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
6 *
7 * Alexey Kuznetsov  <kuznet@ms2.inr.ac.ru>
8 * Ben Greear <greearb@candelatech.com>
9 * Jens Låås <jens.laas@data.slu.se>
10 *
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.
15 *
16 *
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.
22 *
23 * Additional hacking by:
24 *
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>
46 *
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.
53 *
54 * Also moved to /proc/net/pktgen/
55 * --ro
56 *
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>
60 *
61 * Integrated to 2.5.x 021029 --Lucio Maciel (luciomaciel@zipmail.com.br)
62 *
63 *
64 * 021124 Finished major redesign and rewrite for new functionality.
65 * See Documentation/networking/pktgen.txt for how to use this.
66 *
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.
73 * The if_list is RCU protected, and the if_lock remains to protect updating
74 * of if_list, from "add_device" as it invoked from userspace (via proc write).
75 *
76 * By design there should only be *one* "controlling" process. In practice
77 * multiple write accesses gives unpredictable result. Understood by "write"
78 * to /proc gives result code thats should be read be the "writer".
79 * For practical use this should be no problem.
80 *
81 * Note when adding devices to a specific CPU there good idea to also assign
82 * /proc/irq/XX/smp_affinity so TX-interrupts gets bound to the same CPU.
83 * --ro
84 *
85 * Fix refcount off by one if first packet fails, potential null deref,
86 * memleak 030710- KJP
87 *
88 * First "ranges" functionality for ipv6 030726 --ro
89 *
90 * Included flow support. 030802 ANK.
91 *
92 * Fixed unaligned access on IA-64 Grant Grundler <grundler@parisc-linux.org>
93 *
94 * Remove if fix from added Harald Welte <laforge@netfilter.org> 040419
95 * ia64 compilation fix from  Aron Griffis <aron@hp.com> 040604
96 *
97 * New xmit() return, do_div and misc clean up by Stephen Hemminger
98 * <shemminger@osdl.org> 040923
99 *
100 * Randy Dunlap fixed u64 printk compiler warning
101 *
102 * Remove FCS from BW calculation.  Lennert Buytenhek <buytenh@wantstofly.org>
103 * New time handling. Lennert Buytenhek <buytenh@wantstofly.org> 041213
104 *
105 * Corrections from Nikolai Malykh (nmalykh@bilim.com)
106 * Removed unused flags F_SET_SRCMAC & F_SET_SRCIP 041230
107 *
108 * interruptible_sleep_on_timeout() replaced Nishanth Aravamudan <nacc@us.ibm.com>
109 * 050103
110 *
111 * MPLS support by Steven Whitehouse <steve@chygwyn.com>
112 *
113 * 802.1Q/Q-in-Q support by Francesco Fondelli (FF) <francesco.fondelli@gmail.com>
114 *
115 * Fixed src_mac command to set source mac of packet to value specified in
116 * command by Adit Ranadive <adit.262@gmail.com>
117 *
118 */
119
120#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
121
122#include <linux/sys.h>
123#include <linux/types.h>
124#include <linux/module.h>
125#include <linux/moduleparam.h>
126#include <linux/kernel.h>
127#include <linux/mutex.h>
128#include <linux/sched.h>
129#include <linux/slab.h>
130#include <linux/vmalloc.h>
131#include <linux/unistd.h>
132#include <linux/string.h>
133#include <linux/ptrace.h>
134#include <linux/errno.h>
135#include <linux/ioport.h>
136#include <linux/interrupt.h>
137#include <linux/capability.h>
138#include <linux/hrtimer.h>
139#include <linux/freezer.h>
140#include <linux/delay.h>
141#include <linux/timer.h>
142#include <linux/list.h>
143#include <linux/init.h>
144#include <linux/skbuff.h>
145#include <linux/netdevice.h>
146#include <linux/inet.h>
147#include <linux/inetdevice.h>
148#include <linux/rtnetlink.h>
149#include <linux/if_arp.h>
150#include <linux/if_vlan.h>
151#include <linux/in.h>
152#include <linux/ip.h>
153#include <linux/ipv6.h>
154#include <linux/udp.h>
155#include <linux/proc_fs.h>
156#include <linux/seq_file.h>
157#include <linux/wait.h>
158#include <linux/etherdevice.h>
159#include <linux/kthread.h>
160#include <linux/prefetch.h>
161#include <net/net_namespace.h>
162#include <net/checksum.h>
163#include <net/ipv6.h>
164#include <net/udp.h>
165#include <net/ip6_checksum.h>
166#include <net/addrconf.h>
167#ifdef CONFIG_XFRM
168#include <net/xfrm.h>
169#endif
170#include <net/netns/generic.h>
171#include <asm/byteorder.h>
172#include <linux/rcupdate.h>
173#include <linux/bitops.h>
174#include <linux/io.h>
175#include <linux/timex.h>
176#include <linux/uaccess.h>
177#include <asm/dma.h>
178#include <asm/div64.h>		/* do_div */
179
180#define VERSION	"2.74"
181#define IP_NAME_SZ 32
182#define MAX_MPLS_LABELS 16 /* This is the max label stack depth */
183#define MPLS_STACK_BOTTOM htonl(0x00000100)
184
185#define func_enter() pr_debug("entering %s\n", __func__);
186
187/* Device flag bits */
188#define F_IPSRC_RND   (1<<0)	/* IP-Src Random  */
189#define F_IPDST_RND   (1<<1)	/* IP-Dst Random  */
190#define F_UDPSRC_RND  (1<<2)	/* UDP-Src Random */
191#define F_UDPDST_RND  (1<<3)	/* UDP-Dst Random */
192#define F_MACSRC_RND  (1<<4)	/* MAC-Src Random */
193#define F_MACDST_RND  (1<<5)	/* MAC-Dst Random */
194#define F_TXSIZE_RND  (1<<6)	/* Transmit size is random */
195#define F_IPV6        (1<<7)	/* Interface in IPV6 Mode */
196#define F_MPLS_RND    (1<<8)	/* Random MPLS labels */
197#define F_VID_RND     (1<<9)	/* Random VLAN ID */
198#define F_SVID_RND    (1<<10)	/* Random SVLAN ID */
199#define F_FLOW_SEQ    (1<<11)	/* Sequential flows */
200#define F_IPSEC_ON    (1<<12)	/* ipsec on for flows */
201#define F_QUEUE_MAP_RND (1<<13)	/* queue map Random */
202#define F_QUEUE_MAP_CPU (1<<14)	/* queue map mirrors smp_processor_id() */
203#define F_NODE          (1<<15)	/* Node memory alloc*/
204#define F_UDPCSUM       (1<<16)	/* Include UDP checksum */
205#define F_NO_TIMESTAMP  (1<<17)	/* Don't timestamp packets (default TS) */
206
207/* Thread control flag bits */
208#define T_STOP        (1<<0)	/* Stop run */
209#define T_RUN         (1<<1)	/* Start run */
210#define T_REMDEVALL   (1<<2)	/* Remove all devs */
211#define T_REMDEV      (1<<3)	/* Remove one dev */
212
213/* If lock -- protects updating of if_list */
214#define   if_lock(t)           spin_lock(&(t->if_lock));
215#define   if_unlock(t)           spin_unlock(&(t->if_lock));
216
217/* Used to help with determining the pkts on receive */
218#define PKTGEN_MAGIC 0xbe9be955
219#define PG_PROC_DIR "pktgen"
220#define PGCTRL	    "pgctrl"
221
222#define MAX_CFLOWS  65536
223
224#define VLAN_TAG_SIZE(x) ((x)->vlan_id == 0xffff ? 0 : 4)
225#define SVLAN_TAG_SIZE(x) ((x)->svlan_id == 0xffff ? 0 : 4)
226
227struct flow_state {
228	__be32 cur_daddr;
229	int count;
230#ifdef CONFIG_XFRM
231	struct xfrm_state *x;
232#endif
233	__u32 flags;
234};
235
236/* flow flag bits */
237#define F_INIT   (1<<0)		/* flow has been initialized */
238
239struct pktgen_dev {
240	/*
241	 * Try to keep frequent/infrequent used vars. separated.
242	 */
243	struct proc_dir_entry *entry;	/* proc file */
244	struct pktgen_thread *pg_thread;/* the owner */
245	struct list_head list;		/* chaining in the thread's run-queue */
246	struct rcu_head	 rcu;		/* freed by RCU */
247
248	int running;		/* if false, the test will stop */
249
250	/* If min != max, then we will either do a linear iteration, or
251	 * we will do a random selection from within the range.
252	 */
253	__u32 flags;
254	int removal_mark;	/* non-zero => the device is marked for
255				 * removal by worker thread */
256
257	int min_pkt_size;
258	int max_pkt_size;
259	int pkt_overhead;	/* overhead for MPLS, VLANs, IPSEC etc */
260	int nfrags;
261	struct page *page;
262	u64 delay;		/* nano-seconds */
263
264	__u64 count;		/* Default No packets to send */
265	__u64 sofar;		/* How many pkts we've sent so far */
266	__u64 tx_bytes;		/* How many bytes we've transmitted */
267	__u64 errors;		/* Errors when trying to transmit, */
268
269	/* runtime counters relating to clone_skb */
270
271	__u64 allocated_skbs;
272	__u32 clone_count;
273	int last_ok;		/* Was last skb sent?
274				 * Or a failed transmit of some sort?
275				 * This will keep sequence numbers in order
276				 */
277	ktime_t next_tx;
278	ktime_t started_at;
279	ktime_t stopped_at;
280	u64	idle_acc;	/* nano-seconds */
281
282	__u32 seq_num;
283
284	int clone_skb;		/*
285				 * Use multiple SKBs during packet gen.
286				 * If this number is greater than 1, then
287				 * that many copies of the same packet will be
288				 * sent before a new packet is allocated.
289				 * If you want to send 1024 identical packets
290				 * before creating a new packet,
291				 * set clone_skb to 1024.
292				 */
293
294	char dst_min[IP_NAME_SZ];	/* IP, ie 1.2.3.4 */
295	char dst_max[IP_NAME_SZ];	/* IP, ie 1.2.3.4 */
296	char src_min[IP_NAME_SZ];	/* IP, ie 1.2.3.4 */
297	char src_max[IP_NAME_SZ];	/* IP, ie 1.2.3.4 */
298
299	struct in6_addr in6_saddr;
300	struct in6_addr in6_daddr;
301	struct in6_addr cur_in6_daddr;
302	struct in6_addr cur_in6_saddr;
303	/* For ranges */
304	struct in6_addr min_in6_daddr;
305	struct in6_addr max_in6_daddr;
306	struct in6_addr min_in6_saddr;
307	struct in6_addr max_in6_saddr;
308
309	/* If we're doing ranges, random or incremental, then this
310	 * defines the min/max for those ranges.
311	 */
312	__be32 saddr_min;	/* inclusive, source IP address */
313	__be32 saddr_max;	/* exclusive, source IP address */
314	__be32 daddr_min;	/* inclusive, dest IP address */
315	__be32 daddr_max;	/* exclusive, dest IP address */
316
317	__u16 udp_src_min;	/* inclusive, source UDP port */
318	__u16 udp_src_max;	/* exclusive, source UDP port */
319	__u16 udp_dst_min;	/* inclusive, dest UDP port */
320	__u16 udp_dst_max;	/* exclusive, dest UDP port */
321
322	/* DSCP + ECN */
323	__u8 tos;            /* six MSB of (former) IPv4 TOS
324				are for dscp codepoint */
325	__u8 traffic_class;  /* ditto for the (former) Traffic Class in IPv6
326				(see RFC 3260, sec. 4) */
327
328	/* MPLS */
329	unsigned int nr_labels;	/* Depth of stack, 0 = no MPLS */
330	__be32 labels[MAX_MPLS_LABELS];
331
332	/* VLAN/SVLAN (802.1Q/Q-in-Q) */
333	__u8  vlan_p;
334	__u8  vlan_cfi;
335	__u16 vlan_id;  /* 0xffff means no vlan tag */
336
337	__u8  svlan_p;
338	__u8  svlan_cfi;
339	__u16 svlan_id; /* 0xffff means no svlan tag */
340
341	__u32 src_mac_count;	/* How many MACs to iterate through */
342	__u32 dst_mac_count;	/* How many MACs to iterate through */
343
344	unsigned char dst_mac[ETH_ALEN];
345	unsigned char src_mac[ETH_ALEN];
346
347	__u32 cur_dst_mac_offset;
348	__u32 cur_src_mac_offset;
349	__be32 cur_saddr;
350	__be32 cur_daddr;
351	__u16 ip_id;
352	__u16 cur_udp_dst;
353	__u16 cur_udp_src;
354	__u16 cur_queue_map;
355	__u32 cur_pkt_size;
356	__u32 last_pkt_size;
357
358	__u8 hh[14];
359	/* = {
360	   0x00, 0x80, 0xC8, 0x79, 0xB3, 0xCB,
361
362	   We fill in SRC address later
363	   0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
364	   0x08, 0x00
365	   };
366	 */
367	__u16 pad;		/* pad out the hh struct to an even 16 bytes */
368
369	struct sk_buff *skb;	/* skb we are to transmit next, used for when we
370				 * are transmitting the same one multiple times
371				 */
372	struct net_device *odev; /* The out-going device.
373				  * Note that the device should have it's
374				  * pg_info pointer pointing back to this
375				  * device.
376				  * Set when the user specifies the out-going
377				  * device name (not when the inject is
378				  * started as it used to do.)
379				  */
380	char odevname[32];
381	struct flow_state *flows;
382	unsigned int cflows;	/* Concurrent flows (config) */
383	unsigned int lflow;		/* Flow length  (config) */
384	unsigned int nflows;	/* accumulated flows (stats) */
385	unsigned int curfl;		/* current sequenced flow (state)*/
386
387	u16 queue_map_min;
388	u16 queue_map_max;
389	__u32 skb_priority;	/* skb priority field */
390	unsigned int burst;	/* number of duplicated packets to burst */
391	int node;               /* Memory node */
392
393#ifdef CONFIG_XFRM
394	__u8	ipsmode;		/* IPSEC mode (config) */
395	__u8	ipsproto;		/* IPSEC type (config) */
396	__u32	spi;
397	struct dst_entry dst;
398	struct dst_ops dstops;
399#endif
400	char result[512];
401};
402
403struct pktgen_hdr {
404	__be32 pgh_magic;
405	__be32 seq_num;
406	__be32 tv_sec;
407	__be32 tv_usec;
408};
409
410
411static int pg_net_id __read_mostly;
412
413struct pktgen_net {
414	struct net		*net;
415	struct proc_dir_entry	*proc_dir;
416	struct list_head	pktgen_threads;
417	bool			pktgen_exiting;
418};
419
420struct pktgen_thread {
421	spinlock_t if_lock;		/* for list of devices */
422	struct list_head if_list;	/* All device here */
423	struct list_head th_list;
424	struct task_struct *tsk;
425	char result[512];
426
427	/* Field for thread to receive "posted" events terminate,
428	   stop ifs etc. */
429
430	u32 control;
431	int cpu;
432
433	wait_queue_head_t queue;
434	struct completion start_done;
435	struct pktgen_net *net;
436};
437
438#define REMOVE 1
439#define FIND   0
440
441static const char version[] =
442	"Packet Generator for packet performance testing. "
443	"Version: " VERSION "\n";
444
445static int pktgen_remove_device(struct pktgen_thread *t, struct pktgen_dev *i);
446static int pktgen_add_device(struct pktgen_thread *t, const char *ifname);
447static struct pktgen_dev *pktgen_find_dev(struct pktgen_thread *t,
448					  const char *ifname, bool exact);
449static int pktgen_device_event(struct notifier_block *, unsigned long, void *);
450static void pktgen_run_all_threads(struct pktgen_net *pn);
451static void pktgen_reset_all_threads(struct pktgen_net *pn);
452static void pktgen_stop_all_threads_ifs(struct pktgen_net *pn);
453
454static void pktgen_stop(struct pktgen_thread *t);
455static void pktgen_clear_counters(struct pktgen_dev *pkt_dev);
456
457/* Module parameters, defaults. */
458static int pg_count_d __read_mostly = 1000;
459static int pg_delay_d __read_mostly;
460static int pg_clone_skb_d  __read_mostly;
461static int debug  __read_mostly;
462
463static DEFINE_MUTEX(pktgen_thread_lock);
464
465static struct notifier_block pktgen_notifier_block = {
466	.notifier_call = pktgen_device_event,
467};
468
469/*
470 * /proc handling functions
471 *
472 */
473
474static int pgctrl_show(struct seq_file *seq, void *v)
475{
476	seq_puts(seq, version);
477	return 0;
478}
479
480static ssize_t pgctrl_write(struct file *file, const char __user *buf,
481			    size_t count, loff_t *ppos)
482{
483	char data[128];
484	struct pktgen_net *pn = net_generic(current->nsproxy->net_ns, pg_net_id);
485
486	if (!capable(CAP_NET_ADMIN))
487		return -EPERM;
488
489	if (count == 0)
490		return -EINVAL;
491
492	if (count > sizeof(data))
493		count = sizeof(data);
494
495	if (copy_from_user(data, buf, count))
496		return -EFAULT;
497
498	data[count - 1] = 0;	/* Strip trailing '\n' and terminate string */
499
500	if (!strcmp(data, "stop"))
501		pktgen_stop_all_threads_ifs(pn);
502
503	else if (!strcmp(data, "start"))
504		pktgen_run_all_threads(pn);
505
506	else if (!strcmp(data, "reset"))
507		pktgen_reset_all_threads(pn);
508
509	else
510		pr_warn("Unknown command: %s\n", data);
511
512	return count;
513}
514
515static int pgctrl_open(struct inode *inode, struct file *file)
516{
517	return single_open(file, pgctrl_show, PDE_DATA(inode));
518}
519
520static const struct file_operations pktgen_fops = {
521	.owner   = THIS_MODULE,
522	.open    = pgctrl_open,
523	.read    = seq_read,
524	.llseek  = seq_lseek,
525	.write   = pgctrl_write,
526	.release = single_release,
527};
528
529static int pktgen_if_show(struct seq_file *seq, void *v)
530{
531	const struct pktgen_dev *pkt_dev = seq->private;
532	ktime_t stopped;
533	u64 idle;
534
535	seq_printf(seq,
536		   "Params: count %llu  min_pkt_size: %u  max_pkt_size: %u\n",
537		   (unsigned long long)pkt_dev->count, pkt_dev->min_pkt_size,
538		   pkt_dev->max_pkt_size);
539
540	seq_printf(seq,
541		   "     frags: %d  delay: %llu  clone_skb: %d  ifname: %s\n",
542		   pkt_dev->nfrags, (unsigned long long) pkt_dev->delay,
543		   pkt_dev->clone_skb, pkt_dev->odevname);
544
545	seq_printf(seq, "     flows: %u flowlen: %u\n", pkt_dev->cflows,
546		   pkt_dev->lflow);
547
548	seq_printf(seq,
549		   "     queue_map_min: %u  queue_map_max: %u\n",
550		   pkt_dev->queue_map_min,
551		   pkt_dev->queue_map_max);
552
553	if (pkt_dev->skb_priority)
554		seq_printf(seq, "     skb_priority: %u\n",
555			   pkt_dev->skb_priority);
556
557	if (pkt_dev->flags & F_IPV6) {
558		seq_printf(seq,
559			   "     saddr: %pI6c  min_saddr: %pI6c  max_saddr: %pI6c\n"
560			   "     daddr: %pI6c  min_daddr: %pI6c  max_daddr: %pI6c\n",
561			   &pkt_dev->in6_saddr,
562			   &pkt_dev->min_in6_saddr, &pkt_dev->max_in6_saddr,
563			   &pkt_dev->in6_daddr,
564			   &pkt_dev->min_in6_daddr, &pkt_dev->max_in6_daddr);
565	} else {
566		seq_printf(seq,
567			   "     dst_min: %s  dst_max: %s\n",
568			   pkt_dev->dst_min, pkt_dev->dst_max);
569		seq_printf(seq,
570			   "        src_min: %s  src_max: %s\n",
571			   pkt_dev->src_min, pkt_dev->src_max);
572	}
573
574	seq_puts(seq, "     src_mac: ");
575
576	seq_printf(seq, "%pM ",
577		   is_zero_ether_addr(pkt_dev->src_mac) ?
578			     pkt_dev->odev->dev_addr : pkt_dev->src_mac);
579
580	seq_puts(seq, "dst_mac: ");
581	seq_printf(seq, "%pM\n", pkt_dev->dst_mac);
582
583	seq_printf(seq,
584		   "     udp_src_min: %d  udp_src_max: %d"
585		   "  udp_dst_min: %d  udp_dst_max: %d\n",
586		   pkt_dev->udp_src_min, pkt_dev->udp_src_max,
587		   pkt_dev->udp_dst_min, pkt_dev->udp_dst_max);
588
589	seq_printf(seq,
590		   "     src_mac_count: %d  dst_mac_count: %d\n",
591		   pkt_dev->src_mac_count, pkt_dev->dst_mac_count);
592
593	if (pkt_dev->nr_labels) {
594		unsigned int i;
595		seq_puts(seq, "     mpls: ");
596		for (i = 0; i < pkt_dev->nr_labels; i++)
597			seq_printf(seq, "%08x%s", ntohl(pkt_dev->labels[i]),
598				   i == pkt_dev->nr_labels-1 ? "\n" : ", ");
599	}
600
601	if (pkt_dev->vlan_id != 0xffff)
602		seq_printf(seq, "     vlan_id: %u  vlan_p: %u  vlan_cfi: %u\n",
603			   pkt_dev->vlan_id, pkt_dev->vlan_p,
604			   pkt_dev->vlan_cfi);
605
606	if (pkt_dev->svlan_id != 0xffff)
607		seq_printf(seq, "     svlan_id: %u  vlan_p: %u  vlan_cfi: %u\n",
608			   pkt_dev->svlan_id, pkt_dev->svlan_p,
609			   pkt_dev->svlan_cfi);
610
611	if (pkt_dev->tos)
612		seq_printf(seq, "     tos: 0x%02x\n", pkt_dev->tos);
613
614	if (pkt_dev->traffic_class)
615		seq_printf(seq, "     traffic_class: 0x%02x\n", pkt_dev->traffic_class);
616
617	if (pkt_dev->burst > 1)
618		seq_printf(seq, "     burst: %d\n", pkt_dev->burst);
619
620	if (pkt_dev->node >= 0)
621		seq_printf(seq, "     node: %d\n", pkt_dev->node);
622
623	seq_puts(seq, "     Flags: ");
624
625	if (pkt_dev->flags & F_IPV6)
626		seq_puts(seq, "IPV6  ");
627
628	if (pkt_dev->flags & F_IPSRC_RND)
629		seq_puts(seq, "IPSRC_RND  ");
630
631	if (pkt_dev->flags & F_IPDST_RND)
632		seq_puts(seq, "IPDST_RND  ");
633
634	if (pkt_dev->flags & F_TXSIZE_RND)
635		seq_puts(seq, "TXSIZE_RND  ");
636
637	if (pkt_dev->flags & F_UDPSRC_RND)
638		seq_puts(seq, "UDPSRC_RND  ");
639
640	if (pkt_dev->flags & F_UDPDST_RND)
641		seq_puts(seq, "UDPDST_RND  ");
642
643	if (pkt_dev->flags & F_UDPCSUM)
644		seq_puts(seq, "UDPCSUM  ");
645
646	if (pkt_dev->flags & F_NO_TIMESTAMP)
647		seq_puts(seq, "NO_TIMESTAMP  ");
648
649	if (pkt_dev->flags & F_MPLS_RND)
650		seq_puts(seq,  "MPLS_RND  ");
651
652	if (pkt_dev->flags & F_QUEUE_MAP_RND)
653		seq_puts(seq,  "QUEUE_MAP_RND  ");
654
655	if (pkt_dev->flags & F_QUEUE_MAP_CPU)
656		seq_puts(seq,  "QUEUE_MAP_CPU  ");
657
658	if (pkt_dev->cflows) {
659		if (pkt_dev->flags & F_FLOW_SEQ)
660			seq_puts(seq,  "FLOW_SEQ  "); /*in sequence flows*/
661		else
662			seq_puts(seq,  "FLOW_RND  ");
663	}
664
665#ifdef CONFIG_XFRM
666	if (pkt_dev->flags & F_IPSEC_ON) {
667		seq_puts(seq,  "IPSEC  ");
668		if (pkt_dev->spi)
669			seq_printf(seq, "spi:%u", pkt_dev->spi);
670	}
671#endif
672
673	if (pkt_dev->flags & F_MACSRC_RND)
674		seq_puts(seq, "MACSRC_RND  ");
675
676	if (pkt_dev->flags & F_MACDST_RND)
677		seq_puts(seq, "MACDST_RND  ");
678
679	if (pkt_dev->flags & F_VID_RND)
680		seq_puts(seq, "VID_RND  ");
681
682	if (pkt_dev->flags & F_SVID_RND)
683		seq_puts(seq, "SVID_RND  ");
684
685	if (pkt_dev->flags & F_NODE)
686		seq_puts(seq, "NODE_ALLOC  ");
687
688	seq_puts(seq, "\n");
689
690	/* not really stopped, more like last-running-at */
691	stopped = pkt_dev->running ? ktime_get() : pkt_dev->stopped_at;
692	idle = pkt_dev->idle_acc;
693	do_div(idle, NSEC_PER_USEC);
694
695	seq_printf(seq,
696		   "Current:\n     pkts-sofar: %llu  errors: %llu\n",
697		   (unsigned long long)pkt_dev->sofar,
698		   (unsigned long long)pkt_dev->errors);
699
700	seq_printf(seq,
701		   "     started: %lluus  stopped: %lluus idle: %lluus\n",
702		   (unsigned long long) ktime_to_us(pkt_dev->started_at),
703		   (unsigned long long) ktime_to_us(stopped),
704		   (unsigned long long) idle);
705
706	seq_printf(seq,
707		   "     seq_num: %d  cur_dst_mac_offset: %d  cur_src_mac_offset: %d\n",
708		   pkt_dev->seq_num, pkt_dev->cur_dst_mac_offset,
709		   pkt_dev->cur_src_mac_offset);
710
711	if (pkt_dev->flags & F_IPV6) {
712		seq_printf(seq, "     cur_saddr: %pI6c  cur_daddr: %pI6c\n",
713				&pkt_dev->cur_in6_saddr,
714				&pkt_dev->cur_in6_daddr);
715	} else
716		seq_printf(seq, "     cur_saddr: %pI4  cur_daddr: %pI4\n",
717			   &pkt_dev->cur_saddr, &pkt_dev->cur_daddr);
718
719	seq_printf(seq, "     cur_udp_dst: %d  cur_udp_src: %d\n",
720		   pkt_dev->cur_udp_dst, pkt_dev->cur_udp_src);
721
722	seq_printf(seq, "     cur_queue_map: %u\n", pkt_dev->cur_queue_map);
723
724	seq_printf(seq, "     flows: %u\n", pkt_dev->nflows);
725
726	if (pkt_dev->result[0])
727		seq_printf(seq, "Result: %s\n", pkt_dev->result);
728	else
729		seq_puts(seq, "Result: Idle\n");
730
731	return 0;
732}
733
734
735static int hex32_arg(const char __user *user_buffer, unsigned long maxlen,
736		     __u32 *num)
737{
738	int i = 0;
739	*num = 0;
740
741	for (; i < maxlen; i++) {
742		int value;
743		char c;
744		*num <<= 4;
745		if (get_user(c, &user_buffer[i]))
746			return -EFAULT;
747		value = hex_to_bin(c);
748		if (value >= 0)
749			*num |= value;
750		else
751			break;
752	}
753	return i;
754}
755
756static int count_trail_chars(const char __user * user_buffer,
757			     unsigned int maxlen)
758{
759	int i;
760
761	for (i = 0; i < maxlen; i++) {
762		char c;
763		if (get_user(c, &user_buffer[i]))
764			return -EFAULT;
765		switch (c) {
766		case '\"':
767		case '\n':
768		case '\r':
769		case '\t':
770		case ' ':
771		case '=':
772			break;
773		default:
774			goto done;
775		}
776	}
777done:
778	return i;
779}
780
781static long num_arg(const char __user *user_buffer, unsigned long maxlen,
782				unsigned long *num)
783{
784	int i;
785	*num = 0;
786
787	for (i = 0; i < maxlen; i++) {
788		char c;
789		if (get_user(c, &user_buffer[i]))
790			return -EFAULT;
791		if ((c >= '0') && (c <= '9')) {
792			*num *= 10;
793			*num += c - '0';
794		} else
795			break;
796	}
797	return i;
798}
799
800static int strn_len(const char __user * user_buffer, unsigned int maxlen)
801{
802	int i;
803
804	for (i = 0; i < maxlen; i++) {
805		char c;
806		if (get_user(c, &user_buffer[i]))
807			return -EFAULT;
808		switch (c) {
809		case '\"':
810		case '\n':
811		case '\r':
812		case '\t':
813		case ' ':
814			goto done_str;
815		default:
816			break;
817		}
818	}
819done_str:
820	return i;
821}
822
823static ssize_t get_labels(const char __user *buffer, struct pktgen_dev *pkt_dev)
824{
825	unsigned int n = 0;
826	char c;
827	ssize_t i = 0;
828	int len;
829
830	pkt_dev->nr_labels = 0;
831	do {
832		__u32 tmp;
833		len = hex32_arg(&buffer[i], 8, &tmp);
834		if (len <= 0)
835			return len;
836		pkt_dev->labels[n] = htonl(tmp);
837		if (pkt_dev->labels[n] & MPLS_STACK_BOTTOM)
838			pkt_dev->flags |= F_MPLS_RND;
839		i += len;
840		if (get_user(c, &buffer[i]))
841			return -EFAULT;
842		i++;
843		n++;
844		if (n >= MAX_MPLS_LABELS)
845			return -E2BIG;
846	} while (c == ',');
847
848	pkt_dev->nr_labels = n;
849	return i;
850}
851
852static ssize_t pktgen_if_write(struct file *file,
853			       const char __user * user_buffer, size_t count,
854			       loff_t * offset)
855{
856	struct seq_file *seq = file->private_data;
857	struct pktgen_dev *pkt_dev = seq->private;
858	int i, max, len;
859	char name[16], valstr[32];
860	unsigned long value = 0;
861	char *pg_result = NULL;
862	int tmp = 0;
863	char buf[128];
864
865	pg_result = &(pkt_dev->result[0]);
866
867	if (count < 1) {
868		pr_warn("wrong command format\n");
869		return -EINVAL;
870	}
871
872	max = count;
873	tmp = count_trail_chars(user_buffer, max);
874	if (tmp < 0) {
875		pr_warn("illegal format\n");
876		return tmp;
877	}
878	i = tmp;
879
880	/* Read variable name */
881
882	len = strn_len(&user_buffer[i], sizeof(name) - 1);
883	if (len < 0)
884		return len;
885
886	memset(name, 0, sizeof(name));
887	if (copy_from_user(name, &user_buffer[i], len))
888		return -EFAULT;
889	i += len;
890
891	max = count - i;
892	len = count_trail_chars(&user_buffer[i], max);
893	if (len < 0)
894		return len;
895
896	i += len;
897
898	if (debug) {
899		size_t copy = min_t(size_t, count, 1023);
900		char tb[copy + 1];
901		if (copy_from_user(tb, user_buffer, copy))
902			return -EFAULT;
903		tb[copy] = 0;
904		pr_debug("%s,%lu  buffer -:%s:-\n",
905			 name, (unsigned long)count, tb);
906	}
907
908	if (!strcmp(name, "min_pkt_size")) {
909		len = num_arg(&user_buffer[i], 10, &value);
910		if (len < 0)
911			return len;
912
913		i += len;
914		if (value < 14 + 20 + 8)
915			value = 14 + 20 + 8;
916		if (value != pkt_dev->min_pkt_size) {
917			pkt_dev->min_pkt_size = value;
918			pkt_dev->cur_pkt_size = value;
919		}
920		sprintf(pg_result, "OK: min_pkt_size=%u",
921			pkt_dev->min_pkt_size);
922		return count;
923	}
924
925	if (!strcmp(name, "max_pkt_size")) {
926		len = num_arg(&user_buffer[i], 10, &value);
927		if (len < 0)
928			return len;
929
930		i += len;
931		if (value < 14 + 20 + 8)
932			value = 14 + 20 + 8;
933		if (value != pkt_dev->max_pkt_size) {
934			pkt_dev->max_pkt_size = value;
935			pkt_dev->cur_pkt_size = value;
936		}
937		sprintf(pg_result, "OK: max_pkt_size=%u",
938			pkt_dev->max_pkt_size);
939		return count;
940	}
941
942	/* Shortcut for min = max */
943
944	if (!strcmp(name, "pkt_size")) {
945		len = num_arg(&user_buffer[i], 10, &value);
946		if (len < 0)
947			return len;
948
949		i += len;
950		if (value < 14 + 20 + 8)
951			value = 14 + 20 + 8;
952		if (value != pkt_dev->min_pkt_size) {
953			pkt_dev->min_pkt_size = value;
954			pkt_dev->max_pkt_size = value;
955			pkt_dev->cur_pkt_size = value;
956		}
957		sprintf(pg_result, "OK: pkt_size=%u", pkt_dev->min_pkt_size);
958		return count;
959	}
960
961	if (!strcmp(name, "debug")) {
962		len = num_arg(&user_buffer[i], 10, &value);
963		if (len < 0)
964			return len;
965
966		i += len;
967		debug = value;
968		sprintf(pg_result, "OK: debug=%u", debug);
969		return count;
970	}
971
972	if (!strcmp(name, "frags")) {
973		len = num_arg(&user_buffer[i], 10, &value);
974		if (len < 0)
975			return len;
976
977		i += len;
978		pkt_dev->nfrags = value;
979		sprintf(pg_result, "OK: frags=%u", pkt_dev->nfrags);
980		return count;
981	}
982	if (!strcmp(name, "delay")) {
983		len = num_arg(&user_buffer[i], 10, &value);
984		if (len < 0)
985			return len;
986
987		i += len;
988		if (value == 0x7FFFFFFF)
989			pkt_dev->delay = ULLONG_MAX;
990		else
991			pkt_dev->delay = (u64)value;
992
993		sprintf(pg_result, "OK: delay=%llu",
994			(unsigned long long) pkt_dev->delay);
995		return count;
996	}
997	if (!strcmp(name, "rate")) {
998		len = num_arg(&user_buffer[i], 10, &value);
999		if (len < 0)
1000			return len;
1001
1002		i += len;
1003		if (!value)
1004			return len;
1005		pkt_dev->delay = pkt_dev->min_pkt_size*8*NSEC_PER_USEC/value;
1006		if (debug)
1007			pr_info("Delay set at: %llu ns\n", pkt_dev->delay);
1008
1009		sprintf(pg_result, "OK: rate=%lu", value);
1010		return count;
1011	}
1012	if (!strcmp(name, "ratep")) {
1013		len = num_arg(&user_buffer[i], 10, &value);
1014		if (len < 0)
1015			return len;
1016
1017		i += len;
1018		if (!value)
1019			return len;
1020		pkt_dev->delay = NSEC_PER_SEC/value;
1021		if (debug)
1022			pr_info("Delay set at: %llu ns\n", pkt_dev->delay);
1023
1024		sprintf(pg_result, "OK: rate=%lu", value);
1025		return count;
1026	}
1027	if (!strcmp(name, "udp_src_min")) {
1028		len = num_arg(&user_buffer[i], 10, &value);
1029		if (len < 0)
1030			return len;
1031
1032		i += len;
1033		if (value != pkt_dev->udp_src_min) {
1034			pkt_dev->udp_src_min = value;
1035			pkt_dev->cur_udp_src = value;
1036		}
1037		sprintf(pg_result, "OK: udp_src_min=%u", pkt_dev->udp_src_min);
1038		return count;
1039	}
1040	if (!strcmp(name, "udp_dst_min")) {
1041		len = num_arg(&user_buffer[i], 10, &value);
1042		if (len < 0)
1043			return len;
1044
1045		i += len;
1046		if (value != pkt_dev->udp_dst_min) {
1047			pkt_dev->udp_dst_min = value;
1048			pkt_dev->cur_udp_dst = value;
1049		}
1050		sprintf(pg_result, "OK: udp_dst_min=%u", pkt_dev->udp_dst_min);
1051		return count;
1052	}
1053	if (!strcmp(name, "udp_src_max")) {
1054		len = num_arg(&user_buffer[i], 10, &value);
1055		if (len < 0)
1056			return len;
1057
1058		i += len;
1059		if (value != pkt_dev->udp_src_max) {
1060			pkt_dev->udp_src_max = value;
1061			pkt_dev->cur_udp_src = value;
1062		}
1063		sprintf(pg_result, "OK: udp_src_max=%u", pkt_dev->udp_src_max);
1064		return count;
1065	}
1066	if (!strcmp(name, "udp_dst_max")) {
1067		len = num_arg(&user_buffer[i], 10, &value);
1068		if (len < 0)
1069			return len;
1070
1071		i += len;
1072		if (value != pkt_dev->udp_dst_max) {
1073			pkt_dev->udp_dst_max = value;
1074			pkt_dev->cur_udp_dst = value;
1075		}
1076		sprintf(pg_result, "OK: udp_dst_max=%u", pkt_dev->udp_dst_max);
1077		return count;
1078	}
1079	if (!strcmp(name, "clone_skb")) {
1080		len = num_arg(&user_buffer[i], 10, &value);
1081		if (len < 0)
1082			return len;
1083		if ((value > 0) &&
1084		    (!(pkt_dev->odev->priv_flags & IFF_TX_SKB_SHARING)))
1085			return -ENOTSUPP;
1086		i += len;
1087		pkt_dev->clone_skb = value;
1088
1089		sprintf(pg_result, "OK: clone_skb=%d", pkt_dev->clone_skb);
1090		return count;
1091	}
1092	if (!strcmp(name, "count")) {
1093		len = num_arg(&user_buffer[i], 10, &value);
1094		if (len < 0)
1095			return len;
1096
1097		i += len;
1098		pkt_dev->count = value;
1099		sprintf(pg_result, "OK: count=%llu",
1100			(unsigned long long)pkt_dev->count);
1101		return count;
1102	}
1103	if (!strcmp(name, "src_mac_count")) {
1104		len = num_arg(&user_buffer[i], 10, &value);
1105		if (len < 0)
1106			return len;
1107
1108		i += len;
1109		if (pkt_dev->src_mac_count != value) {
1110			pkt_dev->src_mac_count = value;
1111			pkt_dev->cur_src_mac_offset = 0;
1112		}
1113		sprintf(pg_result, "OK: src_mac_count=%d",
1114			pkt_dev->src_mac_count);
1115		return count;
1116	}
1117	if (!strcmp(name, "dst_mac_count")) {
1118		len = num_arg(&user_buffer[i], 10, &value);
1119		if (len < 0)
1120			return len;
1121
1122		i += len;
1123		if (pkt_dev->dst_mac_count != value) {
1124			pkt_dev->dst_mac_count = value;
1125			pkt_dev->cur_dst_mac_offset = 0;
1126		}
1127		sprintf(pg_result, "OK: dst_mac_count=%d",
1128			pkt_dev->dst_mac_count);
1129		return count;
1130	}
1131	if (!strcmp(name, "burst")) {
1132		len = num_arg(&user_buffer[i], 10, &value);
1133		if (len < 0)
1134			return len;
1135
1136		i += len;
1137		if ((value > 1) &&
1138		    (!(pkt_dev->odev->priv_flags & IFF_TX_SKB_SHARING)))
1139			return -ENOTSUPP;
1140		pkt_dev->burst = value < 1 ? 1 : value;
1141		sprintf(pg_result, "OK: burst=%d", pkt_dev->burst);
1142		return count;
1143	}
1144	if (!strcmp(name, "node")) {
1145		len = num_arg(&user_buffer[i], 10, &value);
1146		if (len < 0)
1147			return len;
1148
1149		i += len;
1150
1151		if (node_possible(value)) {
1152			pkt_dev->node = value;
1153			sprintf(pg_result, "OK: node=%d", pkt_dev->node);
1154			if (pkt_dev->page) {
1155				put_page(pkt_dev->page);
1156				pkt_dev->page = NULL;
1157			}
1158		}
1159		else
1160			sprintf(pg_result, "ERROR: node not possible");
1161		return count;
1162	}
1163	if (!strcmp(name, "flag")) {
1164		char f[32];
1165		memset(f, 0, 32);
1166		len = strn_len(&user_buffer[i], sizeof(f) - 1);
1167		if (len < 0)
1168			return len;
1169
1170		if (copy_from_user(f, &user_buffer[i], len))
1171			return -EFAULT;
1172		i += len;
1173		if (strcmp(f, "IPSRC_RND") == 0)
1174			pkt_dev->flags |= F_IPSRC_RND;
1175
1176		else if (strcmp(f, "!IPSRC_RND") == 0)
1177			pkt_dev->flags &= ~F_IPSRC_RND;
1178
1179		else if (strcmp(f, "TXSIZE_RND") == 0)
1180			pkt_dev->flags |= F_TXSIZE_RND;
1181
1182		else if (strcmp(f, "!TXSIZE_RND") == 0)
1183			pkt_dev->flags &= ~F_TXSIZE_RND;
1184
1185		else if (strcmp(f, "IPDST_RND") == 0)
1186			pkt_dev->flags |= F_IPDST_RND;
1187
1188		else if (strcmp(f, "!IPDST_RND") == 0)
1189			pkt_dev->flags &= ~F_IPDST_RND;
1190
1191		else if (strcmp(f, "UDPSRC_RND") == 0)
1192			pkt_dev->flags |= F_UDPSRC_RND;
1193
1194		else if (strcmp(f, "!UDPSRC_RND") == 0)
1195			pkt_dev->flags &= ~F_UDPSRC_RND;
1196
1197		else if (strcmp(f, "UDPDST_RND") == 0)
1198			pkt_dev->flags |= F_UDPDST_RND;
1199
1200		else if (strcmp(f, "!UDPDST_RND") == 0)
1201			pkt_dev->flags &= ~F_UDPDST_RND;
1202
1203		else if (strcmp(f, "MACSRC_RND") == 0)
1204			pkt_dev->flags |= F_MACSRC_RND;
1205
1206		else if (strcmp(f, "!MACSRC_RND") == 0)
1207			pkt_dev->flags &= ~F_MACSRC_RND;
1208
1209		else if (strcmp(f, "MACDST_RND") == 0)
1210			pkt_dev->flags |= F_MACDST_RND;
1211
1212		else if (strcmp(f, "!MACDST_RND") == 0)
1213			pkt_dev->flags &= ~F_MACDST_RND;
1214
1215		else if (strcmp(f, "MPLS_RND") == 0)
1216			pkt_dev->flags |= F_MPLS_RND;
1217
1218		else if (strcmp(f, "!MPLS_RND") == 0)
1219			pkt_dev->flags &= ~F_MPLS_RND;
1220
1221		else if (strcmp(f, "VID_RND") == 0)
1222			pkt_dev->flags |= F_VID_RND;
1223
1224		else if (strcmp(f, "!VID_RND") == 0)
1225			pkt_dev->flags &= ~F_VID_RND;
1226
1227		else if (strcmp(f, "SVID_RND") == 0)
1228			pkt_dev->flags |= F_SVID_RND;
1229
1230		else if (strcmp(f, "!SVID_RND") == 0)
1231			pkt_dev->flags &= ~F_SVID_RND;
1232
1233		else if (strcmp(f, "FLOW_SEQ") == 0)
1234			pkt_dev->flags |= F_FLOW_SEQ;
1235
1236		else if (strcmp(f, "QUEUE_MAP_RND") == 0)
1237			pkt_dev->flags |= F_QUEUE_MAP_RND;
1238
1239		else if (strcmp(f, "!QUEUE_MAP_RND") == 0)
1240			pkt_dev->flags &= ~F_QUEUE_MAP_RND;
1241
1242		else if (strcmp(f, "QUEUE_MAP_CPU") == 0)
1243			pkt_dev->flags |= F_QUEUE_MAP_CPU;
1244
1245		else if (strcmp(f, "!QUEUE_MAP_CPU") == 0)
1246			pkt_dev->flags &= ~F_QUEUE_MAP_CPU;
1247#ifdef CONFIG_XFRM
1248		else if (strcmp(f, "IPSEC") == 0)
1249			pkt_dev->flags |= F_IPSEC_ON;
1250#endif
1251
1252		else if (strcmp(f, "!IPV6") == 0)
1253			pkt_dev->flags &= ~F_IPV6;
1254
1255		else if (strcmp(f, "NODE_ALLOC") == 0)
1256			pkt_dev->flags |= F_NODE;
1257
1258		else if (strcmp(f, "!NODE_ALLOC") == 0)
1259			pkt_dev->flags &= ~F_NODE;
1260
1261		else if (strcmp(f, "UDPCSUM") == 0)
1262			pkt_dev->flags |= F_UDPCSUM;
1263
1264		else if (strcmp(f, "!UDPCSUM") == 0)
1265			pkt_dev->flags &= ~F_UDPCSUM;
1266
1267		else if (strcmp(f, "NO_TIMESTAMP") == 0)
1268			pkt_dev->flags |= F_NO_TIMESTAMP;
1269
1270		else {
1271			sprintf(pg_result,
1272				"Flag -:%s:- unknown\nAvailable flags, (prepend ! to un-set flag):\n%s",
1273				f,
1274				"IPSRC_RND, IPDST_RND, UDPSRC_RND, UDPDST_RND, "
1275				"MACSRC_RND, MACDST_RND, TXSIZE_RND, IPV6, "
1276				"MPLS_RND, VID_RND, SVID_RND, FLOW_SEQ, "
1277				"QUEUE_MAP_RND, QUEUE_MAP_CPU, UDPCSUM, "
1278				"NO_TIMESTAMP, "
1279#ifdef CONFIG_XFRM
1280				"IPSEC, "
1281#endif
1282				"NODE_ALLOC\n");
1283			return count;
1284		}
1285		sprintf(pg_result, "OK: flags=0x%x", pkt_dev->flags);
1286		return count;
1287	}
1288	if (!strcmp(name, "dst_min") || !strcmp(name, "dst")) {
1289		len = strn_len(&user_buffer[i], sizeof(pkt_dev->dst_min) - 1);
1290		if (len < 0)
1291			return len;
1292
1293		if (copy_from_user(buf, &user_buffer[i], len))
1294			return -EFAULT;
1295		buf[len] = 0;
1296		if (strcmp(buf, pkt_dev->dst_min) != 0) {
1297			memset(pkt_dev->dst_min, 0, sizeof(pkt_dev->dst_min));
1298			strncpy(pkt_dev->dst_min, buf, len);
1299			pkt_dev->daddr_min = in_aton(pkt_dev->dst_min);
1300			pkt_dev->cur_daddr = pkt_dev->daddr_min;
1301		}
1302		if (debug)
1303			pr_debug("dst_min set to: %s\n", pkt_dev->dst_min);
1304		i += len;
1305		sprintf(pg_result, "OK: dst_min=%s", pkt_dev->dst_min);
1306		return count;
1307	}
1308	if (!strcmp(name, "dst_max")) {
1309		len = strn_len(&user_buffer[i], sizeof(pkt_dev->dst_max) - 1);
1310		if (len < 0)
1311			return len;
1312
1313
1314		if (copy_from_user(buf, &user_buffer[i], len))
1315			return -EFAULT;
1316
1317		buf[len] = 0;
1318		if (strcmp(buf, pkt_dev->dst_max) != 0) {
1319			memset(pkt_dev->dst_max, 0, sizeof(pkt_dev->dst_max));
1320			strncpy(pkt_dev->dst_max, buf, len);
1321			pkt_dev->daddr_max = in_aton(pkt_dev->dst_max);
1322			pkt_dev->cur_daddr = pkt_dev->daddr_max;
1323		}
1324		if (debug)
1325			pr_debug("dst_max set to: %s\n", pkt_dev->dst_max);
1326		i += len;
1327		sprintf(pg_result, "OK: dst_max=%s", pkt_dev->dst_max);
1328		return count;
1329	}
1330	if (!strcmp(name, "dst6")) {
1331		len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1332		if (len < 0)
1333			return len;
1334
1335		pkt_dev->flags |= F_IPV6;
1336
1337		if (copy_from_user(buf, &user_buffer[i], len))
1338			return -EFAULT;
1339		buf[len] = 0;
1340
1341		in6_pton(buf, -1, pkt_dev->in6_daddr.s6_addr, -1, NULL);
1342		snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->in6_daddr);
1343
1344		pkt_dev->cur_in6_daddr = pkt_dev->in6_daddr;
1345
1346		if (debug)
1347			pr_debug("dst6 set to: %s\n", buf);
1348
1349		i += len;
1350		sprintf(pg_result, "OK: dst6=%s", buf);
1351		return count;
1352	}
1353	if (!strcmp(name, "dst6_min")) {
1354		len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1355		if (len < 0)
1356			return len;
1357
1358		pkt_dev->flags |= F_IPV6;
1359
1360		if (copy_from_user(buf, &user_buffer[i], len))
1361			return -EFAULT;
1362		buf[len] = 0;
1363
1364		in6_pton(buf, -1, pkt_dev->min_in6_daddr.s6_addr, -1, NULL);
1365		snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->min_in6_daddr);
1366
1367		pkt_dev->cur_in6_daddr = pkt_dev->min_in6_daddr;
1368		if (debug)
1369			pr_debug("dst6_min set to: %s\n", buf);
1370
1371		i += len;
1372		sprintf(pg_result, "OK: dst6_min=%s", buf);
1373		return count;
1374	}
1375	if (!strcmp(name, "dst6_max")) {
1376		len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1377		if (len < 0)
1378			return len;
1379
1380		pkt_dev->flags |= F_IPV6;
1381
1382		if (copy_from_user(buf, &user_buffer[i], len))
1383			return -EFAULT;
1384		buf[len] = 0;
1385
1386		in6_pton(buf, -1, pkt_dev->max_in6_daddr.s6_addr, -1, NULL);
1387		snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->max_in6_daddr);
1388
1389		if (debug)
1390			pr_debug("dst6_max set to: %s\n", buf);
1391
1392		i += len;
1393		sprintf(pg_result, "OK: dst6_max=%s", buf);
1394		return count;
1395	}
1396	if (!strcmp(name, "src6")) {
1397		len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1398		if (len < 0)
1399			return len;
1400
1401		pkt_dev->flags |= F_IPV6;
1402
1403		if (copy_from_user(buf, &user_buffer[i], len))
1404			return -EFAULT;
1405		buf[len] = 0;
1406
1407		in6_pton(buf, -1, pkt_dev->in6_saddr.s6_addr, -1, NULL);
1408		snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->in6_saddr);
1409
1410		pkt_dev->cur_in6_saddr = pkt_dev->in6_saddr;
1411
1412		if (debug)
1413			pr_debug("src6 set to: %s\n", buf);
1414
1415		i += len;
1416		sprintf(pg_result, "OK: src6=%s", buf);
1417		return count;
1418	}
1419	if (!strcmp(name, "src_min")) {
1420		len = strn_len(&user_buffer[i], sizeof(pkt_dev->src_min) - 1);
1421		if (len < 0)
1422			return len;
1423
1424		if (copy_from_user(buf, &user_buffer[i], len))
1425			return -EFAULT;
1426		buf[len] = 0;
1427		if (strcmp(buf, pkt_dev->src_min) != 0) {
1428			memset(pkt_dev->src_min, 0, sizeof(pkt_dev->src_min));
1429			strncpy(pkt_dev->src_min, buf, len);
1430			pkt_dev->saddr_min = in_aton(pkt_dev->src_min);
1431			pkt_dev->cur_saddr = pkt_dev->saddr_min;
1432		}
1433		if (debug)
1434			pr_debug("src_min set to: %s\n", pkt_dev->src_min);
1435		i += len;
1436		sprintf(pg_result, "OK: src_min=%s", pkt_dev->src_min);
1437		return count;
1438	}
1439	if (!strcmp(name, "src_max")) {
1440		len = strn_len(&user_buffer[i], sizeof(pkt_dev->src_max) - 1);
1441		if (len < 0)
1442			return len;
1443
1444		if (copy_from_user(buf, &user_buffer[i], len))
1445			return -EFAULT;
1446		buf[len] = 0;
1447		if (strcmp(buf, pkt_dev->src_max) != 0) {
1448			memset(pkt_dev->src_max, 0, sizeof(pkt_dev->src_max));
1449			strncpy(pkt_dev->src_max, buf, len);
1450			pkt_dev->saddr_max = in_aton(pkt_dev->src_max);
1451			pkt_dev->cur_saddr = pkt_dev->saddr_max;
1452		}
1453		if (debug)
1454			pr_debug("src_max set to: %s\n", pkt_dev->src_max);
1455		i += len;
1456		sprintf(pg_result, "OK: src_max=%s", pkt_dev->src_max);
1457		return count;
1458	}
1459	if (!strcmp(name, "dst_mac")) {
1460		len = strn_len(&user_buffer[i], sizeof(valstr) - 1);
1461		if (len < 0)
1462			return len;
1463
1464		memset(valstr, 0, sizeof(valstr));
1465		if (copy_from_user(valstr, &user_buffer[i], len))
1466			return -EFAULT;
1467
1468		if (!mac_pton(valstr, pkt_dev->dst_mac))
1469			return -EINVAL;
1470		/* Set up Dest MAC */
1471		ether_addr_copy(&pkt_dev->hh[0], pkt_dev->dst_mac);
1472
1473		sprintf(pg_result, "OK: dstmac %pM", pkt_dev->dst_mac);
1474		return count;
1475	}
1476	if (!strcmp(name, "src_mac")) {
1477		len = strn_len(&user_buffer[i], sizeof(valstr) - 1);
1478		if (len < 0)
1479			return len;
1480
1481		memset(valstr, 0, sizeof(valstr));
1482		if (copy_from_user(valstr, &user_buffer[i], len))
1483			return -EFAULT;
1484
1485		if (!mac_pton(valstr, pkt_dev->src_mac))
1486			return -EINVAL;
1487		/* Set up Src MAC */
1488		ether_addr_copy(&pkt_dev->hh[6], pkt_dev->src_mac);
1489
1490		sprintf(pg_result, "OK: srcmac %pM", pkt_dev->src_mac);
1491		return count;
1492	}
1493
1494	if (!strcmp(name, "clear_counters")) {
1495		pktgen_clear_counters(pkt_dev);
1496		sprintf(pg_result, "OK: Clearing counters.\n");
1497		return count;
1498	}
1499
1500	if (!strcmp(name, "flows")) {
1501		len = num_arg(&user_buffer[i], 10, &value);
1502		if (len < 0)
1503			return len;
1504
1505		i += len;
1506		if (value > MAX_CFLOWS)
1507			value = MAX_CFLOWS;
1508
1509		pkt_dev->cflows = value;
1510		sprintf(pg_result, "OK: flows=%u", pkt_dev->cflows);
1511		return count;
1512	}
1513#ifdef CONFIG_XFRM
1514	if (!strcmp(name, "spi")) {
1515		len = num_arg(&user_buffer[i], 10, &value);
1516		if (len < 0)
1517			return len;
1518
1519		i += len;
1520		pkt_dev->spi = value;
1521		sprintf(pg_result, "OK: spi=%u", pkt_dev->spi);
1522		return count;
1523	}
1524#endif
1525	if (!strcmp(name, "flowlen")) {
1526		len = num_arg(&user_buffer[i], 10, &value);
1527		if (len < 0)
1528			return len;
1529
1530		i += len;
1531		pkt_dev->lflow = value;
1532		sprintf(pg_result, "OK: flowlen=%u", pkt_dev->lflow);
1533		return count;
1534	}
1535
1536	if (!strcmp(name, "queue_map_min")) {
1537		len = num_arg(&user_buffer[i], 5, &value);
1538		if (len < 0)
1539			return len;
1540
1541		i += len;
1542		pkt_dev->queue_map_min = value;
1543		sprintf(pg_result, "OK: queue_map_min=%u", pkt_dev->queue_map_min);
1544		return count;
1545	}
1546
1547	if (!strcmp(name, "queue_map_max")) {
1548		len = num_arg(&user_buffer[i], 5, &value);
1549		if (len < 0)
1550			return len;
1551
1552		i += len;
1553		pkt_dev->queue_map_max = value;
1554		sprintf(pg_result, "OK: queue_map_max=%u", pkt_dev->queue_map_max);
1555		return count;
1556	}
1557
1558	if (!strcmp(name, "mpls")) {
1559		unsigned int n, cnt;
1560
1561		len = get_labels(&user_buffer[i], pkt_dev);
1562		if (len < 0)
1563			return len;
1564		i += len;
1565		cnt = sprintf(pg_result, "OK: mpls=");
1566		for (n = 0; n < pkt_dev->nr_labels; n++)
1567			cnt += sprintf(pg_result + cnt,
1568				       "%08x%s", ntohl(pkt_dev->labels[n]),
1569				       n == pkt_dev->nr_labels-1 ? "" : ",");
1570
1571		if (pkt_dev->nr_labels && pkt_dev->vlan_id != 0xffff) {
1572			pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1573			pkt_dev->svlan_id = 0xffff;
1574
1575			if (debug)
1576				pr_debug("VLAN/SVLAN auto turned off\n");
1577		}
1578		return count;
1579	}
1580
1581	if (!strcmp(name, "vlan_id")) {
1582		len = num_arg(&user_buffer[i], 4, &value);
1583		if (len < 0)
1584			return len;
1585
1586		i += len;
1587		if (value <= 4095) {
1588			pkt_dev->vlan_id = value;  /* turn on VLAN */
1589
1590			if (debug)
1591				pr_debug("VLAN turned on\n");
1592
1593			if (debug && pkt_dev->nr_labels)
1594				pr_debug("MPLS auto turned off\n");
1595
1596			pkt_dev->nr_labels = 0;    /* turn off MPLS */
1597			sprintf(pg_result, "OK: vlan_id=%u", pkt_dev->vlan_id);
1598		} else {
1599			pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1600			pkt_dev->svlan_id = 0xffff;
1601
1602			if (debug)
1603				pr_debug("VLAN/SVLAN turned off\n");
1604		}
1605		return count;
1606	}
1607
1608	if (!strcmp(name, "vlan_p")) {
1609		len = num_arg(&user_buffer[i], 1, &value);
1610		if (len < 0)
1611			return len;
1612
1613		i += len;
1614		if ((value <= 7) && (pkt_dev->vlan_id != 0xffff)) {
1615			pkt_dev->vlan_p = value;
1616			sprintf(pg_result, "OK: vlan_p=%u", pkt_dev->vlan_p);
1617		} else {
1618			sprintf(pg_result, "ERROR: vlan_p must be 0-7");
1619		}
1620		return count;
1621	}
1622
1623	if (!strcmp(name, "vlan_cfi")) {
1624		len = num_arg(&user_buffer[i], 1, &value);
1625		if (len < 0)
1626			return len;
1627
1628		i += len;
1629		if ((value <= 1) && (pkt_dev->vlan_id != 0xffff)) {
1630			pkt_dev->vlan_cfi = value;
1631			sprintf(pg_result, "OK: vlan_cfi=%u", pkt_dev->vlan_cfi);
1632		} else {
1633			sprintf(pg_result, "ERROR: vlan_cfi must be 0-1");
1634		}
1635		return count;
1636	}
1637
1638	if (!strcmp(name, "svlan_id")) {
1639		len = num_arg(&user_buffer[i], 4, &value);
1640		if (len < 0)
1641			return len;
1642
1643		i += len;
1644		if ((value <= 4095) && ((pkt_dev->vlan_id != 0xffff))) {
1645			pkt_dev->svlan_id = value;  /* turn on SVLAN */
1646
1647			if (debug)
1648				pr_debug("SVLAN turned on\n");
1649
1650			if (debug && pkt_dev->nr_labels)
1651				pr_debug("MPLS auto turned off\n");
1652
1653			pkt_dev->nr_labels = 0;    /* turn off MPLS */
1654			sprintf(pg_result, "OK: svlan_id=%u", pkt_dev->svlan_id);
1655		} else {
1656			pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1657			pkt_dev->svlan_id = 0xffff;
1658
1659			if (debug)
1660				pr_debug("VLAN/SVLAN turned off\n");
1661		}
1662		return count;
1663	}
1664
1665	if (!strcmp(name, "svlan_p")) {
1666		len = num_arg(&user_buffer[i], 1, &value);
1667		if (len < 0)
1668			return len;
1669
1670		i += len;
1671		if ((value <= 7) && (pkt_dev->svlan_id != 0xffff)) {
1672			pkt_dev->svlan_p = value;
1673			sprintf(pg_result, "OK: svlan_p=%u", pkt_dev->svlan_p);
1674		} else {
1675			sprintf(pg_result, "ERROR: svlan_p must be 0-7");
1676		}
1677		return count;
1678	}
1679
1680	if (!strcmp(name, "svlan_cfi")) {
1681		len = num_arg(&user_buffer[i], 1, &value);
1682		if (len < 0)
1683			return len;
1684
1685		i += len;
1686		if ((value <= 1) && (pkt_dev->svlan_id != 0xffff)) {
1687			pkt_dev->svlan_cfi = value;
1688			sprintf(pg_result, "OK: svlan_cfi=%u", pkt_dev->svlan_cfi);
1689		} else {
1690			sprintf(pg_result, "ERROR: svlan_cfi must be 0-1");
1691		}
1692		return count;
1693	}
1694
1695	if (!strcmp(name, "tos")) {
1696		__u32 tmp_value = 0;
1697		len = hex32_arg(&user_buffer[i], 2, &tmp_value);
1698		if (len < 0)
1699			return len;
1700
1701		i += len;
1702		if (len == 2) {
1703			pkt_dev->tos = tmp_value;
1704			sprintf(pg_result, "OK: tos=0x%02x", pkt_dev->tos);
1705		} else {
1706			sprintf(pg_result, "ERROR: tos must be 00-ff");
1707		}
1708		return count;
1709	}
1710
1711	if (!strcmp(name, "traffic_class")) {
1712		__u32 tmp_value = 0;
1713		len = hex32_arg(&user_buffer[i], 2, &tmp_value);
1714		if (len < 0)
1715			return len;
1716
1717		i += len;
1718		if (len == 2) {
1719			pkt_dev->traffic_class = tmp_value;
1720			sprintf(pg_result, "OK: traffic_class=0x%02x", pkt_dev->traffic_class);
1721		} else {
1722			sprintf(pg_result, "ERROR: traffic_class must be 00-ff");
1723		}
1724		return count;
1725	}
1726
1727	if (!strcmp(name, "skb_priority")) {
1728		len = num_arg(&user_buffer[i], 9, &value);
1729		if (len < 0)
1730			return len;
1731
1732		i += len;
1733		pkt_dev->skb_priority = value;
1734		sprintf(pg_result, "OK: skb_priority=%i",
1735			pkt_dev->skb_priority);
1736		return count;
1737	}
1738
1739	sprintf(pkt_dev->result, "No such parameter \"%s\"", name);
1740	return -EINVAL;
1741}
1742
1743static int pktgen_if_open(struct inode *inode, struct file *file)
1744{
1745	return single_open(file, pktgen_if_show, PDE_DATA(inode));
1746}
1747
1748static const struct file_operations pktgen_if_fops = {
1749	.owner   = THIS_MODULE,
1750	.open    = pktgen_if_open,
1751	.read    = seq_read,
1752	.llseek  = seq_lseek,
1753	.write   = pktgen_if_write,
1754	.release = single_release,
1755};
1756
1757static int pktgen_thread_show(struct seq_file *seq, void *v)
1758{
1759	struct pktgen_thread *t = seq->private;
1760	const struct pktgen_dev *pkt_dev;
1761
1762	BUG_ON(!t);
1763
1764	seq_puts(seq, "Running: ");
1765
1766	rcu_read_lock();
1767	list_for_each_entry_rcu(pkt_dev, &t->if_list, list)
1768		if (pkt_dev->running)
1769			seq_printf(seq, "%s ", pkt_dev->odevname);
1770
1771	seq_puts(seq, "\nStopped: ");
1772
1773	list_for_each_entry_rcu(pkt_dev, &t->if_list, list)
1774		if (!pkt_dev->running)
1775			seq_printf(seq, "%s ", pkt_dev->odevname);
1776
1777	if (t->result[0])
1778		seq_printf(seq, "\nResult: %s\n", t->result);
1779	else
1780		seq_puts(seq, "\nResult: NA\n");
1781
1782	rcu_read_unlock();
1783
1784	return 0;
1785}
1786
1787static ssize_t pktgen_thread_write(struct file *file,
1788				   const char __user * user_buffer,
1789				   size_t count, loff_t * offset)
1790{
1791	struct seq_file *seq = file->private_data;
1792	struct pktgen_thread *t = seq->private;
1793	int i, max, len, ret;
1794	char name[40];
1795	char *pg_result;
1796
1797	if (count < 1) {
1798		//      sprintf(pg_result, "Wrong command format");
1799		return -EINVAL;
1800	}
1801
1802	max = count;
1803	len = count_trail_chars(user_buffer, max);
1804	if (len < 0)
1805		return len;
1806
1807	i = len;
1808
1809	/* Read variable name */
1810
1811	len = strn_len(&user_buffer[i], sizeof(name) - 1);
1812	if (len < 0)
1813		return len;
1814
1815	memset(name, 0, sizeof(name));
1816	if (copy_from_user(name, &user_buffer[i], len))
1817		return -EFAULT;
1818	i += len;
1819
1820	max = count - i;
1821	len = count_trail_chars(&user_buffer[i], max);
1822	if (len < 0)
1823		return len;
1824
1825	i += len;
1826
1827	if (debug)
1828		pr_debug("t=%s, count=%lu\n", name, (unsigned long)count);
1829
1830	if (!t) {
1831		pr_err("ERROR: No thread\n");
1832		ret = -EINVAL;
1833		goto out;
1834	}
1835
1836	pg_result = &(t->result[0]);
1837
1838	if (!strcmp(name, "add_device")) {
1839		char f[32];
1840		memset(f, 0, 32);
1841		len = strn_len(&user_buffer[i], sizeof(f) - 1);
1842		if (len < 0) {
1843			ret = len;
1844			goto out;
1845		}
1846		if (copy_from_user(f, &user_buffer[i], len))
1847			return -EFAULT;
1848		i += len;
1849		mutex_lock(&pktgen_thread_lock);
1850		ret = pktgen_add_device(t, f);
1851		mutex_unlock(&pktgen_thread_lock);
1852		if (!ret) {
1853			ret = count;
1854			sprintf(pg_result, "OK: add_device=%s", f);
1855		} else
1856			sprintf(pg_result, "ERROR: can not add device %s", f);
1857		goto out;
1858	}
1859
1860	if (!strcmp(name, "rem_device_all")) {
1861		mutex_lock(&pktgen_thread_lock);
1862		t->control |= T_REMDEVALL;
1863		mutex_unlock(&pktgen_thread_lock);
1864		schedule_timeout_interruptible(msecs_to_jiffies(125));	/* Propagate thread->control  */
1865		ret = count;
1866		sprintf(pg_result, "OK: rem_device_all");
1867		goto out;
1868	}
1869
1870	if (!strcmp(name, "max_before_softirq")) {
1871		sprintf(pg_result, "OK: Note! max_before_softirq is obsoleted -- Do not use");
1872		ret = count;
1873		goto out;
1874	}
1875
1876	ret = -EINVAL;
1877out:
1878	return ret;
1879}
1880
1881static int pktgen_thread_open(struct inode *inode, struct file *file)
1882{
1883	return single_open(file, pktgen_thread_show, PDE_DATA(inode));
1884}
1885
1886static const struct file_operations pktgen_thread_fops = {
1887	.owner   = THIS_MODULE,
1888	.open    = pktgen_thread_open,
1889	.read    = seq_read,
1890	.llseek  = seq_lseek,
1891	.write   = pktgen_thread_write,
1892	.release = single_release,
1893};
1894
1895/* Think find or remove for NN */
1896static struct pktgen_dev *__pktgen_NN_threads(const struct pktgen_net *pn,
1897					      const char *ifname, int remove)
1898{
1899	struct pktgen_thread *t;
1900	struct pktgen_dev *pkt_dev = NULL;
1901	bool exact = (remove == FIND);
1902
1903	list_for_each_entry(t, &pn->pktgen_threads, th_list) {
1904		pkt_dev = pktgen_find_dev(t, ifname, exact);
1905		if (pkt_dev) {
1906			if (remove) {
1907				pkt_dev->removal_mark = 1;
1908				t->control |= T_REMDEV;
1909			}
1910			break;
1911		}
1912	}
1913	return pkt_dev;
1914}
1915
1916/*
1917 * mark a device for removal
1918 */
1919static void pktgen_mark_device(const struct pktgen_net *pn, const char *ifname)
1920{
1921	struct pktgen_dev *pkt_dev = NULL;
1922	const int max_tries = 10, msec_per_try = 125;
1923	int i = 0;
1924
1925	mutex_lock(&pktgen_thread_lock);
1926	pr_debug("%s: marking %s for removal\n", __func__, ifname);
1927
1928	while (1) {
1929
1930		pkt_dev = __pktgen_NN_threads(pn, ifname, REMOVE);
1931		if (pkt_dev == NULL)
1932			break;	/* success */
1933
1934		mutex_unlock(&pktgen_thread_lock);
1935		pr_debug("%s: waiting for %s to disappear....\n",
1936			 __func__, ifname);
1937		schedule_timeout_interruptible(msecs_to_jiffies(msec_per_try));
1938		mutex_lock(&pktgen_thread_lock);
1939
1940		if (++i >= max_tries) {
1941			pr_err("%s: timed out after waiting %d msec for device %s to be removed\n",
1942			       __func__, msec_per_try * i, ifname);
1943			break;
1944		}
1945
1946	}
1947
1948	mutex_unlock(&pktgen_thread_lock);
1949}
1950
1951static void pktgen_change_name(const struct pktgen_net *pn, struct net_device *dev)
1952{
1953	struct pktgen_thread *t;
1954
1955	list_for_each_entry(t, &pn->pktgen_threads, th_list) {
1956		struct pktgen_dev *pkt_dev;
1957
1958		rcu_read_lock();
1959		list_for_each_entry_rcu(pkt_dev, &t->if_list, list) {
1960			if (pkt_dev->odev != dev)
1961				continue;
1962
1963			proc_remove(pkt_dev->entry);
1964
1965			pkt_dev->entry = proc_create_data(dev->name, 0600,
1966							  pn->proc_dir,
1967							  &pktgen_if_fops,
1968							  pkt_dev);
1969			if (!pkt_dev->entry)
1970				pr_err("can't move proc entry for '%s'\n",
1971				       dev->name);
1972			break;
1973		}
1974		rcu_read_unlock();
1975	}
1976}
1977
1978static int pktgen_device_event(struct notifier_block *unused,
1979			       unsigned long event, void *ptr)
1980{
1981	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1982	struct pktgen_net *pn = net_generic(dev_net(dev), pg_net_id);
1983
1984	if (pn->pktgen_exiting)
1985		return NOTIFY_DONE;
1986
1987	/* It is OK that we do not hold the group lock right now,
1988	 * as we run under the RTNL lock.
1989	 */
1990
1991	switch (event) {
1992	case NETDEV_CHANGENAME:
1993		pktgen_change_name(pn, dev);
1994		break;
1995
1996	case NETDEV_UNREGISTER:
1997		pktgen_mark_device(pn, dev->name);
1998		break;
1999	}
2000
2001	return NOTIFY_DONE;
2002}
2003
2004static struct net_device *pktgen_dev_get_by_name(const struct pktgen_net *pn,
2005						 struct pktgen_dev *pkt_dev,
2006						 const char *ifname)
2007{
2008	char b[IFNAMSIZ+5];
2009	int i;
2010
2011	for (i = 0; ifname[i] != '@'; i++) {
2012		if (i == IFNAMSIZ)
2013			break;
2014
2015		b[i] = ifname[i];
2016	}
2017	b[i] = 0;
2018
2019	return dev_get_by_name(pn->net, b);
2020}
2021
2022
2023/* Associate pktgen_dev with a device. */
2024
2025static int pktgen_setup_dev(const struct pktgen_net *pn,
2026			    struct pktgen_dev *pkt_dev, const char *ifname)
2027{
2028	struct net_device *odev;
2029	int err;
2030
2031	/* Clean old setups */
2032	if (pkt_dev->odev) {
2033		dev_put(pkt_dev->odev);
2034		pkt_dev->odev = NULL;
2035	}
2036
2037	odev = pktgen_dev_get_by_name(pn, pkt_dev, ifname);
2038	if (!odev) {
2039		pr_err("no such netdevice: \"%s\"\n", ifname);
2040		return -ENODEV;
2041	}
2042
2043	if (odev->type != ARPHRD_ETHER) {
2044		pr_err("not an ethernet device: \"%s\"\n", ifname);
2045		err = -EINVAL;
2046	} else if (!netif_running(odev)) {
2047		pr_err("device is down: \"%s\"\n", ifname);
2048		err = -ENETDOWN;
2049	} else {
2050		pkt_dev->odev = odev;
2051		return 0;
2052	}
2053
2054	dev_put(odev);
2055	return err;
2056}
2057
2058/* Read pkt_dev from the interface and set up internal pktgen_dev
2059 * structure to have the right information to create/send packets
2060 */
2061static void pktgen_setup_inject(struct pktgen_dev *pkt_dev)
2062{
2063	int ntxq;
2064
2065	if (!pkt_dev->odev) {
2066		pr_err("ERROR: pkt_dev->odev == NULL in setup_inject\n");
2067		sprintf(pkt_dev->result,
2068			"ERROR: pkt_dev->odev == NULL in setup_inject.\n");
2069		return;
2070	}
2071
2072	/* make sure that we don't pick a non-existing transmit queue */
2073	ntxq = pkt_dev->odev->real_num_tx_queues;
2074
2075	if (ntxq <= pkt_dev->queue_map_min) {
2076		pr_warn("WARNING: Requested queue_map_min (zero-based) (%d) exceeds valid range [0 - %d] for (%d) queues on %s, resetting\n",
2077			pkt_dev->queue_map_min, (ntxq ?: 1) - 1, ntxq,
2078			pkt_dev->odevname);
2079		pkt_dev->queue_map_min = (ntxq ?: 1) - 1;
2080	}
2081	if (pkt_dev->queue_map_max >= ntxq) {
2082		pr_warn("WARNING: Requested queue_map_max (zero-based) (%d) exceeds valid range [0 - %d] for (%d) queues on %s, resetting\n",
2083			pkt_dev->queue_map_max, (ntxq ?: 1) - 1, ntxq,
2084			pkt_dev->odevname);
2085		pkt_dev->queue_map_max = (ntxq ?: 1) - 1;
2086	}
2087
2088	/* Default to the interface's mac if not explicitly set. */
2089
2090	if (is_zero_ether_addr(pkt_dev->src_mac))
2091		ether_addr_copy(&(pkt_dev->hh[6]), pkt_dev->odev->dev_addr);
2092
2093	/* Set up Dest MAC */
2094	ether_addr_copy(&(pkt_dev->hh[0]), pkt_dev->dst_mac);
2095
2096	if (pkt_dev->flags & F_IPV6) {
2097		int i, set = 0, err = 1;
2098		struct inet6_dev *idev;
2099
2100		if (pkt_dev->min_pkt_size == 0) {
2101			pkt_dev->min_pkt_size = 14 + sizeof(struct ipv6hdr)
2102						+ sizeof(struct udphdr)
2103						+ sizeof(struct pktgen_hdr)
2104						+ pkt_dev->pkt_overhead;
2105		}
2106
2107		for (i = 0; i < IN6_ADDR_HSIZE; i++)
2108			if (pkt_dev->cur_in6_saddr.s6_addr[i]) {
2109				set = 1;
2110				break;
2111			}
2112
2113		if (!set) {
2114
2115			/*
2116			 * Use linklevel address if unconfigured.
2117			 *
2118			 * use ipv6_get_lladdr if/when it's get exported
2119			 */
2120
2121			rcu_read_lock();
2122			idev = __in6_dev_get(pkt_dev->odev);
2123			if (idev) {
2124				struct inet6_ifaddr *ifp;
2125
2126				read_lock_bh(&idev->lock);
2127				list_for_each_entry(ifp, &idev->addr_list, if_list) {
2128					if ((ifp->scope & IFA_LINK) &&
2129					    !(ifp->flags & IFA_F_TENTATIVE)) {
2130						pkt_dev->cur_in6_saddr = ifp->addr;
2131						err = 0;
2132						break;
2133					}
2134				}
2135				read_unlock_bh(&idev->lock);
2136			}
2137			rcu_read_unlock();
2138			if (err)
2139				pr_err("ERROR: IPv6 link address not available\n");
2140		}
2141	} else {
2142		if (pkt_dev->min_pkt_size == 0) {
2143			pkt_dev->min_pkt_size = 14 + sizeof(struct iphdr)
2144						+ sizeof(struct udphdr)
2145						+ sizeof(struct pktgen_hdr)
2146						+ pkt_dev->pkt_overhead;
2147		}
2148
2149		pkt_dev->saddr_min = 0;
2150		pkt_dev->saddr_max = 0;
2151		if (strlen(pkt_dev->src_min) == 0) {
2152
2153			struct in_device *in_dev;
2154
2155			rcu_read_lock();
2156			in_dev = __in_dev_get_rcu(pkt_dev->odev);
2157			if (in_dev) {
2158				if (in_dev->ifa_list) {
2159					pkt_dev->saddr_min =
2160					    in_dev->ifa_list->ifa_address;
2161					pkt_dev->saddr_max = pkt_dev->saddr_min;
2162				}
2163			}
2164			rcu_read_unlock();
2165		} else {
2166			pkt_dev->saddr_min = in_aton(pkt_dev->src_min);
2167			pkt_dev->saddr_max = in_aton(pkt_dev->src_max);
2168		}
2169
2170		pkt_dev->daddr_min = in_aton(pkt_dev->dst_min);
2171		pkt_dev->daddr_max = in_aton(pkt_dev->dst_max);
2172	}
2173	/* Initialize current values. */
2174	pkt_dev->cur_pkt_size = pkt_dev->min_pkt_size;
2175	if (pkt_dev->min_pkt_size > pkt_dev->max_pkt_size)
2176		pkt_dev->max_pkt_size = pkt_dev->min_pkt_size;
2177
2178	pkt_dev->cur_dst_mac_offset = 0;
2179	pkt_dev->cur_src_mac_offset = 0;
2180	pkt_dev->cur_saddr = pkt_dev->saddr_min;
2181	pkt_dev->cur_daddr = pkt_dev->daddr_min;
2182	pkt_dev->cur_udp_dst = pkt_dev->udp_dst_min;
2183	pkt_dev->cur_udp_src = pkt_dev->udp_src_min;
2184	pkt_dev->nflows = 0;
2185}
2186
2187
2188static void spin(struct pktgen_dev *pkt_dev, ktime_t spin_until)
2189{
2190	ktime_t start_time, end_time;
2191	s64 remaining;
2192	struct hrtimer_sleeper t;
2193
2194	hrtimer_init_on_stack(&t.timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS);
2195	hrtimer_set_expires(&t.timer, spin_until);
2196
2197	remaining = ktime_to_ns(hrtimer_expires_remaining(&t.timer));
2198	if (remaining <= 0) {
2199		pkt_dev->next_tx = ktime_add_ns(spin_until, pkt_dev->delay);
2200		return;
2201	}
2202
2203	start_time = ktime_get();
2204	if (remaining < 100000) {
2205		/* for small delays (<100us), just loop until limit is reached */
2206		do {
2207			end_time = ktime_get();
2208		} while (ktime_compare(end_time, spin_until) < 0);
2209	} else {
2210		/* see do_nanosleep */
2211		hrtimer_init_sleeper(&t, current);
2212		do {
2213			set_current_state(TASK_INTERRUPTIBLE);
2214			hrtimer_start_expires(&t.timer, HRTIMER_MODE_ABS);
2215			if (!hrtimer_active(&t.timer))
2216				t.task = NULL;
2217
2218			if (likely(t.task))
2219				schedule();
2220
2221			hrtimer_cancel(&t.timer);
2222		} while (t.task && pkt_dev->running && !signal_pending(current));
2223		__set_current_state(TASK_RUNNING);
2224		end_time = ktime_get();
2225	}
2226
2227	pkt_dev->idle_acc += ktime_to_ns(ktime_sub(end_time, start_time));
2228	pkt_dev->next_tx = ktime_add_ns(spin_until, pkt_dev->delay);
2229}
2230
2231static inline void set_pkt_overhead(struct pktgen_dev *pkt_dev)
2232{
2233	pkt_dev->pkt_overhead = 0;
2234	pkt_dev->pkt_overhead += pkt_dev->nr_labels*sizeof(u32);
2235	pkt_dev->pkt_overhead += VLAN_TAG_SIZE(pkt_dev);
2236	pkt_dev->pkt_overhead += SVLAN_TAG_SIZE(pkt_dev);
2237}
2238
2239static inline int f_seen(const struct pktgen_dev *pkt_dev, int flow)
2240{
2241	return !!(pkt_dev->flows[flow].flags & F_INIT);
2242}
2243
2244static inline int f_pick(struct pktgen_dev *pkt_dev)
2245{
2246	int flow = pkt_dev->curfl;
2247
2248	if (pkt_dev->flags & F_FLOW_SEQ) {
2249		if (pkt_dev->flows[flow].count >= pkt_dev->lflow) {
2250			/* reset time */
2251			pkt_dev->flows[flow].count = 0;
2252			pkt_dev->flows[flow].flags = 0;
2253			pkt_dev->curfl += 1;
2254			if (pkt_dev->curfl >= pkt_dev->cflows)
2255				pkt_dev->curfl = 0; /*reset */
2256		}
2257	} else {
2258		flow = prandom_u32() % pkt_dev->cflows;
2259		pkt_dev->curfl = flow;
2260
2261		if (pkt_dev->flows[flow].count > pkt_dev->lflow) {
2262			pkt_dev->flows[flow].count = 0;
2263			pkt_dev->flows[flow].flags = 0;
2264		}
2265	}
2266
2267	return pkt_dev->curfl;
2268}
2269
2270
2271#ifdef CONFIG_XFRM
2272/* If there was already an IPSEC SA, we keep it as is, else
2273 * we go look for it ...
2274*/
2275#define DUMMY_MARK 0
2276static void get_ipsec_sa(struct pktgen_dev *pkt_dev, int flow)
2277{
2278	struct xfrm_state *x = pkt_dev->flows[flow].x;
2279	struct pktgen_net *pn = net_generic(dev_net(pkt_dev->odev), pg_net_id);
2280	if (!x) {
2281
2282		if (pkt_dev->spi) {
2283			/* We need as quick as possible to find the right SA
2284			 * Searching with minimum criteria to archieve this.
2285			 */
2286			x = xfrm_state_lookup_byspi(pn->net, htonl(pkt_dev->spi), AF_INET);
2287		} else {
2288			/* slow path: we dont already have xfrm_state */
2289			x = xfrm_stateonly_find(pn->net, DUMMY_MARK,
2290						(xfrm_address_t *)&pkt_dev->cur_daddr,
2291						(xfrm_address_t *)&pkt_dev->cur_saddr,
2292						AF_INET,
2293						pkt_dev->ipsmode,
2294						pkt_dev->ipsproto, 0);
2295		}
2296		if (x) {
2297			pkt_dev->flows[flow].x = x;
2298			set_pkt_overhead(pkt_dev);
2299			pkt_dev->pkt_overhead += x->props.header_len;
2300		}
2301
2302	}
2303}
2304#endif
2305static void set_cur_queue_map(struct pktgen_dev *pkt_dev)
2306{
2307
2308	if (pkt_dev->flags & F_QUEUE_MAP_CPU)
2309		pkt_dev->cur_queue_map = smp_processor_id();
2310
2311	else if (pkt_dev->queue_map_min <= pkt_dev->queue_map_max) {
2312		__u16 t;
2313		if (pkt_dev->flags & F_QUEUE_MAP_RND) {
2314			t = prandom_u32() %
2315				(pkt_dev->queue_map_max -
2316				 pkt_dev->queue_map_min + 1)
2317				+ pkt_dev->queue_map_min;
2318		} else {
2319			t = pkt_dev->cur_queue_map + 1;
2320			if (t > pkt_dev->queue_map_max)
2321				t = pkt_dev->queue_map_min;
2322		}
2323		pkt_dev->cur_queue_map = t;
2324	}
2325	pkt_dev->cur_queue_map  = pkt_dev->cur_queue_map % pkt_dev->odev->real_num_tx_queues;
2326}
2327
2328/* Increment/randomize headers according to flags and current values
2329 * for IP src/dest, UDP src/dst port, MAC-Addr src/dst
2330 */
2331static void mod_cur_headers(struct pktgen_dev *pkt_dev)
2332{
2333	__u32 imn;
2334	__u32 imx;
2335	int flow = 0;
2336
2337	if (pkt_dev->cflows)
2338		flow = f_pick(pkt_dev);
2339
2340	/*  Deal with source MAC */
2341	if (pkt_dev->src_mac_count > 1) {
2342		__u32 mc;
2343		__u32 tmp;
2344
2345		if (pkt_dev->flags & F_MACSRC_RND)
2346			mc = prandom_u32() % pkt_dev->src_mac_count;
2347		else {
2348			mc = pkt_dev->cur_src_mac_offset++;
2349			if (pkt_dev->cur_src_mac_offset >=
2350			    pkt_dev->src_mac_count)
2351				pkt_dev->cur_src_mac_offset = 0;
2352		}
2353
2354		tmp = pkt_dev->src_mac[5] + (mc & 0xFF);
2355		pkt_dev->hh[11] = tmp;
2356		tmp = (pkt_dev->src_mac[4] + ((mc >> 8) & 0xFF) + (tmp >> 8));
2357		pkt_dev->hh[10] = tmp;
2358		tmp = (pkt_dev->src_mac[3] + ((mc >> 16) & 0xFF) + (tmp >> 8));
2359		pkt_dev->hh[9] = tmp;
2360		tmp = (pkt_dev->src_mac[2] + ((mc >> 24) & 0xFF) + (tmp >> 8));
2361		pkt_dev->hh[8] = tmp;
2362		tmp = (pkt_dev->src_mac[1] + (tmp >> 8));
2363		pkt_dev->hh[7] = tmp;
2364	}
2365
2366	/*  Deal with Destination MAC */
2367	if (pkt_dev->dst_mac_count > 1) {
2368		__u32 mc;
2369		__u32 tmp;
2370
2371		if (pkt_dev->flags & F_MACDST_RND)
2372			mc = prandom_u32() % pkt_dev->dst_mac_count;
2373
2374		else {
2375			mc = pkt_dev->cur_dst_mac_offset++;
2376			if (pkt_dev->cur_dst_mac_offset >=
2377			    pkt_dev->dst_mac_count) {
2378				pkt_dev->cur_dst_mac_offset = 0;
2379			}
2380		}
2381
2382		tmp = pkt_dev->dst_mac[5] + (mc & 0xFF);
2383		pkt_dev->hh[5] = tmp;
2384		tmp = (pkt_dev->dst_mac[4] + ((mc >> 8) & 0xFF) + (tmp >> 8));
2385		pkt_dev->hh[4] = tmp;
2386		tmp = (pkt_dev->dst_mac[3] + ((mc >> 16) & 0xFF) + (tmp >> 8));
2387		pkt_dev->hh[3] = tmp;
2388		tmp = (pkt_dev->dst_mac[2] + ((mc >> 24) & 0xFF) + (tmp >> 8));
2389		pkt_dev->hh[2] = tmp;
2390		tmp = (pkt_dev->dst_mac[1] + (tmp >> 8));
2391		pkt_dev->hh[1] = tmp;
2392	}
2393
2394	if (pkt_dev->flags & F_MPLS_RND) {
2395		unsigned int i;
2396		for (i = 0; i < pkt_dev->nr_labels; i++)
2397			if (pkt_dev->labels[i] & MPLS_STACK_BOTTOM)
2398				pkt_dev->labels[i] = MPLS_STACK_BOTTOM |
2399					     ((__force __be32)prandom_u32() &
2400						      htonl(0x000fffff));
2401	}
2402
2403	if ((pkt_dev->flags & F_VID_RND) && (pkt_dev->vlan_id != 0xffff)) {
2404		pkt_dev->vlan_id = prandom_u32() & (4096 - 1);
2405	}
2406
2407	if ((pkt_dev->flags & F_SVID_RND) && (pkt_dev->svlan_id != 0xffff)) {
2408		pkt_dev->svlan_id = prandom_u32() & (4096 - 1);
2409	}
2410
2411	if (pkt_dev->udp_src_min < pkt_dev->udp_src_max) {
2412		if (pkt_dev->flags & F_UDPSRC_RND)
2413			pkt_dev->cur_udp_src = prandom_u32() %
2414				(pkt_dev->udp_src_max - pkt_dev->udp_src_min)
2415				+ pkt_dev->udp_src_min;
2416
2417		else {
2418			pkt_dev->cur_udp_src++;
2419			if (pkt_dev->cur_udp_src >= pkt_dev->udp_src_max)
2420				pkt_dev->cur_udp_src = pkt_dev->udp_src_min;
2421		}
2422	}
2423
2424	if (pkt_dev->udp_dst_min < pkt_dev->udp_dst_max) {
2425		if (pkt_dev->flags & F_UDPDST_RND) {
2426			pkt_dev->cur_udp_dst = prandom_u32() %
2427				(pkt_dev->udp_dst_max - pkt_dev->udp_dst_min)
2428				+ pkt_dev->udp_dst_min;
2429		} else {
2430			pkt_dev->cur_udp_dst++;
2431			if (pkt_dev->cur_udp_dst >= pkt_dev->udp_dst_max)
2432				pkt_dev->cur_udp_dst = pkt_dev->udp_dst_min;
2433		}
2434	}
2435
2436	if (!(pkt_dev->flags & F_IPV6)) {
2437
2438		imn = ntohl(pkt_dev->saddr_min);
2439		imx = ntohl(pkt_dev->saddr_max);
2440		if (imn < imx) {
2441			__u32 t;
2442			if (pkt_dev->flags & F_IPSRC_RND)
2443				t = prandom_u32() % (imx - imn) + imn;
2444			else {
2445				t = ntohl(pkt_dev->cur_saddr);
2446				t++;
2447				if (t > imx)
2448					t = imn;
2449
2450			}
2451			pkt_dev->cur_saddr = htonl(t);
2452		}
2453
2454		if (pkt_dev->cflows && f_seen(pkt_dev, flow)) {
2455			pkt_dev->cur_daddr = pkt_dev->flows[flow].cur_daddr;
2456		} else {
2457			imn = ntohl(pkt_dev->daddr_min);
2458			imx = ntohl(pkt_dev->daddr_max);
2459			if (imn < imx) {
2460				__u32 t;
2461				__be32 s;
2462				if (pkt_dev->flags & F_IPDST_RND) {
2463
2464					do {
2465						t = prandom_u32() %
2466							(imx - imn) + imn;
2467						s = htonl(t);
2468					} while (ipv4_is_loopback(s) ||
2469						ipv4_is_multicast(s) ||
2470						ipv4_is_lbcast(s) ||
2471						ipv4_is_zeronet(s) ||
2472						ipv4_is_local_multicast(s));
2473					pkt_dev->cur_daddr = s;
2474				} else {
2475					t = ntohl(pkt_dev->cur_daddr);
2476					t++;
2477					if (t > imx) {
2478						t = imn;
2479					}
2480					pkt_dev->cur_daddr = htonl(t);
2481				}
2482			}
2483			if (pkt_dev->cflows) {
2484				pkt_dev->flows[flow].flags |= F_INIT;
2485				pkt_dev->flows[flow].cur_daddr =
2486				    pkt_dev->cur_daddr;
2487#ifdef CONFIG_XFRM
2488				if (pkt_dev->flags & F_IPSEC_ON)
2489					get_ipsec_sa(pkt_dev, flow);
2490#endif
2491				pkt_dev->nflows++;
2492			}
2493		}
2494	} else {		/* IPV6 * */
2495
2496		if (!ipv6_addr_any(&pkt_dev->min_in6_daddr)) {
2497			int i;
2498
2499			/* Only random destinations yet */
2500
2501			for (i = 0; i < 4; i++) {
2502				pkt_dev->cur_in6_daddr.s6_addr32[i] =
2503				    (((__force __be32)prandom_u32() |
2504				      pkt_dev->min_in6_daddr.s6_addr32[i]) &
2505				     pkt_dev->max_in6_daddr.s6_addr32[i]);
2506			}
2507		}
2508	}
2509
2510	if (pkt_dev->min_pkt_size < pkt_dev->max_pkt_size) {
2511		__u32 t;
2512		if (pkt_dev->flags & F_TXSIZE_RND) {
2513			t = prandom_u32() %
2514				(pkt_dev->max_pkt_size - pkt_dev->min_pkt_size)
2515				+ pkt_dev->min_pkt_size;
2516		} else {
2517			t = pkt_dev->cur_pkt_size + 1;
2518			if (t > pkt_dev->max_pkt_size)
2519				t = pkt_dev->min_pkt_size;
2520		}
2521		pkt_dev->cur_pkt_size = t;
2522	}
2523
2524	set_cur_queue_map(pkt_dev);
2525
2526	pkt_dev->flows[flow].count++;
2527}
2528
2529
2530#ifdef CONFIG_XFRM
2531static u32 pktgen_dst_metrics[RTAX_MAX + 1] = {
2532
2533	[RTAX_HOPLIMIT] = 0x5, /* Set a static hoplimit */
2534};
2535
2536static int pktgen_output_ipsec(struct sk_buff *skb, struct pktgen_dev *pkt_dev)
2537{
2538	struct xfrm_state *x = pkt_dev->flows[pkt_dev->curfl].x;
2539	int err = 0;
2540	struct net *net = dev_net(pkt_dev->odev);
2541
2542	if (!x)
2543		return 0;
2544	/* XXX: we dont support tunnel mode for now until
2545	 * we resolve the dst issue */
2546	if ((x->props.mode != XFRM_MODE_TRANSPORT) && (pkt_dev->spi == 0))
2547		return 0;
2548
2549	/* But when user specify an valid SPI, transformation
2550	 * supports both transport/tunnel mode + ESP/AH type.
2551	 */
2552	if ((x->props.mode == XFRM_MODE_TUNNEL) && (pkt_dev->spi != 0))
2553		skb->_skb_refdst = (unsigned long)&pkt_dev->dst | SKB_DST_NOREF;
2554
2555	rcu_read_lock_bh();
2556	err = x->outer_mode->output(x, skb);
2557	rcu_read_unlock_bh();
2558	if (err) {
2559		XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATEMODEERROR);
2560		goto error;
2561	}
2562	err = x->type->output(x, skb);
2563	if (err) {
2564		XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATEPROTOERROR);
2565		goto error;
2566	}
2567	spin_lock_bh(&x->lock);
2568	x->curlft.bytes += skb->len;
2569	x->curlft.packets++;
2570	spin_unlock_bh(&x->lock);
2571error:
2572	return err;
2573}
2574
2575static void free_SAs(struct pktgen_dev *pkt_dev)
2576{
2577	if (pkt_dev->cflows) {
2578		/* let go of the SAs if we have them */
2579		int i;
2580		for (i = 0; i < pkt_dev->cflows; i++) {
2581			struct xfrm_state *x = pkt_dev->flows[i].x;
2582			if (x) {
2583				xfrm_state_put(x);
2584				pkt_dev->flows[i].x = NULL;
2585			}
2586		}
2587	}
2588}
2589
2590static int process_ipsec(struct pktgen_dev *pkt_dev,
2591			      struct sk_buff *skb, __be16 protocol)
2592{
2593	if (pkt_dev->flags & F_IPSEC_ON) {
2594		struct xfrm_state *x = pkt_dev->flows[pkt_dev->curfl].x;
2595		int nhead = 0;
2596		if (x) {
2597			int ret;
2598			__u8 *eth;
2599			struct iphdr *iph;
2600
2601			nhead = x->props.header_len - skb_headroom(skb);
2602			if (nhead > 0) {
2603				ret = pskb_expand_head(skb, nhead, 0, GFP_ATOMIC);
2604				if (ret < 0) {
2605					pr_err("Error expanding ipsec packet %d\n",
2606					       ret);
2607					goto err;
2608				}
2609			}
2610
2611			/* ipsec is not expecting ll header */
2612			skb_pull(skb, ETH_HLEN);
2613			ret = pktgen_output_ipsec(skb, pkt_dev);
2614			if (ret) {
2615				pr_err("Error creating ipsec packet %d\n", ret);
2616				goto err;
2617			}
2618			/* restore ll */
2619			eth = (__u8 *) skb_push(skb, ETH_HLEN);
2620			memcpy(eth, pkt_dev->hh, 12);
2621			*(u16 *) &eth[12] = protocol;
2622
2623			/* Update IPv4 header len as well as checksum value */
2624			iph = ip_hdr(skb);
2625			iph->tot_len = htons(skb->len - ETH_HLEN);
2626			ip_send_check(iph);
2627		}
2628	}
2629	return 1;
2630err:
2631	kfree_skb(skb);
2632	return 0;
2633}
2634#endif
2635
2636static void mpls_push(__be32 *mpls, struct pktgen_dev *pkt_dev)
2637{
2638	unsigned int i;
2639	for (i = 0; i < pkt_dev->nr_labels; i++)
2640		*mpls++ = pkt_dev->labels[i] & ~MPLS_STACK_BOTTOM;
2641
2642	mpls--;
2643	*mpls |= MPLS_STACK_BOTTOM;
2644}
2645
2646static inline __be16 build_tci(unsigned int id, unsigned int cfi,
2647			       unsigned int prio)
2648{
2649	return htons(id | (cfi << 12) | (prio << 13));
2650}
2651
2652static void pktgen_finalize_skb(struct pktgen_dev *pkt_dev, struct sk_buff *skb,
2653				int datalen)
2654{
2655	struct timeval timestamp;
2656	struct pktgen_hdr *pgh;
2657
2658	pgh = (struct pktgen_hdr *)skb_put(skb, sizeof(*pgh));
2659	datalen -= sizeof(*pgh);
2660
2661	if (pkt_dev->nfrags <= 0) {
2662		memset(skb_put(skb, datalen), 0, datalen);
2663	} else {
2664		int frags = pkt_dev->nfrags;
2665		int i, len;
2666		int frag_len;
2667
2668
2669		if (frags > MAX_SKB_FRAGS)
2670			frags = MAX_SKB_FRAGS;
2671		len = datalen - frags * PAGE_SIZE;
2672		if (len > 0) {
2673			memset(skb_put(skb, len), 0, len);
2674			datalen = frags * PAGE_SIZE;
2675		}
2676
2677		i = 0;
2678		frag_len = (datalen/frags) < PAGE_SIZE ?
2679			   (datalen/frags) : PAGE_SIZE;
2680		while (datalen > 0) {
2681			if (unlikely(!pkt_dev->page)) {
2682				int node = numa_node_id();
2683
2684				if (pkt_dev->node >= 0 && (pkt_dev->flags & F_NODE))
2685					node = pkt_dev->node;
2686				pkt_dev->page = alloc_pages_node(node, GFP_KERNEL | __GFP_ZERO, 0);
2687				if (!pkt_dev->page)
2688					break;
2689			}
2690			get_page(pkt_dev->page);
2691			skb_frag_set_page(skb, i, pkt_dev->page);
2692			skb_shinfo(skb)->frags[i].page_offset = 0;
2693			/*last fragment, fill rest of data*/
2694			if (i == (frags - 1))
2695				skb_frag_size_set(&skb_shinfo(skb)->frags[i],
2696				    (datalen < PAGE_SIZE ? datalen : PAGE_SIZE));
2697			else
2698				skb_frag_size_set(&skb_shinfo(skb)->frags[i], frag_len);
2699			datalen -= skb_frag_size(&skb_shinfo(skb)->frags[i]);
2700			skb->len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2701			skb->data_len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2702			i++;
2703			skb_shinfo(skb)->nr_frags = i;
2704		}
2705	}
2706
2707	/* Stamp the time, and sequence number,
2708	 * convert them to network byte order
2709	 */
2710	pgh->pgh_magic = htonl(PKTGEN_MAGIC);
2711	pgh->seq_num = htonl(pkt_dev->seq_num);
2712
2713	if (pkt_dev->flags & F_NO_TIMESTAMP) {
2714		pgh->tv_sec = 0;
2715		pgh->tv_usec = 0;
2716	} else {
2717		do_gettimeofday(&timestamp);
2718		pgh->tv_sec = htonl(timestamp.tv_sec);
2719		pgh->tv_usec = htonl(timestamp.tv_usec);
2720	}
2721}
2722
2723static struct sk_buff *pktgen_alloc_skb(struct net_device *dev,
2724					struct pktgen_dev *pkt_dev,
2725					unsigned int extralen)
2726{
2727	struct sk_buff *skb = NULL;
2728	unsigned int size = pkt_dev->cur_pkt_size + 64 + extralen +
2729			    pkt_dev->pkt_overhead;
2730
2731	if (pkt_dev->flags & F_NODE) {
2732		int node = pkt_dev->node >= 0 ? pkt_dev->node : numa_node_id();
2733
2734		skb = __alloc_skb(NET_SKB_PAD + size, GFP_NOWAIT, 0, node);
2735		if (likely(skb)) {
2736			skb_reserve(skb, NET_SKB_PAD);
2737			skb->dev = dev;
2738		}
2739	} else {
2740		 skb = __netdev_alloc_skb(dev, size, GFP_NOWAIT);
2741	}
2742
2743	return skb;
2744}
2745
2746static struct sk_buff *fill_packet_ipv4(struct net_device *odev,
2747					struct pktgen_dev *pkt_dev)
2748{
2749	struct sk_buff *skb = NULL;
2750	__u8 *eth;
2751	struct udphdr *udph;
2752	int datalen, iplen;
2753	struct iphdr *iph;
2754	__be16 protocol = htons(ETH_P_IP);
2755	__be32 *mpls;
2756	__be16 *vlan_tci = NULL;                 /* Encapsulates priority and VLAN ID */
2757	__be16 *vlan_encapsulated_proto = NULL;  /* packet type ID field (or len) for VLAN tag */
2758	__be16 *svlan_tci = NULL;                /* Encapsulates priority and SVLAN ID */
2759	__be16 *svlan_encapsulated_proto = NULL; /* packet type ID field (or len) for SVLAN tag */
2760	u16 queue_map;
2761
2762	if (pkt_dev->nr_labels)
2763		protocol = htons(ETH_P_MPLS_UC);
2764
2765	if (pkt_dev->vlan_id != 0xffff)
2766		protocol = htons(ETH_P_8021Q);
2767
2768	/* Update any of the values, used when we're incrementing various
2769	 * fields.
2770	 */
2771	mod_cur_headers(pkt_dev);
2772	queue_map = pkt_dev->cur_queue_map;
2773
2774	datalen = (odev->hard_header_len + 16) & ~0xf;
2775
2776	skb = pktgen_alloc_skb(odev, pkt_dev, datalen);
2777	if (!skb) {
2778		sprintf(pkt_dev->result, "No memory");
2779		return NULL;
2780	}
2781
2782	prefetchw(skb->data);
2783	skb_reserve(skb, datalen);
2784
2785	/*  Reserve for ethernet and IP header  */
2786	eth = (__u8 *) skb_push(skb, 14);
2787	mpls = (__be32 *)skb_put(skb, pkt_dev->nr_labels*sizeof(__u32));
2788	if (pkt_dev->nr_labels)
2789		mpls_push(mpls, pkt_dev);
2790
2791	if (pkt_dev->vlan_id != 0xffff) {
2792		if (pkt_dev->svlan_id != 0xffff) {
2793			svlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2794			*svlan_tci = build_tci(pkt_dev->svlan_id,
2795					       pkt_dev->svlan_cfi,
2796					       pkt_dev->svlan_p);
2797			svlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2798			*svlan_encapsulated_proto = htons(ETH_P_8021Q);
2799		}
2800		vlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2801		*vlan_tci = build_tci(pkt_dev->vlan_id,
2802				      pkt_dev->vlan_cfi,
2803				      pkt_dev->vlan_p);
2804		vlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2805		*vlan_encapsulated_proto = htons(ETH_P_IP);
2806	}
2807
2808	skb_set_mac_header(skb, 0);
2809	skb_set_network_header(skb, skb->len);
2810	iph = (struct iphdr *) skb_put(skb, sizeof(struct iphdr));
2811
2812	skb_set_transport_header(skb, skb->len);
2813	udph = (struct udphdr *) skb_put(skb, sizeof(struct udphdr));
2814	skb_set_queue_mapping(skb, queue_map);
2815	skb->priority = pkt_dev->skb_priority;
2816
2817	memcpy(eth, pkt_dev->hh, 12);
2818	*(__be16 *) & eth[12] = protocol;
2819
2820	/* Eth + IPh + UDPh + mpls */
2821	datalen = pkt_dev->cur_pkt_size - 14 - 20 - 8 -
2822		  pkt_dev->pkt_overhead;
2823	if (datalen < 0 || datalen < sizeof(struct pktgen_hdr))
2824		datalen = sizeof(struct pktgen_hdr);
2825
2826	udph->source = htons(pkt_dev->cur_udp_src);
2827	udph->dest = htons(pkt_dev->cur_udp_dst);
2828	udph->len = htons(datalen + 8);	/* DATA + udphdr */
2829	udph->check = 0;
2830
2831	iph->ihl = 5;
2832	iph->version = 4;
2833	iph->ttl = 32;
2834	iph->tos = pkt_dev->tos;
2835	iph->protocol = IPPROTO_UDP;	/* UDP */
2836	iph->saddr = pkt_dev->cur_saddr;
2837	iph->daddr = pkt_dev->cur_daddr;
2838	iph->id = htons(pkt_dev->ip_id);
2839	pkt_dev->ip_id++;
2840	iph->frag_off = 0;
2841	iplen = 20 + 8 + datalen;
2842	iph->tot_len = htons(iplen);
2843	ip_send_check(iph);
2844	skb->protocol = protocol;
2845	skb->dev = odev;
2846	skb->pkt_type = PACKET_HOST;
2847
2848	pktgen_finalize_skb(pkt_dev, skb, datalen);
2849
2850	if (!(pkt_dev->flags & F_UDPCSUM)) {
2851		skb->ip_summed = CHECKSUM_NONE;
2852	} else if (odev->features & NETIF_F_V4_CSUM) {
2853		skb->ip_summed = CHECKSUM_PARTIAL;
2854		skb->csum = 0;
2855		udp4_hwcsum(skb, iph->saddr, iph->daddr);
2856	} else {
2857		__wsum csum = skb_checksum(skb, skb_transport_offset(skb), datalen + 8, 0);
2858
2859		/* add protocol-dependent pseudo-header */
2860		udph->check = csum_tcpudp_magic(iph->saddr, iph->daddr,
2861						datalen + 8, IPPROTO_UDP, csum);
2862
2863		if (udph->check == 0)
2864			udph->check = CSUM_MANGLED_0;
2865	}
2866
2867#ifdef CONFIG_XFRM
2868	if (!process_ipsec(pkt_dev, skb, protocol))
2869		return NULL;
2870#endif
2871
2872	return skb;
2873}
2874
2875static struct sk_buff *fill_packet_ipv6(struct net_device *odev,
2876					struct pktgen_dev *pkt_dev)
2877{
2878	struct sk_buff *skb = NULL;
2879	__u8 *eth;
2880	struct udphdr *udph;
2881	int datalen, udplen;
2882	struct ipv6hdr *iph;
2883	__be16 protocol = htons(ETH_P_IPV6);
2884	__be32 *mpls;
2885	__be16 *vlan_tci = NULL;                 /* Encapsulates priority and VLAN ID */
2886	__be16 *vlan_encapsulated_proto = NULL;  /* packet type ID field (or len) for VLAN tag */
2887	__be16 *svlan_tci = NULL;                /* Encapsulates priority and SVLAN ID */
2888	__be16 *svlan_encapsulated_proto = NULL; /* packet type ID field (or len) for SVLAN tag */
2889	u16 queue_map;
2890
2891	if (pkt_dev->nr_labels)
2892		protocol = htons(ETH_P_MPLS_UC);
2893
2894	if (pkt_dev->vlan_id != 0xffff)
2895		protocol = htons(ETH_P_8021Q);
2896
2897	/* Update any of the values, used when we're incrementing various
2898	 * fields.
2899	 */
2900	mod_cur_headers(pkt_dev);
2901	queue_map = pkt_dev->cur_queue_map;
2902
2903	skb = pktgen_alloc_skb(odev, pkt_dev, 16);
2904	if (!skb) {
2905		sprintf(pkt_dev->result, "No memory");
2906		return NULL;
2907	}
2908
2909	prefetchw(skb->data);
2910	skb_reserve(skb, 16);
2911
2912	/*  Reserve for ethernet and IP header  */
2913	eth = (__u8 *) skb_push(skb, 14);
2914	mpls = (__be32 *)skb_put(skb, pkt_dev->nr_labels*sizeof(__u32));
2915	if (pkt_dev->nr_labels)
2916		mpls_push(mpls, pkt_dev);
2917
2918	if (pkt_dev->vlan_id != 0xffff) {
2919		if (pkt_dev->svlan_id != 0xffff) {
2920			svlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2921			*svlan_tci = build_tci(pkt_dev->svlan_id,
2922					       pkt_dev->svlan_cfi,
2923					       pkt_dev->svlan_p);
2924			svlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2925			*svlan_encapsulated_proto = htons(ETH_P_8021Q);
2926		}
2927		vlan_tci = (__be16 *)skb_put(skb, sizeof(__be16));
2928		*vlan_tci = build_tci(pkt_dev->vlan_id,
2929				      pkt_dev->vlan_cfi,
2930				      pkt_dev->vlan_p);
2931		vlan_encapsulated_proto = (__be16 *)skb_put(skb, sizeof(__be16));
2932		*vlan_encapsulated_proto = htons(ETH_P_IPV6);
2933	}
2934
2935	skb_set_mac_header(skb, 0);
2936	skb_set_network_header(skb, skb->len);
2937	iph = (struct ipv6hdr *) skb_put(skb, sizeof(struct ipv6hdr));
2938
2939	skb_set_transport_header(skb, skb->len);
2940	udph = (struct udphdr *) skb_put(skb, sizeof(struct udphdr));
2941	skb_set_queue_mapping(skb, queue_map);
2942	skb->priority = pkt_dev->skb_priority;
2943
2944	memcpy(eth, pkt_dev->hh, 12);
2945	*(__be16 *) &eth[12] = protocol;
2946
2947	/* Eth + IPh + UDPh + mpls */
2948	datalen = pkt_dev->cur_pkt_size - 14 -
2949		  sizeof(struct ipv6hdr) - sizeof(struct udphdr) -
2950		  pkt_dev->pkt_overhead;
2951
2952	if (datalen < 0 || datalen < sizeof(struct pktgen_hdr)) {
2953		datalen = sizeof(struct pktgen_hdr);
2954		net_info_ratelimited("increased datalen to %d\n", datalen);
2955	}
2956
2957	udplen = datalen + sizeof(struct udphdr);
2958	udph->source = htons(pkt_dev->cur_udp_src);
2959	udph->dest = htons(pkt_dev->cur_udp_dst);
2960	udph->len = htons(udplen);
2961	udph->check = 0;
2962
2963	*(__be32 *) iph = htonl(0x60000000);	/* Version + flow */
2964
2965	if (pkt_dev->traffic_class) {
2966		/* Version + traffic class + flow (0) */
2967		*(__be32 *)iph |= htonl(0x60000000 | (pkt_dev->traffic_class << 20));
2968	}
2969
2970	iph->hop_limit = 32;
2971
2972	iph->payload_len = htons(udplen);
2973	iph->nexthdr = IPPROTO_UDP;
2974
2975	iph->daddr = pkt_dev->cur_in6_daddr;
2976	iph->saddr = pkt_dev->cur_in6_saddr;
2977
2978	skb->protocol = protocol;
2979	skb->dev = odev;
2980	skb->pkt_type = PACKET_HOST;
2981
2982	pktgen_finalize_skb(pkt_dev, skb, datalen);
2983
2984	if (!(pkt_dev->flags & F_UDPCSUM)) {
2985		skb->ip_summed = CHECKSUM_NONE;
2986	} else if (odev->features & NETIF_F_V6_CSUM) {
2987		skb->ip_summed = CHECKSUM_PARTIAL;
2988		skb->csum_start = skb_transport_header(skb) - skb->head;
2989		skb->csum_offset = offsetof(struct udphdr, check);
2990		udph->check = ~csum_ipv6_magic(&iph->saddr, &iph->daddr, udplen, IPPROTO_UDP, 0);
2991	} else {
2992		__wsum csum = skb_checksum(skb, skb_transport_offset(skb), udplen, 0);
2993
2994		/* add protocol-dependent pseudo-header */
2995		udph->check = csum_ipv6_magic(&iph->saddr, &iph->daddr, udplen, IPPROTO_UDP, csum);
2996
2997		if (udph->check == 0)
2998			udph->check = CSUM_MANGLED_0;
2999	}
3000
3001	return skb;
3002}
3003
3004static struct sk_buff *fill_packet(struct net_device *odev,
3005				   struct pktgen_dev *pkt_dev)
3006{
3007	if (pkt_dev->flags & F_IPV6)
3008		return fill_packet_ipv6(odev, pkt_dev);
3009	else
3010		return fill_packet_ipv4(odev, pkt_dev);
3011}
3012
3013static void pktgen_clear_counters(struct pktgen_dev *pkt_dev)
3014{
3015	pkt_dev->seq_num = 1;
3016	pkt_dev->idle_acc = 0;
3017	pkt_dev->sofar = 0;
3018	pkt_dev->tx_bytes = 0;
3019	pkt_dev->errors = 0;
3020}
3021
3022/* Set up structure for sending pkts, clear counters */
3023
3024static void pktgen_run(struct pktgen_thread *t)
3025{
3026	struct pktgen_dev *pkt_dev;
3027	int started = 0;
3028
3029	func_enter();
3030
3031	rcu_read_lock();
3032	list_for_each_entry_rcu(pkt_dev, &t->if_list, list) {
3033
3034		/*
3035		 * setup odev and create initial packet.
3036		 */
3037		pktgen_setup_inject(pkt_dev);
3038
3039		if (pkt_dev->odev) {
3040			pktgen_clear_counters(pkt_dev);
3041			pkt_dev->skb = NULL;
3042			pkt_dev->started_at = pkt_dev->next_tx = ktime_get();
3043
3044			set_pkt_overhead(pkt_dev);
3045
3046			strcpy(pkt_dev->result, "Starting");
3047			pkt_dev->running = 1;	/* Cranke yeself! */
3048			started++;
3049		} else
3050			strcpy(pkt_dev->result, "Error starting");
3051	}
3052	rcu_read_unlock();
3053	if (started)
3054		t->control &= ~(T_STOP);
3055}
3056
3057static void pktgen_stop_all_threads_ifs(struct pktgen_net *pn)
3058{
3059	struct pktgen_thread *t;
3060
3061	func_enter();
3062
3063	mutex_lock(&pktgen_thread_lock);
3064
3065	list_for_each_entry(t, &pn->pktgen_threads, th_list)
3066		t->control |= T_STOP;
3067
3068	mutex_unlock(&pktgen_thread_lock);
3069}
3070
3071static int thread_is_running(const struct pktgen_thread *t)
3072{
3073	const struct pktgen_dev *pkt_dev;
3074
3075	rcu_read_lock();
3076	list_for_each_entry_rcu(pkt_dev, &t->if_list, list)
3077		if (pkt_dev->running) {
3078			rcu_read_unlock();
3079			return 1;
3080		}
3081	rcu_read_unlock();
3082	return 0;
3083}
3084
3085static int pktgen_wait_thread_run(struct pktgen_thread *t)
3086{
3087	while (thread_is_running(t)) {
3088
3089		msleep_interruptible(100);
3090
3091		if (signal_pending(current))
3092			goto signal;
3093	}
3094	return 1;
3095signal:
3096	return 0;
3097}
3098
3099static int pktgen_wait_all_threads_run(struct pktgen_net *pn)
3100{
3101	struct pktgen_thread *t;
3102	int sig = 1;
3103
3104	mutex_lock(&pktgen_thread_lock);
3105
3106	list_for_each_entry(t, &pn->pktgen_threads, th_list) {
3107		sig = pktgen_wait_thread_run(t);
3108		if (sig == 0)
3109			break;
3110	}
3111
3112	if (sig == 0)
3113		list_for_each_entry(t, &pn->pktgen_threads, th_list)
3114			t->control |= (T_STOP);
3115
3116	mutex_unlock(&pktgen_thread_lock);
3117	return sig;
3118}
3119
3120static void pktgen_run_all_threads(struct pktgen_net *pn)
3121{
3122	struct pktgen_thread *t;
3123
3124	func_enter();
3125
3126	mutex_lock(&pktgen_thread_lock);
3127
3128	list_for_each_entry(t, &pn->pktgen_threads, th_list)
3129		t->control |= (T_RUN);
3130
3131	mutex_unlock(&pktgen_thread_lock);
3132
3133	/* Propagate thread->control  */
3134	schedule_timeout_interruptible(msecs_to_jiffies(125));
3135
3136	pktgen_wait_all_threads_run(pn);
3137}
3138
3139static void pktgen_reset_all_threads(struct pktgen_net *pn)
3140{
3141	struct pktgen_thread *t;
3142
3143	func_enter();
3144
3145	mutex_lock(&pktgen_thread_lock);
3146
3147	list_for_each_entry(t, &pn->pktgen_threads, th_list)
3148		t->control |= (T_REMDEVALL);
3149
3150	mutex_unlock(&pktgen_thread_lock);
3151
3152	/* Propagate thread->control  */
3153	schedule_timeout_interruptible(msecs_to_jiffies(125));
3154
3155	pktgen_wait_all_threads_run(pn);
3156}
3157
3158static void show_results(struct pktgen_dev *pkt_dev, int nr_frags)
3159{
3160	__u64 bps, mbps, pps;
3161	char *p = pkt_dev->result;
3162	ktime_t elapsed = ktime_sub(pkt_dev->stopped_at,
3163				    pkt_dev->started_at);
3164	ktime_t idle = ns_to_ktime(pkt_dev->idle_acc);
3165
3166	p += sprintf(p, "OK: %llu(c%llu+d%llu) usec, %llu (%dbyte,%dfrags)\n",
3167		     (unsigned long long)ktime_to_us(elapsed),
3168		     (unsigned long long)ktime_to_us(ktime_sub(elapsed, idle)),
3169		     (unsigned long long)ktime_to_us(idle),
3170		     (unsigned long long)pkt_dev->sofar,
3171		     pkt_dev->cur_pkt_size, nr_frags);
3172
3173	pps = div64_u64(pkt_dev->sofar * NSEC_PER_SEC,
3174			ktime_to_ns(elapsed));
3175
3176	bps = pps * 8 * pkt_dev->cur_pkt_size;
3177
3178	mbps = bps;
3179	do_div(mbps, 1000000);
3180	p += sprintf(p, "  %llupps %lluMb/sec (%llubps) errors: %llu",
3181		     (unsigned long long)pps,
3182		     (unsigned long long)mbps,
3183		     (unsigned long long)bps,
3184		     (unsigned long long)pkt_dev->errors);
3185}
3186
3187/* Set stopped-at timer, remove from running list, do counters & statistics */
3188static int pktgen_stop_device(struct pktgen_dev *pkt_dev)
3189{
3190	int nr_frags = pkt_dev->skb ? skb_shinfo(pkt_dev->skb)->nr_frags : -1;
3191
3192	if (!pkt_dev->running) {
3193		pr_warn("interface: %s is already stopped\n",
3194			pkt_dev->odevname);
3195		return -EINVAL;
3196	}
3197
3198	pkt_dev->running = 0;
3199	kfree_skb(pkt_dev->skb);
3200	pkt_dev->skb = NULL;
3201	pkt_dev->stopped_at = ktime_get();
3202
3203	show_results(pkt_dev, nr_frags);
3204
3205	return 0;
3206}
3207
3208static struct pktgen_dev *next_to_run(struct pktgen_thread *t)
3209{
3210	struct pktgen_dev *pkt_dev, *best = NULL;
3211
3212	rcu_read_lock();
3213	list_for_each_entry_rcu(pkt_dev, &t->if_list, list) {
3214		if (!pkt_dev->running)
3215			continue;
3216		if (best == NULL)
3217			best = pkt_dev;
3218		else if (ktime_compare(pkt_dev->next_tx, best->next_tx) < 0)
3219			best = pkt_dev;
3220	}
3221	rcu_read_unlock();
3222
3223	return best;
3224}
3225
3226static void pktgen_stop(struct pktgen_thread *t)
3227{
3228	struct pktgen_dev *pkt_dev;
3229
3230	func_enter();
3231
3232	rcu_read_lock();
3233
3234	list_for_each_entry_rcu(pkt_dev, &t->if_list, list) {
3235		pktgen_stop_device(pkt_dev);
3236	}
3237
3238	rcu_read_unlock();
3239}
3240
3241/*
3242 * one of our devices needs to be removed - find it
3243 * and remove it
3244 */
3245static void pktgen_rem_one_if(struct pktgen_thread *t)
3246{
3247	struct list_head *q, *n;
3248	struct pktgen_dev *cur;
3249
3250	func_enter();
3251
3252	list_for_each_safe(q, n, &t->if_list) {
3253		cur = list_entry(q, struct pktgen_dev, list);
3254
3255		if (!cur->removal_mark)
3256			continue;
3257
3258		kfree_skb(cur->skb);
3259		cur->skb = NULL;
3260
3261		pktgen_remove_device(t, cur);
3262
3263		break;
3264	}
3265}
3266
3267static void pktgen_rem_all_ifs(struct pktgen_thread *t)
3268{
3269	struct list_head *q, *n;
3270	struct pktgen_dev *cur;
3271
3272	func_enter();
3273
3274	/* Remove all devices, free mem */
3275
3276	list_for_each_safe(q, n, &t->if_list) {
3277		cur = list_entry(q, struct pktgen_dev, list);
3278
3279		kfree_skb(cur->skb);
3280		cur->skb = NULL;
3281
3282		pktgen_remove_device(t, cur);
3283	}
3284}
3285
3286static void pktgen_rem_thread(struct pktgen_thread *t)
3287{
3288	/* Remove from the thread list */
3289	remove_proc_entry(t->tsk->comm, t->net->proc_dir);
3290}
3291
3292static void pktgen_resched(struct pktgen_dev *pkt_dev)
3293{
3294	ktime_t idle_start = ktime_get();
3295	schedule();
3296	pkt_dev->idle_acc += ktime_to_ns(ktime_sub(ktime_get(), idle_start));
3297}
3298
3299static void pktgen_wait_for_skb(struct pktgen_dev *pkt_dev)
3300{
3301	ktime_t idle_start = ktime_get();
3302
3303	while (atomic_read(&(pkt_dev->skb->users)) != 1) {
3304		if (signal_pending(current))
3305			break;
3306
3307		if (need_resched())
3308			pktgen_resched(pkt_dev);
3309		else
3310			cpu_relax();
3311	}
3312	pkt_dev->idle_acc += ktime_to_ns(ktime_sub(ktime_get(), idle_start));
3313}
3314
3315static void pktgen_xmit(struct pktgen_dev *pkt_dev)
3316{
3317	unsigned int burst = ACCESS_ONCE(pkt_dev->burst);
3318	struct net_device *odev = pkt_dev->odev;
3319	struct netdev_queue *txq;
3320	int ret;
3321
3322	/* If device is offline, then don't send */
3323	if (unlikely(!netif_running(odev) || !netif_carrier_ok(odev))) {
3324		pktgen_stop_device(pkt_dev);
3325		return;
3326	}
3327
3328	/* This is max DELAY, this has special meaning of
3329	 * "never transmit"
3330	 */
3331	if (unlikely(pkt_dev->delay == ULLONG_MAX)) {
3332		pkt_dev->next_tx = ktime_add_ns(ktime_get(), ULONG_MAX);
3333		return;
3334	}
3335
3336	/* If no skb or clone count exhausted then get new one */
3337	if (!pkt_dev->skb || (pkt_dev->last_ok &&
3338			      ++pkt_dev->clone_count >= pkt_dev->clone_skb)) {
3339		/* build a new pkt */
3340		kfree_skb(pkt_dev->skb);
3341
3342		pkt_dev->skb = fill_packet(odev, pkt_dev);
3343		if (pkt_dev->skb == NULL) {
3344			pr_err("ERROR: couldn't allocate skb in fill_packet\n");
3345			schedule();
3346			pkt_dev->clone_count--;	/* back out increment, OOM */
3347			return;
3348		}
3349		pkt_dev->last_pkt_size = pkt_dev->skb->len;
3350		pkt_dev->allocated_skbs++;
3351		pkt_dev->clone_count = 0;	/* reset counter */
3352	}
3353
3354	if (pkt_dev->delay && pkt_dev->last_ok)
3355		spin(pkt_dev, pkt_dev->next_tx);
3356
3357	txq = skb_get_tx_queue(odev, pkt_dev->skb);
3358
3359	local_bh_disable();
3360
3361	HARD_TX_LOCK(odev, txq, smp_processor_id());
3362
3363	if (unlikely(netif_xmit_frozen_or_drv_stopped(txq))) {
3364		ret = NETDEV_TX_BUSY;
3365		pkt_dev->last_ok = 0;
3366		goto unlock;
3367	}
3368	atomic_add(burst, &pkt_dev->skb->users);
3369
3370xmit_more:
3371	ret = netdev_start_xmit(pkt_dev->skb, odev, txq, --burst > 0);
3372
3373	switch (ret) {
3374	case NETDEV_TX_OK:
3375		pkt_dev->last_ok = 1;
3376		pkt_dev->sofar++;
3377		pkt_dev->seq_num++;
3378		pkt_dev->tx_bytes += pkt_dev->last_pkt_size;
3379		if (burst > 0 && !netif_xmit_frozen_or_drv_stopped(txq))
3380			goto xmit_more;
3381		break;
3382	case NET_XMIT_DROP:
3383	case NET_XMIT_CN:
3384	case NET_XMIT_POLICED:
3385		/* skb has been consumed */
3386		pkt_dev->errors++;
3387		break;
3388	default: /* Drivers are not supposed to return other values! */
3389		net_info_ratelimited("%s xmit error: %d\n",
3390				     pkt_dev->odevname, ret);
3391		pkt_dev->errors++;
3392		/* fallthru */
3393	case NETDEV_TX_LOCKED:
3394	case NETDEV_TX_BUSY:
3395		/* Retry it next time */
3396		atomic_dec(&(pkt_dev->skb->users));
3397		pkt_dev->last_ok = 0;
3398	}
3399	if (unlikely(burst))
3400		atomic_sub(burst, &pkt_dev->skb->users);
3401unlock:
3402	HARD_TX_UNLOCK(odev, txq);
3403
3404	local_bh_enable();
3405
3406	/* If pkt_dev->count is zero, then run forever */
3407	if ((pkt_dev->count != 0) && (pkt_dev->sofar >= pkt_dev->count)) {
3408		pktgen_wait_for_skb(pkt_dev);
3409
3410		/* Done with this */
3411		pktgen_stop_device(pkt_dev);
3412	}
3413}
3414
3415/*
3416 * Main loop of the thread goes here
3417 */
3418
3419static int pktgen_thread_worker(void *arg)
3420{
3421	DEFINE_WAIT(wait);
3422	struct pktgen_thread *t = arg;
3423	struct pktgen_dev *pkt_dev = NULL;
3424	int cpu = t->cpu;
3425
3426	BUG_ON(smp_processor_id() != cpu);
3427
3428	init_waitqueue_head(&t->queue);
3429	complete(&t->start_done);
3430
3431	pr_debug("starting pktgen/%d:  pid=%d\n", cpu, task_pid_nr(current));
3432
3433	set_freezable();
3434
3435	__set_current_state(TASK_RUNNING);
3436
3437	while (!kthread_should_stop()) {
3438		pkt_dev = next_to_run(t);
3439
3440		if (unlikely(!pkt_dev && t->control == 0)) {
3441			if (t->net->pktgen_exiting)
3442				break;
3443			wait_event_interruptible_timeout(t->queue,
3444							 t->control != 0,
3445							 HZ/10);
3446			try_to_freeze();
3447			continue;
3448		}
3449
3450		if (likely(pkt_dev)) {
3451			pktgen_xmit(pkt_dev);
3452
3453			if (need_resched())
3454				pktgen_resched(pkt_dev);
3455			else
3456				cpu_relax();
3457		}
3458
3459		if (t->control & T_STOP) {
3460			pktgen_stop(t);
3461			t->control &= ~(T_STOP);
3462		}
3463
3464		if (t->control & T_RUN) {
3465			pktgen_run(t);
3466			t->control &= ~(T_RUN);
3467		}
3468
3469		if (t->control & T_REMDEVALL) {
3470			pktgen_rem_all_ifs(t);
3471			t->control &= ~(T_REMDEVALL);
3472		}
3473
3474		if (t->control & T_REMDEV) {
3475			pktgen_rem_one_if(t);
3476			t->control &= ~(T_REMDEV);
3477		}
3478
3479		try_to_freeze();
3480	}
3481	set_current_state(TASK_INTERRUPTIBLE);
3482
3483	pr_debug("%s stopping all device\n", t->tsk->comm);
3484	pktgen_stop(t);
3485
3486	pr_debug("%s removing all device\n", t->tsk->comm);
3487	pktgen_rem_all_ifs(t);
3488
3489	pr_debug("%s removing thread\n", t->tsk->comm);
3490	pktgen_rem_thread(t);
3491
3492	/* Wait for kthread_stop */
3493	for (;;) {
3494		set_current_state(TASK_INTERRUPTIBLE);
3495		if (kthread_should_stop())
3496			break;
3497		schedule();
3498	}
3499	__set_current_state(TASK_RUNNING);
3500
3501	return 0;
3502}
3503
3504static struct pktgen_dev *pktgen_find_dev(struct pktgen_thread *t,
3505					  const char *ifname, bool exact)
3506{
3507	struct pktgen_dev *p, *pkt_dev = NULL;
3508	size_t len = strlen(ifname);
3509
3510	rcu_read_lock();
3511	list_for_each_entry_rcu(p, &t->if_list, list)
3512		if (strncmp(p->odevname, ifname, len) == 0) {
3513			if (p->odevname[len]) {
3514				if (exact || p->odevname[len] != '@')
3515					continue;
3516			}
3517			pkt_dev = p;
3518			break;
3519		}
3520
3521	rcu_read_unlock();
3522	pr_debug("find_dev(%s) returning %p\n", ifname, pkt_dev);
3523	return pkt_dev;
3524}
3525
3526/*
3527 * Adds a dev at front of if_list.
3528 */
3529
3530static int add_dev_to_thread(struct pktgen_thread *t,
3531			     struct pktgen_dev *pkt_dev)
3532{
3533	int rv = 0;
3534
3535	/* This function cannot be called concurrently, as its called
3536	 * under pktgen_thread_lock mutex, but it can run from
3537	 * userspace on another CPU than the kthread.  The if_lock()
3538	 * is used here to sync with concurrent instances of
3539	 * _rem_dev_from_if_list() invoked via kthread, which is also
3540	 * updating the if_list */
3541	if_lock(t);
3542
3543	if (pkt_dev->pg_thread) {
3544		pr_err("ERROR: already assigned to a thread\n");
3545		rv = -EBUSY;
3546		goto out;
3547	}
3548
3549	pkt_dev->running = 0;
3550	pkt_dev->pg_thread = t;
3551	list_add_rcu(&pkt_dev->list, &t->if_list);
3552
3553out:
3554	if_unlock(t);
3555	return rv;
3556}
3557
3558/* Called under thread lock */
3559
3560static int pktgen_add_device(struct pktgen_thread *t, const char *ifname)
3561{
3562	struct pktgen_dev *pkt_dev;
3563	int err;
3564	int node = cpu_to_node(t->cpu);
3565
3566	/* We don't allow a device to be on several threads */
3567
3568	pkt_dev = __pktgen_NN_threads(t->net, ifname, FIND);
3569	if (pkt_dev) {
3570		pr_err("ERROR: interface already used\n");
3571		return -EBUSY;
3572	}
3573
3574	pkt_dev = kzalloc_node(sizeof(struct pktgen_dev), GFP_KERNEL, node);
3575	if (!pkt_dev)
3576		return -ENOMEM;
3577
3578	strcpy(pkt_dev->odevname, ifname);
3579	pkt_dev->flows = vzalloc_node(MAX_CFLOWS * sizeof(struct flow_state),
3580				      node);
3581	if (pkt_dev->flows == NULL) {
3582		kfree(pkt_dev);
3583		return -ENOMEM;
3584	}
3585
3586	pkt_dev->removal_mark = 0;
3587	pkt_dev->nfrags = 0;
3588	pkt_dev->delay = pg_delay_d;
3589	pkt_dev->count = pg_count_d;
3590	pkt_dev->sofar = 0;
3591	pkt_dev->udp_src_min = 9;	/* sink port */
3592	pkt_dev->udp_src_max = 9;
3593	pkt_dev->udp_dst_min = 9;
3594	pkt_dev->udp_dst_max = 9;
3595	pkt_dev->vlan_p = 0;
3596	pkt_dev->vlan_cfi = 0;
3597	pkt_dev->vlan_id = 0xffff;
3598	pkt_dev->svlan_p = 0;
3599	pkt_dev->svlan_cfi = 0;
3600	pkt_dev->svlan_id = 0xffff;
3601	pkt_dev->burst = 1;
3602	pkt_dev->node = -1;
3603
3604	err = pktgen_setup_dev(t->net, pkt_dev, ifname);
3605	if (err)
3606		goto out1;
3607	if (pkt_dev->odev->priv_flags & IFF_TX_SKB_SHARING)
3608		pkt_dev->clone_skb = pg_clone_skb_d;
3609
3610	pkt_dev->entry = proc_create_data(ifname, 0600, t->net->proc_dir,
3611					  &pktgen_if_fops, pkt_dev);
3612	if (!pkt_dev->entry) {
3613		pr_err("cannot create %s/%s procfs entry\n",
3614		       PG_PROC_DIR, ifname);
3615		err = -EINVAL;
3616		goto out2;
3617	}
3618#ifdef CONFIG_XFRM
3619	pkt_dev->ipsmode = XFRM_MODE_TRANSPORT;
3620	pkt_dev->ipsproto = IPPROTO_ESP;
3621
3622	/* xfrm tunnel mode needs additional dst to extract outter
3623	 * ip header protocol/ttl/id field, here creat a phony one.
3624	 * instead of looking for a valid rt, which definitely hurting
3625	 * performance under such circumstance.
3626	 */
3627	pkt_dev->dstops.family = AF_INET;
3628	pkt_dev->dst.dev = pkt_dev->odev;
3629	dst_init_metrics(&pkt_dev->dst, pktgen_dst_metrics, false);
3630	pkt_dev->dst.child = &pkt_dev->dst;
3631	pkt_dev->dst.ops = &pkt_dev->dstops;
3632#endif
3633
3634	return add_dev_to_thread(t, pkt_dev);
3635out2:
3636	dev_put(pkt_dev->odev);
3637out1:
3638#ifdef CONFIG_XFRM
3639	free_SAs(pkt_dev);
3640#endif
3641	vfree(pkt_dev->flows);
3642	kfree(pkt_dev);
3643	return err;
3644}
3645
3646static int __net_init pktgen_create_thread(int cpu, struct pktgen_net *pn)
3647{
3648	struct pktgen_thread *t;
3649	struct proc_dir_entry *pe;
3650	struct task_struct *p;
3651
3652	t = kzalloc_node(sizeof(struct pktgen_thread), GFP_KERNEL,
3653			 cpu_to_node(cpu));
3654	if (!t) {
3655		pr_err("ERROR: out of memory, can't create new thread\n");
3656		return -ENOMEM;
3657	}
3658
3659	spin_lock_init(&t->if_lock);
3660	t->cpu = cpu;
3661
3662	INIT_LIST_HEAD(&t->if_list);
3663
3664	list_add_tail(&t->th_list, &pn->pktgen_threads);
3665	init_completion(&t->start_done);
3666
3667	p = kthread_create_on_node(pktgen_thread_worker,
3668				   t,
3669				   cpu_to_node(cpu),
3670				   "kpktgend_%d", cpu);
3671	if (IS_ERR(p)) {
3672		pr_err("kernel_thread() failed for cpu %d\n", t->cpu);
3673		list_del(&t->th_list);
3674		kfree(t);
3675		return PTR_ERR(p);
3676	}
3677	kthread_bind(p, cpu);
3678	t->tsk = p;
3679
3680	pe = proc_create_data(t->tsk->comm, 0600, pn->proc_dir,
3681			      &pktgen_thread_fops, t);
3682	if (!pe) {
3683		pr_err("cannot create %s/%s procfs entry\n",
3684		       PG_PROC_DIR, t->tsk->comm);
3685		kthread_stop(p);
3686		list_del(&t->th_list);
3687		kfree(t);
3688		return -EINVAL;
3689	}
3690
3691	t->net = pn;
3692	wake_up_process(p);
3693	wait_for_completion(&t->start_done);
3694
3695	return 0;
3696}
3697
3698/*
3699 * Removes a device from the thread if_list.
3700 */
3701static void _rem_dev_from_if_list(struct pktgen_thread *t,
3702				  struct pktgen_dev *pkt_dev)
3703{
3704	struct list_head *q, *n;
3705	struct pktgen_dev *p;
3706
3707	if_lock(t);
3708	list_for_each_safe(q, n, &t->if_list) {
3709		p = list_entry(q, struct pktgen_dev, list);
3710		if (p == pkt_dev)
3711			list_del_rcu(&p->list);
3712	}
3713	if_unlock(t);
3714}
3715
3716static int pktgen_remove_device(struct pktgen_thread *t,
3717				struct pktgen_dev *pkt_dev)
3718{
3719	pr_debug("remove_device pkt_dev=%p\n", pkt_dev);
3720
3721	if (pkt_dev->running) {
3722		pr_warn("WARNING: trying to remove a running interface, stopping it now\n");
3723		pktgen_stop_device(pkt_dev);
3724	}
3725
3726	/* Dis-associate from the interface */
3727
3728	if (pkt_dev->odev) {
3729		dev_put(pkt_dev->odev);
3730		pkt_dev->odev = NULL;
3731	}
3732
3733	/* Remove proc before if_list entry, because add_device uses
3734	 * list to determine if interface already exist, avoid race
3735	 * with proc_create_data() */
3736	proc_remove(pkt_dev->entry);
3737
3738	/* And update the thread if_list */
3739	_rem_dev_from_if_list(t, pkt_dev);
3740
3741#ifdef CONFIG_XFRM
3742	free_SAs(pkt_dev);
3743#endif
3744	vfree(pkt_dev->flows);
3745	if (pkt_dev->page)
3746		put_page(pkt_dev->page);
3747	kfree_rcu(pkt_dev, rcu);
3748	return 0;
3749}
3750
3751static int __net_init pg_net_init(struct net *net)
3752{
3753	struct pktgen_net *pn = net_generic(net, pg_net_id);
3754	struct proc_dir_entry *pe;
3755	int cpu, ret = 0;
3756
3757	pn->net = net;
3758	INIT_LIST_HEAD(&pn->pktgen_threads);
3759	pn->pktgen_exiting = false;
3760	pn->proc_dir = proc_mkdir(PG_PROC_DIR, pn->net->proc_net);
3761	if (!pn->proc_dir) {
3762		pr_warn("cannot create /proc/net/%s\n", PG_PROC_DIR);
3763		return -ENODEV;
3764	}
3765	pe = proc_create(PGCTRL, 0600, pn->proc_dir, &pktgen_fops);
3766	if (pe == NULL) {
3767		pr_err("cannot create %s procfs entry\n", PGCTRL);
3768		ret = -EINVAL;
3769		goto remove;
3770	}
3771
3772	for_each_online_cpu(cpu) {
3773		int err;
3774
3775		err = pktgen_create_thread(cpu, pn);
3776		if (err)
3777			pr_warn("Cannot create thread for cpu %d (%d)\n",
3778				   cpu, err);
3779	}
3780
3781	if (list_empty(&pn->pktgen_threads)) {
3782		pr_err("Initialization failed for all threads\n");
3783		ret = -ENODEV;
3784		goto remove_entry;
3785	}
3786
3787	return 0;
3788
3789remove_entry:
3790	remove_proc_entry(PGCTRL, pn->proc_dir);
3791remove:
3792	remove_proc_entry(PG_PROC_DIR, pn->net->proc_net);
3793	return ret;
3794}
3795
3796static void __net_exit pg_net_exit(struct net *net)
3797{
3798	struct pktgen_net *pn = net_generic(net, pg_net_id);
3799	struct pktgen_thread *t;
3800	struct list_head *q, *n;
3801	LIST_HEAD(list);
3802
3803	/* Stop all interfaces & threads */
3804	pn->pktgen_exiting = true;
3805
3806	mutex_lock(&pktgen_thread_lock);
3807	list_splice_init(&pn->pktgen_threads, &list);
3808	mutex_unlock(&pktgen_thread_lock);
3809
3810	list_for_each_safe(q, n, &list) {
3811		t = list_entry(q, struct pktgen_thread, th_list);
3812		list_del(&t->th_list);
3813		kthread_stop(t->tsk);
3814		kfree(t);
3815	}
3816
3817	remove_proc_entry(PGCTRL, pn->proc_dir);
3818	remove_proc_entry(PG_PROC_DIR, pn->net->proc_net);
3819}
3820
3821static struct pernet_operations pg_net_ops = {
3822	.init = pg_net_init,
3823	.exit = pg_net_exit,
3824	.id   = &pg_net_id,
3825	.size = sizeof(struct pktgen_net),
3826};
3827
3828static int __init pg_init(void)
3829{
3830	int ret = 0;
3831
3832	pr_info("%s", version);
3833	ret = register_pernet_subsys(&pg_net_ops);
3834	if (ret)
3835		return ret;
3836	ret = register_netdevice_notifier(&pktgen_notifier_block);
3837	if (ret)
3838		unregister_pernet_subsys(&pg_net_ops);
3839
3840	return ret;
3841}
3842
3843static void __exit pg_cleanup(void)
3844{
3845	unregister_netdevice_notifier(&pktgen_notifier_block);
3846	unregister_pernet_subsys(&pg_net_ops);
3847	/* Don't need rcu_barrier() due to use of kfree_rcu() */
3848}
3849
3850module_init(pg_init);
3851module_exit(pg_cleanup);
3852
3853MODULE_AUTHOR("Robert Olsson <robert.olsson@its.uu.se>");
3854MODULE_DESCRIPTION("Packet Generator tool");
3855MODULE_LICENSE("GPL");
3856MODULE_VERSION(VERSION);
3857module_param(pg_count_d, int, 0);
3858MODULE_PARM_DESC(pg_count_d, "Default number of packets to inject");
3859module_param(pg_delay_d, int, 0);
3860MODULE_PARM_DESC(pg_delay_d, "Default delay between packets (nanoseconds)");
3861module_param(pg_clone_skb_d, int, 0);
3862MODULE_PARM_DESC(pg_clone_skb_d, "Default number of copies of the same packet");
3863module_param(debug, int, 0);
3864MODULE_PARM_DESC(debug, "Enable debugging of pktgen module");
3865