_threads_ifs(struct pktgen_net *pn)
2939 struct pktgen_thread *t;
2941 func_enter();
2943 mutex_lock(&pktgen_thread_lock);
2945 list_for_each_entry(t, &pn->pktgen_threads, th_list)
2946 t->control |= T_STOP;
2948 mutex_unlock(&pktgen_thread_lock);
2951 static int thread_is_running(const struct pktgen_thread *t)
2953 const struct pktgen_dev *pkt_dev;
2955 list_for_each_entry(pkt_dev, &t->if_list, list)
2956 if (pkt_dev->running)
2957 return 1;
2958 return 0;
2961 static int pktgen_wait_thread_run(struct pktgen_thread *t)
2963 if_lock(t);
2965 while (thread_is_running(t)) {
2967 if_unlock(t);
2969 msleep_interruptible(100);
2971 if (signal_pending(current))
2972 goto signal;
2973 if_lock(t);
2975 if_unlock(t);
2976 return 1;
2977 signal:
2978 return 0;
2981 static int pktgen_wait_all_threads_run(struct pktgen_net *pn)
2983 struct pktgen_thread *t;
2984 int sig = 1;
2986 mutex_lock(&pktgen_thread_lock);
2988 list_for_each_entry(t, &pn->pktgen_threads, th_list) {
2989 sig = pktgen_wait_thread_run(t);
2990 if (sig == 0)
2991 break;
2994 if (sig == 0)
2995 list_for_each_entry(t, &pn->pktgen_threads, th_list)
2996 t->control |= (T_STOP);
2998 mutex_unlock(&pktgen_thread_lock);
2999 return sig;
3002 static void pktgen_run_all_threads(struct pktgen_net *pn)
3004 struct pktgen_thread *t;
3006 func_enter();
3008 mutex_lock(&pktgen_thread_lock);
3010 list_for_each_entry(t, &pn->pktgen_threads, th_list)
3011 t->control |= (T_RUN);
3013 mutex_unlock(&pktgen_thread_lock);
3015 /* Propagate thread->control */
3016 schedule_timeout_interruptible(msecs_to_jiffies(125));
3018 pktgen_wait_all_threads_run(pn);
3021 static void pktgen_reset_all_threads(struct pktgen_net *pn)
3023 struct pktgen_thread *t;
3025 func_enter();
3027 mutex_lock(&pktgen_thread_lock);
3029 list_for_each_entry(t, &pn->pktgen_threads, th_list)
3030 t->control |= (T_REMDEVALL);
3032 mutex_unlock(&pktgen_thread_lock);
3034 /* Propagate thread->control */
3035 schedule_timeout_interruptible(msecs_to_jiffies(125));
3037 pktgen_wait_all_threads_run(pn);
3040 static void show_results(struct pktgen_dev *pkt_dev, int nr_frags)
3042 __u64 bps, mbps, pps;
3043 char *p = pkt_dev->result;
3044 ktime_t elapsed = ktime_sub(pkt_dev->stopped_at,
3045 pkt_dev->started_at);
3046 ktime_t idle = ns_to_ktime(pkt_dev->idle_acc);
3048 p += sprintf(p, "OK: %llu(c%llu+d%llu) usec, %llu (%dbyte,%dfrags)\n",
3049 (unsigned long long)ktime_to_us(elapsed),
3050 (unsigned long long)ktime_to_us(ktime_sub(elapsed, idle)),
3051 (unsigned long long)ktime_to_us(idle),
3052 (unsigned long long)pkt_dev->sofar,
3053 pkt_dev->cur_pkt_size, nr_frags);
3055 pps = div64_u64(pkt_dev->sofar * NSEC_PER_SEC,
3056 ktime_to_ns(elapsed));
3058 bps = pps * 8 * pkt_dev->cur_pkt_size;
3060 mbps = bps;
3061 do_div(mbps, 1000000);
3062 p += sprintf(p, " %llupps %lluMb/sec (%llubps) errors: %llu",
3063 (unsigned long long)pps,
3064 (unsigned long long)mbps,
3065 (unsigned long long)bps,
3066 (unsigned long long)pkt_dev->errors);
3069 /* Set stopped-at timer, remove from running list, do counters & statistics */
3070 static int pktgen_stop_device(struct pktgen_dev *pkt_dev)
3072 int nr_frags = pkt_dev->skb ? skb_shinfo(pkt_dev->skb)->nr_frags : -1;
3074 if (!pkt_dev->running) {
3075 pr_warning("interface: %s is already stopped\n",
3076 pkt_dev->odevname);
3077 return -EINVAL;
3080 kfree_skb(pkt_dev->skb);
3081 pkt_dev->skb = NULL;
3082 pkt_dev->stopped_at = ktime_get();
3083 pkt_dev->running = 0;
3085 show_results(pkt_dev, nr_frags);
3087 return 0;
3090 static struct pktgen_dev *next_to_run(struct pktgen_thread *t)
3092 struct pktgen_dev *pkt_dev, *best = NULL;
3094 if_lock(t);
3096 list_for_each_entry(pkt_dev, &t->if_list, list) {
3097 if (!pkt_dev->running)
3098 continue;
3099 if (best == NULL)
3100 best = pkt_dev;
3101 else if (ktime_compare(pkt_dev->next_tx, best->next_tx) < 0)
3102 best = pkt_dev;
3104 if_unlock(t);
3105 return best;
3108 static void pktgen_stop(struct pktgen_thread *t)
3110 struct pktgen_dev *pkt_dev;
3112 func_enter();
3114 if_lock(t);
3116 list_for_each_entry(pkt_dev, &t->if_list, list) {
3117 pktgen_stop_device(pkt_dev);
3120 if_unlock(t);
3124 * one of our devices needs to be removed - find it
3125 * and remove it
3127 static void pktgen_rem_one_if(struct pktgen_thread *t)
3129 struct list_head *q, *n;
3130 struct pktgen_dev *cur;
3132 func_enter();
3134 if_lock(t);
3136 list_for_each_safe(q, n, &t->if_list) {
3137 cur = list_entry(q, struct pktgen_dev, list);
3139 if (!cur->removal_mark)
3140 continue;
3142 kfree_skb(cur->skb);
3143 cur->skb = NULL;
3145 pktgen_remove_device(t, cur);
3147 break;
3150 if_unlock(t);
3153 static void pktgen_rem_all_ifs(struct pktgen_thread *t)
3155 struct list_head *q, *n;
3156 struct pktgen_dev *cur;
3158 func_enter();
3160 /* Remove all devices, free mem */
3162 if_lock(t);
3164 list_for_each_safe(q, n, &t->if_list) {
3165 cur = list_entry(q, struct pktgen_dev, list);
3167 kfree_skb(cur->skb);
3168 cur->skb = NULL;
3170 pktgen_remove_device(t, cur);
3173 if_unlock(t);
3176 static void pktgen_rem_thread(struct pktgen_thread *t)
3178 /* Remove from the thread list */
3179 remove_proc_entry(t->tsk->comm, t->net->proc_dir);
3182 static void pktgen_resched(struct pktgen_dev *pkt_dev)
3184 ktime_t idle_start = ktime_get();
3185 schedule();
3186 pkt_dev->idle_acc += ktime_to_ns(ktime_sub(ktime_get(), idle_start));
3189 static void pktgen_wait_for_skb(struct pktgen_dev *pkt_dev)
3191 ktime_t idle_start = ktime_get();
3193 while (atomic_read(&(pkt_dev->skb->users)) != 1) {
3194 if (signal_pending(current))
3195 break;
3197 if (need_resched())
3198 pktgen_resched(pkt_dev);
3199 else
3200 cpu_relax();
3202 pkt_dev->idle_acc += ktime_to_ns(ktime_sub(ktime_get(), idle_start));
3205 static void pktgen_xmit(struct pktgen_dev *pkt_dev)
3207 struct net_device *odev = pkt_dev->odev;
3208 netdev_tx_t (*xmit)(struct sk_buff *, struct net_device *)
3209 = odev->netdev_ops->ndo_start_xmit;
3210 struct netdev_queue *txq;
3211 u16 queue_map;
3212 int ret;
3214 /* If device is offline, then don't send */
3215 if (unlikely(!netif_running(odev) || !netif_carrier_ok(odev))) {
3216 pktgen_stop_device(pkt_dev);
3217 return;
3220 /* This is max DELAY, this has special meaning of
3221 * "never transmit"
3223 if (unlikely(pkt_dev->delay == ULLONG_MAX)) {
3224 pkt_dev->next_tx = ktime_add_ns(ktime_get(), ULONG_MAX);
3225 return;
3228 /* If no skb or clone count exhausted then get new one */
3229 if (!pkt_dev->skb || (pkt_dev->last_ok &&
3230 ++pkt_dev->clone_count >= pkt_dev->clone_skb)) {
3231 /* build a new pkt */
3232 kfree_skb(pkt_dev->skb);
3234 pkt_dev->skb = fill_packet(odev, pkt_dev);
3235 if (pkt_dev->skb == NULL) {
3236 pr_err("ERROR: couldn't allocate skb in fill_packet\n");
3237 schedule();
3238 pkt_dev->clone_count--; /* back out increment, OOM */
3239 return;
3241 pkt_dev->last_pkt_size = pkt_dev->skb->len;
3242 pkt_dev->allocated_skbs++;
3243 pkt_dev->clone_count = 0; /* reset counter */
3246 if (pkt_dev->delay && pkt_dev->last_ok)
3247 spin(pkt_dev, pkt_dev->next_tx);
3249 queue_map = skb_get_queue_mapping(pkt_dev->skb);
3250 txq = netdev_get_tx_queue(odev, queue_map);
3252 __netif_tx_lock_bh(txq);
3254 if (unlikely(netif_xmit_frozen_or_stopped(txq))) {
3255 ret = NETDEV_TX_BUSY;
3256 pkt_dev->last_ok = 0;
3257 goto unlock;
3259 atomic_inc(&(pkt_dev->skb->users));
3260 ret = (*xmit)(pkt_dev->skb, odev);
3262 switch (ret) {
3263 case NETDEV_TX_OK:
3264 txq_trans_update(txq);
3265 pkt_dev->last_ok = 1;
3266 pkt_dev->sofar++;
3267 pkt_dev->seq_num++;
3268 pkt_dev->tx_bytes += pkt_dev->last_pkt_size;
3269 break;
3270 case NET_XMIT_DROP:
3271 case NET_XMIT_CN:
3272 case NET_XMIT_POLICED:
3273 /* skb has been consumed */
3274 pkt_dev->errors++;
3275 break;
3276 default: /* Drivers are not supposed to return other values! */
3277 net_info_ratelimited("%s xmit error: %d\n",
3278 pkt_dev->odevname, ret);
3279 pkt_dev->errors++;
3280 /* fallthru */
3281 case NETDEV_TX_LOCKED:
3282 case NETDEV_TX_BUSY:
3283 /* Retry it next time */
3284 atomic_dec(&(pkt_dev->skb->users));
3285 pkt_dev->last_ok = 0;
3287 unlock:
3288 __netif_tx_unlock_bh(txq);
3290 /* If pkt_dev->count is zero, then run forever */
3291 if ((pkt_dev->count != 0) && (pkt_dev->sofar >= pkt_dev->count)) {
3292 pktgen_wait_for_skb(pkt_dev);
3294 /* Done with this */
3295 pktgen_stop_device(pkt_dev);
3300 * Main loop of the thread goes here
3303 static int pktgen_thread_worker(void *arg)
3305 DEFINE_WAIT(wait);
3306 struct pktgen_thread *t = arg;
3307 struct pktgen_dev *pkt_dev = NULL;
3308 int cpu = t->cpu;
3310 BUG_ON(smp_processor_id() != cpu);
3312 init_waitqueue_head(&t->queue);
3313 complete(&t->start_done);
3315 pr_debug("starting pktgen/%d: pid=%d\n", cpu, task_pid_nr(current));
3317 set_current_state(TASK_INTERRUPTIBLE);
3319 set_freezable();
3321 while (!kthread_should_stop()) {
3322 pkt_dev = next_to_run(t);
3324 if (unlikely(!pkt_dev && t->control == 0)) {
3325 if (t->net->pktgen_exiting)
3326 break;
3327 wait_event_interruptible_timeout(t->queue,
3328 t->control != 0,
3329 HZ/10);
3330 try_to_freeze();
3331 continue;
3334 __set_current_state(TASK_RUNNING);
3336 if (likely(pkt_dev)) {
3337 pktgen_xmit(pkt_dev);
3339 if (need_resched())
3340 pktgen_resched(pkt_dev);
3341 else
3342 cpu_relax();
3345 if (t->control & T_STOP) {
3346 pktgen_stop(t);
3347 t->control &= ~(T_STOP);
3350 if (t->control & T_RUN) {
3351 pktgen_run(t);
3352 t->control &= ~(T_RUN);
3355 if (t->control & T_REMDEVALL) {
3356 pktgen_rem_all_ifs(t);
3357 t->control &= ~(T_REMDEVALL);
3360 if (t->control & T_REMDEV) {
3361 pktgen_rem_one_if(t);
3362 t->control &= ~(T_REMDEV);
3365 try_to_freeze();
3367 set_current_state(TASK_INTERRUPTIBLE);
3370 pr_debug("%s stopping all device\n", t->tsk->comm);
3371 pktgen_stop(t);
3373 pr_debug("%s removing all device\n", t->tsk->comm);
3374 pktgen_rem_all_ifs(t);
3376 pr_debug("%s removing thread\n", t->tsk->comm);
3377 pktgen_rem_thread(t);
3379 /* Wait for kthread_stop */
3380 for (;;) {
3381 set_current_state(TASK_INTERRUPTIBLE);
3382 if (kthread_should_stop())
3383 break;
3384 schedule();
3386 __set_current_state(TASK_RUNNING);
3388 return 0;
3391 static struct pktgen_dev *pktgen_find_dev(struct pktgen_thread *t,
3392 const char *ifname, bool exact)
3394 struct pktgen_dev *p, *pkt_dev = NULL;
3395 size_t len = strlen(ifname);
3397 if_lock(t);
3398 list_for_each_entry(p, &t->if_list, list)
3399 if (strncmp(p->odevname, ifname, len) == 0) {
3400 if (p->odevname[len]) {
3401 if (exact || p->odevname[len] != '@')
3402 continue;
3404 pkt_dev = p;
3405 break;
3408 if_unlock(t);
3409 pr_debug("find_dev(%s) returning %p\n", ifname, pkt_dev);
3410 return pkt_dev;
3414 * Adds a dev at front of if_list.
3417 static int add_dev_to_thread(struct pktgen_thread *t,
3418 struct pktgen_dev *pkt_dev)
3420 int rv = 0;
3422 if_lock(t);
3424 if (pkt_dev->pg_thread) {
3425 pr_err("ERROR: already assigned to a thread\n");
3426 rv = -EBUSY;
3427 goto out;
3430 list_add(&pkt_dev->list, &t->if_list);
3431 pkt_dev->pg_thread = t;
3432 pkt_dev->running = 0;
3434 out:
3435 if_unlock(t);
3436 return rv;
3439 /* Called under thread lock */
3441 static int pktgen_add_device(struct pktgen_thread *t, const char *ifname)
3443 struct pktgen_dev *pkt_dev;
3444 int err;
3445 int node = cpu_to_node(t->cpu);
3447 /* We don't allow a device to be on several threads */
3449 pkt_dev = __pktgen_NN_threads(t->net, ifname, FIND);
3450 if (pkt_dev) {
3451 pr_err("ERROR: interface already used\n");
3452 return -EBUSY;
3455 pkt_dev = kzalloc_node(sizeof(struct pktgen_dev), GFP_KERNEL, node);
3456 if (!pkt_dev)
3457 return -ENOMEM;
3459 strcpy(pkt_dev->odevname, ifname);
3460 pkt_dev->flows = vzalloc_node(MAX_CFLOWS * sizeof(struct flow_state),
3461 node);
3462 if (pkt_dev->flows == NULL) {
3463 kfree(pkt_dev);
3464 return -ENOMEM;
3467 pkt_dev->removal_mark = 0;
3468 pkt_dev->nfrags = 0;
3469 pkt_dev->delay = pg_delay_d;
3470 pkt_dev->count = pg_count_d;
3471 pkt_dev->sofar = 0;
3472 pkt_dev->udp_src_min = 9; /* sink port */
3473 pkt_dev->udp_src_max = 9;
3474 pkt_dev->udp_dst_min = 9;
3475 pkt_dev->udp_dst_max = 9;
3476 pkt_dev->vlan_p = 0;
3477 pkt_dev->vlan_cfi = 0;
3478 pkt_dev->vlan_id = 0xffff;
3479 pkt_dev->svlan_p = 0;
3480 pkt_dev->svlan_cfi = 0;
3481 pkt_dev->svlan_id = 0xffff;
3482 pkt_dev->node = -1;
3484 err = pktgen_setup_dev(t->net, pkt_dev, ifname);
3485 if (err)
3486 goto out1;
3487 if (pkt_dev->odev->priv_flags & IFF_TX_SKB_SHARING)
3488 pkt_dev->clone_skb = pg_clone_skb_d;
3490 pkt_dev->entry = proc_create_data(ifname, 0600, t->net->proc_dir,
3491 &pktgen_if_fops, pkt_dev);
3492 if (!pkt_dev->entry) {
3493 pr_err("cannot create %s/%s procfs entry\n",
3494 PG_PROC_DIR, ifname);
3495 err = -EINVAL;
3496 goto out2;
3498 #ifdef CONFIG_XFRM
3499 pkt_dev->ipsmode = XFRM_MODE_TRANSPORT;
3500 pkt_dev->ipsproto = IPPROTO_ESP;
3501 #endif
3503 return add_dev_to_thread(t, pkt_dev);
3504 out2:
3505 dev_put(pkt_dev->odev);
3506 out1:
3507 #ifdef CONFIG_XFRM
3508 free_SAs(pkt_dev);
3509 #endif
3510 vfree(pkt_dev->flows);
3511 kfree(pkt_dev);
3512 return err;
3515 static int __net_init pktgen_create_thread(int cpu, struct pktgen_net *pn)
3517 struct pktgen_thread *t;
3518 struct proc_dir_entry *pe;
3519 struct task_struct *p;
3521 t = kzalloc_node(sizeof(struct pktgen_thread), GFP_KERNEL,
3522 cpu_to_node(cpu));
3523 if (!t) {
3524 pr_err("ERROR: out of memory, can't create new thread\n");
3525 return -ENOMEM;
3528 spin_lock_init(&t->if_lock);
3529 t->cpu = cpu;
3531 INIT_LIST_HEAD(&t->if_list);
3533 list_add_tail(&t->th_list, &pn->pktgen_threads);
3534 init_completion(&t->start_done);
3536 p = kthread_create_on_node(pktgen_thread_worker,
3538 cpu_to_node(cpu),
3539 "kpktgend_%d", cpu);
3540 if (IS_ERR(p)) {
3541 pr_err("kernel_thread() failed for cpu %d\n", t->cpu);
3542 list_del(&t->th_list);
3543 kfree(t);
3544 return PTR_ERR(p);
3546 kthread_bind(p, cpu);
3547 t->tsk = p;
3549 pe = proc_create_data(t->tsk->comm, 0600, pn->proc_dir,
3550 &pktgen_thread_fops, t);
3551 if (!pe) {
3552 pr_err("cannot create %s/%s procfs entry\n",
3553 PG_PROC_DIR, t->tsk->comm);
3554 kthread_stop(p);
3555 list_del(&t->th_list);
3556 kfree(t);
3557 return -EINVAL;
3560 t->net = pn;
3561 wake_up_process(p);
3562 wait_for_completion(&t->start_done);
3564 return 0;
3568 * Removes a device from the thread if_list.
3570 static void _rem_dev_from_if_list(struct pktgen_thread *t,
3571 struct pktgen_dev *pkt_dev)
3573 struct list_head *q, *n;
3574 struct pktgen_dev *p;
3576 list_for_each_safe(q, n, &t->if_list) {
3577 p = list_entry(q, struct pktgen_dev, list);
3578 if (p == pkt_dev)
3579 list_del(&p->list);
3583 static int pktgen_remove_device(struct pktgen_thread *t,
3584 struct pktgen_dev *pkt_dev)
3586 pr_debug("remove_device pkt_dev=%p\n", pkt_dev);
3588 if (pkt_dev->running) {
3589 pr_warning("WARNING: trying to remove a running interface, stopping it now\n");
3590 pktgen_stop_device(pkt_dev);
3593 /* Dis-associate from the interface */
3595 if (pkt_dev->odev) {
3596 dev_put(pkt_dev->odev);
3597 pkt_dev->odev = NULL;
3600 /* And update the thread if_list */
3602 _rem_dev_from_if_list(t, pkt_dev);
3604 if (pkt_dev->entry)
3605 proc_remove(pkt_dev->entry);
3607 #ifdef CONFIG_XFRM
3608 free_SAs(pkt_dev);
3609 #endif
3610 vfree(pkt_dev->flows);
3611 if (pkt_dev->page)
3612 put_page(pkt_dev->page);
3613 kfree(pkt_dev);
3614 return 0;
3617 static int __net_init pg_net_init(struct net *net)
3619 struct pktgen_net *pn = net_generic(net, pg_net_id);
3620 struct proc_dir_entry *pe;
3621 int cpu, ret = 0;
3623 pn->net = net;
3624 INIT_LIST_HEAD(&pn->pktgen_threads);
3625 pn->pktgen_exiting = false;
3626 pn->proc_dir = proc_mkdir(PG_PROC_DIR, pn->net->proc_net);
3627 if (!pn->proc_dir) {
3628 pr_warn("cannot create /proc/net/%s\n", PG_PROC_DIR);
3629 return -ENODEV;
3631 pe = proc_create(PGCTRL, 0600, pn->proc_dir, &pktgen_fops);
3632 if (pe == NULL) {
3633 pr_err("cannot create %s procfs entry\n", PGCTRL);
3634 ret = -EINVAL;
3635 goto remove;
3638 for_each_online_cpu(cpu) {
3639 int err;
3641 err = pktgen_create_thread(cpu, pn);
3642 if (err)
3643 pr_warn("Cannot create thread for cpu %d (%d)\n",
3644 cpu, err);
3647 if (list_empty(&pn->pktgen_threads)) {
3648 pr_err("Initialization failed for all threads\n");
3649 ret = -ENODEV;
3650 goto remove_entry;
3653 return 0;
3655 remove_entry:
3656 remove_proc_entry(PGCTRL, pn->proc_dir);
3657 remove:
3658 remove_proc_entry(PG_PROC_DIR, pn->net->proc_net);
3659 return ret;
3662 static void __net_exit pg_net_exit(struct net *net)
3664 struct pktgen_net *pn = net_generic(net, pg_net_id);
3665 struct pktgen_thread *t;
3666 struct list_head *q, *n;
3667 LIST_HEAD(list);
3669 /* Stop all interfaces & threads */
3670 pn->pktgen_exiting = true;
3672 mutex_lock(&pktgen_thread_lock);
3673 list_splice_init(&pn->pktgen_threads, &list);
3674 mutex_unlock(&pktgen_thread_lock);
3676 list_for_each_safe(q, n, &list) {
3677 t = list_entry(q, struct pktgen_thread, th_list);
3678 list_del(&t->th_list);
3679 kthread_stop(t->tsk);
3680 kfree(t);
3683 remove_proc_entry(PGCTRL, pn->proc_dir);
3684 remove_proc_entry(PG_PROC_DIR, pn->net->proc_net);
3687 static struct pernet_operations pg_net_ops = {
3688 .init = pg_net_init,
3689 .exit = pg_net_exit,
3690 .id = &pg_net_id,
3691 .size = sizeof(struct pktgen_net),
3694 static int __init pg_init(void)
3696 int ret = 0;
3698 pr_info("%s", version);
3699 ret = register_pernet_subsys(&pg_net_ops);
3700 if (ret)
3701 return ret;
3702 ret = register_netdevice_notifier(&pktgen_notifier_block);
3703 if (ret)
3704 unregister_pernet_subsys(&pg_net_ops);
3706 return ret;
3709 static void __exit pg_cleanup(void)
3711 unregister_netdevice_notifier(&pktgen_notifier_block);
3712 unregister_pernet_subsys(&pg_net_ops);
3715 module_init(pg_init);
3716 module_exit(pg_cleanup);
3718 MODULE_AUTHOR("Robert Olsson <robert.olsson@its.uu.se>");
3719 MODULE_DESCRIPTION("Packet Generator tool");
3720 MODULE_LICENSE("GPL");
3721 MODULE_VERSION(VERSION);
3722 module_param(pg_count_d, int, 0);
3723 MODULE_PARM_DESC(pg_count_d, "Default number of packets to inject");
3724 module_param(pg_delay_d, int, 0);
3725 MODULE_PARM_DESC(pg_delay_d, "Default delay between packets (nanoseconds)");
3726 module_param(pg_clone_skb_d, int, 0);
3727 MODULE_PARM_DESC(pg_clone_skb_d, "Default number of copies of the same packet");
3728 module_param(debug, int, 0);
3729 MODULE_PARM_DESC(debug, "Enable debugging of pktgen module");