net/core/pktgen.c

0001 // SPDX-License-Identifier: GPL-2.0-or-later
0002 /*
0003  * Authors:
0004  * Copyright 2001, 2002 by Robert Olsson <robert.olsson@its.uu.se>
0005  *                             Uppsala University and
0006  *                             Swedish University of Agricultural Sciences
0007  *
0008  * Alexey Kuznetsov  <kuznet@ms2.inr.ac.ru>
0009  * Ben Greear <greearb@candelatech.com>
0010  * Jens Låås <jens.laas@data.slu.se>
0011  *
0012  * A tool for loading the network with preconfigurated packets.
0013  * The tool is implemented as a linux module.  Parameters are output
0014  * device, delay (to hard_xmit), number of packets, and whether
0015  * to use multiple SKBs or just the same one.
0016  * pktgen uses the installed interface's output routine.
0017  *
0018  * Additional hacking by:
0019  *
0020  * Jens.Laas@data.slu.se
0021  * Improved by ANK. 010120.
0022  * Improved by ANK even more. 010212.
0023  * MAC address typo fixed. 010417 --ro
0024  * Integrated.  020301 --DaveM
0025  * Added multiskb option 020301 --DaveM
0026  * Scaling of results. 020417--sigurdur@linpro.no
0027  * Significant re-work of the module:
0028  *   *  Convert to threaded model to more efficiently be able to transmit
0029  *       and receive on multiple interfaces at once.
0030  *   *  Converted many counters to __u64 to allow longer runs.
0031  *   *  Allow configuration of ranges, like min/max IP address, MACs,
0032  *       and UDP-ports, for both source and destination, and can
0033  *       set to use a random distribution or sequentially walk the range.
0034  *   *  Can now change most values after starting.
0035  *   *  Place 12-byte packet in UDP payload with magic number,
0036  *       sequence number, and timestamp.
0037  *   *  Add receiver code that detects dropped pkts, re-ordered pkts, and
0038  *       latencies (with micro-second) precision.
0039  *   *  Add IOCTL interface to easily get counters & configuration.
0040  *   --Ben Greear <greearb@candelatech.com>
0041  *
0042  * Renamed multiskb to clone_skb and cleaned up sending core for two distinct
0043  * skb modes. A clone_skb=0 mode for Ben "ranges" work and a clone_skb != 0
0044  * as a "fastpath" with a configurable number of clones after alloc's.
0045  * clone_skb=0 means all packets are allocated this also means ranges time
0046  * stamps etc can be used. clone_skb=100 means 1 malloc is followed by 100
0047  * clones.
0048  *
0049  * Also moved to /proc/net/pktgen/
0050  * --ro
0051  *
0052  * Sept 10:  Fixed threading/locking.  Lots of bone-headed and more clever
0053  *    mistakes.  Also merged in DaveM's patch in the -pre6 patch.
0054  * --Ben Greear <greearb@candelatech.com>
0055  *
0056  * Integrated to 2.5.x 021029 --Lucio Maciel (luciomaciel@zipmail.com.br)
0057  *
0058  * 021124 Finished major redesign and rewrite for new functionality.
0059  * See Documentation/networking/pktgen.rst for how to use this.
0060  *
0061  * The new operation:
0062  * For each CPU one thread/process is created at start. This process checks
0063  * for running devices in the if_list and sends packets until count is 0 it
0064  * also the thread checks the thread->control which is used for inter-process
0065  * communication. controlling process "posts" operations to the threads this
0066  * way.
0067  * The if_list is RCU protected, and the if_lock remains to protect updating
0068  * of if_list, from "add_device" as it invoked from userspace (via proc write).
0069  *
0070  * By design there should only be *one* "controlling" process. In practice
0071  * multiple write accesses gives unpredictable result. Understood by "write"
0072  * to /proc gives result code thats should be read be the "writer".
0073  * For practical use this should be no problem.
0074  *
0075  * Note when adding devices to a specific CPU there good idea to also assign
0076  * /proc/irq/XX/smp_affinity so TX-interrupts gets bound to the same CPU.
0077  * --ro
0078  *
0079  * Fix refcount off by one if first packet fails, potential null deref,
0080  * memleak 030710- KJP
0081  *
0082  * First "ranges" functionality for ipv6 030726 --ro
0083  *
0084  * Included flow support. 030802 ANK.
0085  *
0086  * Fixed unaligned access on IA-64 Grant Grundler <grundler@parisc-linux.org>
0087  *
0088  * Remove if fix from added Harald Welte <laforge@netfilter.org> 040419
0089  * ia64 compilation fix from  Aron Griffis <aron@hp.com> 040604
0090  *
0091  * New xmit() return, do_div and misc clean up by Stephen Hemminger
0092  * <shemminger@osdl.org> 040923
0093  *
0094  * Randy Dunlap fixed u64 printk compiler warning
0095  *
0096  * Remove FCS from BW calculation.  Lennert Buytenhek <buytenh@wantstofly.org>
0097  * New time handling. Lennert Buytenhek <buytenh@wantstofly.org> 041213
0098  *
0099  * Corrections from Nikolai Malykh (nmalykh@bilim.com)
0100  * Removed unused flags F_SET_SRCMAC & F_SET_SRCIP 041230
0101  *
0102  * interruptible_sleep_on_timeout() replaced Nishanth Aravamudan <nacc@us.ibm.com>
0103  * 050103
0104  *
0105  * MPLS support by Steven Whitehouse <steve@chygwyn.com>
0106  *
0107  * 802.1Q/Q-in-Q support by Francesco Fondelli (FF) <francesco.fondelli@gmail.com>
0108  *
0109  * Fixed src_mac command to set source mac of packet to value specified in
0110  * command by Adit Ranadive <adit.262@gmail.com>
0111  */
0112
0113 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
0114
0115 #include <linux/sys.h>
0116 #include <linux/types.h>
0117 #include <linux/module.h>
0118 #include <linux/moduleparam.h>
0119 #include <linux/kernel.h>
0120 #include <linux/mutex.h>
0121 #include <linux/sched.h>
0122 #include <linux/slab.h>
0123 #include <linux/vmalloc.h>
0124 #include <linux/unistd.h>
0125 #include <linux/string.h>
0126 #include <linux/ptrace.h>
0127 #include <linux/errno.h>
0128 #include <linux/ioport.h>
0129 #include <linux/interrupt.h>
0130 #include <linux/capability.h>
0131 #include <linux/hrtimer.h>
0132 #include <linux/freezer.h>
0133 #include <linux/delay.h>
0134 #include <linux/timer.h>
0135 #include <linux/list.h>
0136 #include <linux/init.h>
0137 #include <linux/skbuff.h>
0138 #include <linux/netdevice.h>
0139 #include <linux/inet.h>
0140 #include <linux/inetdevice.h>
0141 #include <linux/rtnetlink.h>
0142 #include <linux/if_arp.h>
0143 #include <linux/if_vlan.h>
0144 #include <linux/in.h>
0145 #include <linux/ip.h>
0146 #include <linux/ipv6.h>
0147 #include <linux/udp.h>
0148 #include <linux/proc_fs.h>
0149 #include <linux/seq_file.h>
0150 #include <linux/wait.h>
0151 #include <linux/etherdevice.h>
0152 #include <linux/kthread.h>
0153 #include <linux/prefetch.h>
0154 #include <linux/mmzone.h>
0155 #include <net/net_namespace.h>
0156 #include <net/checksum.h>
0157 #include <net/ipv6.h>
0158 #include <net/udp.h>
0159 #include <net/ip6_checksum.h>
0160 #include <net/addrconf.h>
0161 #ifdef CONFIG_XFRM
0162 #include <net/xfrm.h>
0163 #endif
0164 #include <net/netns/generic.h>
0165 #include <asm/byteorder.h>
0166 #include <linux/rcupdate.h>
0167 #include <linux/bitops.h>
0168 #include <linux/io.h>
0169 #include <linux/timex.h>
0170 #include <linux/uaccess.h>
0171 #include <asm/dma.h>
0172 #include <asm/div64.h>      /* do_div */
0173
0174 #define VERSION "2.75"
0175 #define IP_NAME_SZ 32
0176 #define MAX_MPLS_LABELS 16 /* This is the max label stack depth */
0177 #define MPLS_STACK_BOTTOM htonl(0x00000100)
0178 /* Max number of internet mix entries that can be specified in imix_weights. */
0179 #define MAX_IMIX_ENTRIES 20
0180 #define IMIX_PRECISION 100 /* Precision of IMIX distribution */
0181
0182 #define func_enter() pr_debug("entering %s\n", __func__);
0183
0184 #define PKT_FLAGS                           \
0185     pf(IPV6)        /* Interface in IPV6 Mode */        \
0186     pf(IPSRC_RND)       /* IP-Src Random  */            \
0187     pf(IPDST_RND)       /* IP-Dst Random  */            \
0188     pf(TXSIZE_RND)      /* Transmit size is random */       \
0189     pf(UDPSRC_RND)      /* UDP-Src Random */            \
0190     pf(UDPDST_RND)      /* UDP-Dst Random */            \
0191     pf(UDPCSUM)     /* Include UDP checksum */      \
0192     pf(NO_TIMESTAMP)    /* Don't timestamp packets (default TS) */ \
0193     pf(MPLS_RND)        /* Random MPLS labels */        \
0194     pf(QUEUE_MAP_RND)   /* queue map Random */          \
0195     pf(QUEUE_MAP_CPU)   /* queue map mirrors smp_processor_id() */ \
0196     pf(FLOW_SEQ)        /* Sequential flows */          \
0197     pf(IPSEC)       /* ipsec on for flows */        \
0198     pf(MACSRC_RND)      /* MAC-Src Random */            \
0199     pf(MACDST_RND)      /* MAC-Dst Random */            \
0200     pf(VID_RND)     /* Random VLAN ID */            \
0201     pf(SVID_RND)        /* Random SVLAN ID */           \
0202     pf(NODE)        /* Node memory alloc*/          \
0203
0204 #define pf(flag)        flag##_SHIFT,
0205 enum pkt_flags {
0206     PKT_FLAGS
0207 };
0208 #undef pf
0209
0210 /* Device flag bits */
0211 #define pf(flag)        static const __u32 F_##flag = (1<<flag##_SHIFT);
0212 PKT_FLAGS
0213 #undef pf
0214
0215 #define pf(flag)        __stringify(flag),
0216 static char *pkt_flag_names[] = {
0217     PKT_FLAGS
0218 };
0219 #undef pf
0220
0221 #define NR_PKT_FLAGS        ARRAY_SIZE(pkt_flag_names)
0222
0223 /* Thread control flag bits */
0224 #define T_STOP        (1<<0)    /* Stop run */
0225 #define T_RUN         (1<<1)    /* Start run */
0226 #define T_REMDEVALL   (1<<2)    /* Remove all devs */
0227 #define T_REMDEV      (1<<3)    /* Remove one dev */
0228
0229 /* Xmit modes */
0230 #define M_START_XMIT        0   /* Default normal TX */
0231 #define M_NETIF_RECEIVE     1   /* Inject packets into stack */
0232 #define M_QUEUE_XMIT        2   /* Inject packet into qdisc */
0233
0234 /* If lock -- protects updating of if_list */
0235 #define   if_lock(t)           mutex_lock(&(t->if_lock));
0236 #define   if_unlock(t)           mutex_unlock(&(t->if_lock));
0237
0238 /* Used to help with determining the pkts on receive */
0239 #define PKTGEN_MAGIC 0xbe9be955
0240 #define PG_PROC_DIR "pktgen"
0241 #define PGCTRL      "pgctrl"
0242
0243 #define MAX_CFLOWS  65536
0244
0245 #define VLAN_TAG_SIZE(x) ((x)->vlan_id == 0xffff ? 0 : 4)
0246 #define SVLAN_TAG_SIZE(x) ((x)->svlan_id == 0xffff ? 0 : 4)
0247
0248 struct imix_pkt {
0249     u64 size;
0250     u64 weight;
0251     u64 count_so_far;
0252 };
0253
0254 struct flow_state {
0255     __be32 cur_daddr;
0256     int count;
0257 #ifdef CONFIG_XFRM
0258     struct xfrm_state *x;
0259 #endif
0260     __u32 flags;
0261 };
0262
0263 /* flow flag bits */
0264 #define F_INIT   (1<<0)     /* flow has been initialized */
0265
0266 struct pktgen_dev {
0267     /*
0268      * Try to keep frequent/infrequent used vars. separated.
0269      */
0270     struct proc_dir_entry *entry;   /* proc file */
0271     struct pktgen_thread *pg_thread;/* the owner */
0272     struct list_head list;      /* chaining in the thread's run-queue */
0273     struct rcu_head  rcu;       /* freed by RCU */
0274
0275     int running;        /* if false, the test will stop */
0276
0277     /* If min != max, then we will either do a linear iteration, or
0278      * we will do a random selection from within the range.
0279      */
0280     __u32 flags;
0281     int xmit_mode;
0282     int min_pkt_size;
0283     int max_pkt_size;
0284     int pkt_overhead;   /* overhead for MPLS, VLANs, IPSEC etc */
0285     int nfrags;
0286     int removal_mark;   /* non-zero => the device is marked for
0287                  * removal by worker thread */
0288
0289     struct page *page;
0290     u64 delay;      /* nano-seconds */
0291
0292     __u64 count;        /* Default No packets to send */
0293     __u64 sofar;        /* How many pkts we've sent so far */
0294     __u64 tx_bytes;     /* How many bytes we've transmitted */
0295     __u64 errors;       /* Errors when trying to transmit, */
0296
0297     /* runtime counters relating to clone_skb */
0298
0299     __u32 clone_count;
0300     int last_ok;        /* Was last skb sent?
0301                  * Or a failed transmit of some sort?
0302                  * This will keep sequence numbers in order
0303                  */
0304     ktime_t next_tx;
0305     ktime_t started_at;
0306     ktime_t stopped_at;
0307     u64 idle_acc;   /* nano-seconds */
0308
0309     __u32 seq_num;
0310
0311     int clone_skb;      /*
0312                  * Use multiple SKBs during packet gen.
0313                  * If this number is greater than 1, then
0314                  * that many copies of the same packet will be
0315                  * sent before a new packet is allocated.
0316                  * If you want to send 1024 identical packets
0317                  * before creating a new packet,
0318                  * set clone_skb to 1024.
0319                  */
0320
0321     char dst_min[IP_NAME_SZ];   /* IP, ie 1.2.3.4 */
0322     char dst_max[IP_NAME_SZ];   /* IP, ie 1.2.3.4 */
0323     char src_min[IP_NAME_SZ];   /* IP, ie 1.2.3.4 */
0324     char src_max[IP_NAME_SZ];   /* IP, ie 1.2.3.4 */
0325
0326     struct in6_addr in6_saddr;
0327     struct in6_addr in6_daddr;
0328     struct in6_addr cur_in6_daddr;
0329     struct in6_addr cur_in6_saddr;
0330     /* For ranges */
0331     struct in6_addr min_in6_daddr;
0332     struct in6_addr max_in6_daddr;
0333     struct in6_addr min_in6_saddr;
0334     struct in6_addr max_in6_saddr;
0335
0336     /* If we're doing ranges, random or incremental, then this
0337      * defines the min/max for those ranges.
0338      */
0339     __be32 saddr_min;   /* inclusive, source IP address */
0340     __be32 saddr_max;   /* exclusive, source IP address */
0341     __be32 daddr_min;   /* inclusive, dest IP address */
0342     __be32 daddr_max;   /* exclusive, dest IP address */
0343
0344     __u16 udp_src_min;  /* inclusive, source UDP port */
0345     __u16 udp_src_max;  /* exclusive, source UDP port */
0346     __u16 udp_dst_min;  /* inclusive, dest UDP port */
0347     __u16 udp_dst_max;  /* exclusive, dest UDP port */
0348
0349     /* DSCP + ECN */
0350     __u8 tos;            /* six MSB of (former) IPv4 TOS
0351                 are for dscp codepoint */
0352     __u8 traffic_class;  /* ditto for the (former) Traffic Class in IPv6
0353                 (see RFC 3260, sec. 4) */
0354
0355     /* IMIX */
0356     unsigned int n_imix_entries;
0357     struct imix_pkt imix_entries[MAX_IMIX_ENTRIES];
0358     /* Maps 0-IMIX_PRECISION range to imix_entry based on probability*/
0359     __u8 imix_distribution[IMIX_PRECISION];
0360
0361     /* MPLS */
0362     unsigned int nr_labels; /* Depth of stack, 0 = no MPLS */
0363     __be32 labels[MAX_MPLS_LABELS];
0364
0365     /* VLAN/SVLAN (802.1Q/Q-in-Q) */
0366     __u8  vlan_p;
0367     __u8  vlan_cfi;
0368     __u16 vlan_id;  /* 0xffff means no vlan tag */
0369
0370     __u8  svlan_p;
0371     __u8  svlan_cfi;
0372     __u16 svlan_id; /* 0xffff means no svlan tag */
0373
0374     __u32 src_mac_count;    /* How many MACs to iterate through */
0375     __u32 dst_mac_count;    /* How many MACs to iterate through */
0376
0377     unsigned char dst_mac[ETH_ALEN];
0378     unsigned char src_mac[ETH_ALEN];
0379
0380     __u32 cur_dst_mac_offset;
0381     __u32 cur_src_mac_offset;
0382     __be32 cur_saddr;
0383     __be32 cur_daddr;
0384     __u16 ip_id;
0385     __u16 cur_udp_dst;
0386     __u16 cur_udp_src;
0387     __u16 cur_queue_map;
0388     __u32 cur_pkt_size;
0389     __u32 last_pkt_size;
0390
0391     __u8 hh[14];
0392     /* = {
0393        0x00, 0x80, 0xC8, 0x79, 0xB3, 0xCB,
0394
0395        We fill in SRC address later
0396        0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0397        0x08, 0x00
0398        };
0399      */
0400     __u16 pad;      /* pad out the hh struct to an even 16 bytes */
0401
0402     struct sk_buff *skb;    /* skb we are to transmit next, used for when we
0403                  * are transmitting the same one multiple times
0404                  */
0405     struct net_device *odev; /* The out-going device.
0406                   * Note that the device should have it's
0407                   * pg_info pointer pointing back to this
0408                   * device.
0409                   * Set when the user specifies the out-going
0410                   * device name (not when the inject is
0411                   * started as it used to do.)
0412                   */
0413     netdevice_tracker dev_tracker;
0414     char odevname[32];
0415     struct flow_state *flows;
0416     unsigned int cflows;    /* Concurrent flows (config) */
0417     unsigned int lflow;     /* Flow length  (config) */
0418     unsigned int nflows;    /* accumulated flows (stats) */
0419     unsigned int curfl;     /* current sequenced flow (state)*/
0420
0421     u16 queue_map_min;
0422     u16 queue_map_max;
0423     __u32 skb_priority; /* skb priority field */
0424     unsigned int burst; /* number of duplicated packets to burst */
0425     int node;               /* Memory node */
0426
0427 #ifdef CONFIG_XFRM
0428     __u8    ipsmode;        /* IPSEC mode (config) */
0429     __u8    ipsproto;       /* IPSEC type (config) */
0430     __u32   spi;
0431     struct xfrm_dst xdst;
0432     struct dst_ops dstops;
0433 #endif
0434     char result[512];
0435 };
0436
0437 struct pktgen_hdr {
0438     __be32 pgh_magic;
0439     __be32 seq_num;
0440     __be32 tv_sec;
0441     __be32 tv_usec;
0442 };
0443
0444
0445 static unsigned int pg_net_id __read_mostly;
0446
0447 struct pktgen_net {
0448     struct net      *net;
0449     struct proc_dir_entry   *proc_dir;
0450     struct list_head    pktgen_threads;
0451     bool            pktgen_exiting;
0452 };
0453
0454 struct pktgen_thread {
0455     struct mutex if_lock;       /* for list of devices */
0456     struct list_head if_list;   /* All device here */
0457     struct list_head th_list;
0458     struct task_struct *tsk;
0459     char result[512];
0460
0461     /* Field for thread to receive "posted" events terminate,
0462        stop ifs etc. */
0463
0464     u32 control;
0465     int cpu;
0466
0467     wait_queue_head_t queue;
0468     struct completion start_done;
0469     struct pktgen_net *net;
0470 };
0471
0472 #define REMOVE 1
0473 #define FIND   0
0474
0475 static const char version[] =
0476     "Packet Generator for packet performance testing. "
0477     "Version: " VERSION "\n";
0478
0479 static int pktgen_remove_device(struct pktgen_thread *t, struct pktgen_dev *i);
0480 static int pktgen_add_device(struct pktgen_thread *t, const char *ifname);
0481 static struct pktgen_dev *pktgen_find_dev(struct pktgen_thread *t,
0482                       const char *ifname, bool exact);
0483 static int pktgen_device_event(struct notifier_block *, unsigned long, void *);
0484 static void pktgen_run_all_threads(struct pktgen_net *pn);
0485 static void pktgen_reset_all_threads(struct pktgen_net *pn);
0486 static void pktgen_stop_all_threads(struct pktgen_net *pn);
0487
0488 static void pktgen_stop(struct pktgen_thread *t);
0489 static void pktgen_clear_counters(struct pktgen_dev *pkt_dev);
0490 static void fill_imix_distribution(struct pktgen_dev *pkt_dev);
0491
0492 /* Module parameters, defaults. */
0493 static int pg_count_d __read_mostly = 1000;
0494 static int pg_delay_d __read_mostly;
0495 static int pg_clone_skb_d  __read_mostly;
0496 static int debug  __read_mostly;
0497
0498 static DEFINE_MUTEX(pktgen_thread_lock);
0499
0500 static struct notifier_block pktgen_notifier_block = {
0501     .notifier_call = pktgen_device_event,
0502 };
0503
0504 /*
0505  * /proc handling functions
0506  *
0507  */
0508
0509 static int pgctrl_show(struct seq_file *seq, void *v)
0510 {
0511     seq_puts(seq, version);
0512     return 0;
0513 }
0514
0515 static ssize_t pgctrl_write(struct file *file, const char __user *buf,
0516                 size_t count, loff_t *ppos)
0517 {
0518     char data[128];
0519     struct pktgen_net *pn = net_generic(current->nsproxy->net_ns, pg_net_id);
0520
0521     if (!capable(CAP_NET_ADMIN))
0522         return -EPERM;
0523
0524     if (count == 0)
0525         return -EINVAL;
0526
0527     if (count > sizeof(data))
0528         count = sizeof(data);
0529
0530     if (copy_from_user(data, buf, count))
0531         return -EFAULT;
0532
0533     data[count - 1] = 0;    /* Strip trailing '\n' and terminate string */
0534
0535     if (!strcmp(data, "stop"))
0536         pktgen_stop_all_threads(pn);
0537     else if (!strcmp(data, "start"))
0538         pktgen_run_all_threads(pn);
0539     else if (!strcmp(data, "reset"))
0540         pktgen_reset_all_threads(pn);
0541     else
0542         return -EINVAL;
0543
0544     return count;
0545 }
0546
0547 static int pgctrl_open(struct inode *inode, struct file *file)
0548 {
0549     return single_open(file, pgctrl_show, pde_data(inode));
0550 }
0551
0552 static const struct proc_ops pktgen_proc_ops = {
0553     .proc_open  = pgctrl_open,
0554     .proc_read  = seq_read,
0555     .proc_lseek = seq_lseek,
0556     .proc_write = pgctrl_write,
0557     .proc_release   = single_release,
0558 };
0559
0560 static int pktgen_if_show(struct seq_file *seq, void *v)
0561 {
0562     const struct pktgen_dev *pkt_dev = seq->private;
0563     ktime_t stopped;
0564     unsigned int i;
0565     u64 idle;
0566
0567     seq_printf(seq,
0568            "Params: count %llu  min_pkt_size: %u  max_pkt_size: %u\n",
0569            (unsigned long long)pkt_dev->count, pkt_dev->min_pkt_size,
0570            pkt_dev->max_pkt_size);
0571
0572     if (pkt_dev->n_imix_entries > 0) {
0573         seq_puts(seq, "     imix_weights: ");
0574         for (i = 0; i < pkt_dev->n_imix_entries; i++) {
0575             seq_printf(seq, "%llu,%llu ",
0576                    pkt_dev->imix_entries[i].size,
0577                    pkt_dev->imix_entries[i].weight);
0578         }
0579         seq_puts(seq, "\n");
0580     }
0581
0582     seq_printf(seq,
0583            "     frags: %d  delay: %llu  clone_skb: %d  ifname: %s\n",
0584            pkt_dev->nfrags, (unsigned long long) pkt_dev->delay,
0585            pkt_dev->clone_skb, pkt_dev->odevname);
0586
0587     seq_printf(seq, "     flows: %u flowlen: %u\n", pkt_dev->cflows,
0588            pkt_dev->lflow);
0589
0590     seq_printf(seq,
0591            "     queue_map_min: %u  queue_map_max: %u\n",
0592            pkt_dev->queue_map_min,
0593            pkt_dev->queue_map_max);
0594
0595     if (pkt_dev->skb_priority)
0596         seq_printf(seq, "     skb_priority: %u\n",
0597                pkt_dev->skb_priority);
0598
0599     if (pkt_dev->flags & F_IPV6) {
0600         seq_printf(seq,
0601                "     saddr: %pI6c  min_saddr: %pI6c  max_saddr: %pI6c\n"
0602                "     daddr: %pI6c  min_daddr: %pI6c  max_daddr: %pI6c\n",
0603                &pkt_dev->in6_saddr,
0604                &pkt_dev->min_in6_saddr, &pkt_dev->max_in6_saddr,
0605                &pkt_dev->in6_daddr,
0606                &pkt_dev->min_in6_daddr, &pkt_dev->max_in6_daddr);
0607     } else {
0608         seq_printf(seq,
0609                "     dst_min: %s  dst_max: %s\n",
0610                pkt_dev->dst_min, pkt_dev->dst_max);
0611         seq_printf(seq,
0612                "     src_min: %s  src_max: %s\n",
0613                pkt_dev->src_min, pkt_dev->src_max);
0614     }
0615
0616     seq_puts(seq, "     src_mac: ");
0617
0618     seq_printf(seq, "%pM ",
0619            is_zero_ether_addr(pkt_dev->src_mac) ?
0620                  pkt_dev->odev->dev_addr : pkt_dev->src_mac);
0621
0622     seq_puts(seq, "dst_mac: ");
0623     seq_printf(seq, "%pM\n", pkt_dev->dst_mac);
0624
0625     seq_printf(seq,
0626            "     udp_src_min: %d  udp_src_max: %d"
0627            "  udp_dst_min: %d  udp_dst_max: %d\n",
0628            pkt_dev->udp_src_min, pkt_dev->udp_src_max,
0629            pkt_dev->udp_dst_min, pkt_dev->udp_dst_max);
0630
0631     seq_printf(seq,
0632            "     src_mac_count: %d  dst_mac_count: %d\n",
0633            pkt_dev->src_mac_count, pkt_dev->dst_mac_count);
0634
0635     if (pkt_dev->nr_labels) {
0636         seq_puts(seq, "     mpls: ");
0637         for (i = 0; i < pkt_dev->nr_labels; i++)
0638             seq_printf(seq, "%08x%s", ntohl(pkt_dev->labels[i]),
0639                    i == pkt_dev->nr_labels-1 ? "\n" : ", ");
0640     }
0641
0642     if (pkt_dev->vlan_id != 0xffff)
0643         seq_printf(seq, "     vlan_id: %u  vlan_p: %u  vlan_cfi: %u\n",
0644                pkt_dev->vlan_id, pkt_dev->vlan_p,
0645                pkt_dev->vlan_cfi);
0646
0647     if (pkt_dev->svlan_id != 0xffff)
0648         seq_printf(seq, "     svlan_id: %u  vlan_p: %u  vlan_cfi: %u\n",
0649                pkt_dev->svlan_id, pkt_dev->svlan_p,
0650                pkt_dev->svlan_cfi);
0651
0652     if (pkt_dev->tos)
0653         seq_printf(seq, "     tos: 0x%02x\n", pkt_dev->tos);
0654
0655     if (pkt_dev->traffic_class)
0656         seq_printf(seq, "     traffic_class: 0x%02x\n", pkt_dev->traffic_class);
0657
0658     if (pkt_dev->burst > 1)
0659         seq_printf(seq, "     burst: %d\n", pkt_dev->burst);
0660
0661     if (pkt_dev->node >= 0)
0662         seq_printf(seq, "     node: %d\n", pkt_dev->node);
0663
0664     if (pkt_dev->xmit_mode == M_NETIF_RECEIVE)
0665         seq_puts(seq, "     xmit_mode: netif_receive\n");
0666     else if (pkt_dev->xmit_mode == M_QUEUE_XMIT)
0667         seq_puts(seq, "     xmit_mode: xmit_queue\n");
0668
0669     seq_puts(seq, "     Flags: ");
0670
0671     for (i = 0; i < NR_PKT_FLAGS; i++) {
0672         if (i == F_FLOW_SEQ)
0673             if (!pkt_dev->cflows)
0674                 continue;
0675
0676         if (pkt_dev->flags & (1 << i))
0677             seq_printf(seq, "%s  ", pkt_flag_names[i]);
0678         else if (i == F_FLOW_SEQ)
0679             seq_puts(seq, "FLOW_RND  ");
0680
0681 #ifdef CONFIG_XFRM
0682         if (i == F_IPSEC && pkt_dev->spi)
0683             seq_printf(seq, "spi:%u", pkt_dev->spi);
0684 #endif
0685     }
0686
0687     seq_puts(seq, "\n");
0688
0689     /* not really stopped, more like last-running-at */
0690     stopped = pkt_dev->running ? ktime_get() : pkt_dev->stopped_at;
0691     idle = pkt_dev->idle_acc;
0692     do_div(idle, NSEC_PER_USEC);
0693
0694     seq_printf(seq,
0695            "Current:\n     pkts-sofar: %llu  errors: %llu\n",
0696            (unsigned long long)pkt_dev->sofar,
0697            (unsigned long long)pkt_dev->errors);
0698
0699     if (pkt_dev->n_imix_entries > 0) {
0700         int i;
0701
0702         seq_puts(seq, "     imix_size_counts: ");
0703         for (i = 0; i < pkt_dev->n_imix_entries; i++) {
0704             seq_printf(seq, "%llu,%llu ",
0705                    pkt_dev->imix_entries[i].size,
0706                    pkt_dev->imix_entries[i].count_so_far);
0707         }
0708         seq_puts(seq, "\n");
0709     }
0710
0711     seq_printf(seq,
0712            "     started: %lluus  stopped: %lluus idle: %lluus\n",
0713            (unsigned long long) ktime_to_us(pkt_dev->started_at),
0714            (unsigned long long) ktime_to_us(stopped),
0715            (unsigned long long) idle);
0716
0717     seq_printf(seq,
0718            "     seq_num: %d  cur_dst_mac_offset: %d  cur_src_mac_offset: %d\n",
0719            pkt_dev->seq_num, pkt_dev->cur_dst_mac_offset,
0720            pkt_dev->cur_src_mac_offset);
0721
0722     if (pkt_dev->flags & F_IPV6) {
0723         seq_printf(seq, "     cur_saddr: %pI6c  cur_daddr: %pI6c\n",
0724                 &pkt_dev->cur_in6_saddr,
0725                 &pkt_dev->cur_in6_daddr);
0726     } else
0727         seq_printf(seq, "     cur_saddr: %pI4  cur_daddr: %pI4\n",
0728                &pkt_dev->cur_saddr, &pkt_dev->cur_daddr);
0729
0730     seq_printf(seq, "     cur_udp_dst: %d  cur_udp_src: %d\n",
0731            pkt_dev->cur_udp_dst, pkt_dev->cur_udp_src);
0732
0733     seq_printf(seq, "     cur_queue_map: %u\n", pkt_dev->cur_queue_map);
0734
0735     seq_printf(seq, "     flows: %u\n", pkt_dev->nflows);
0736
0737     if (pkt_dev->result[0])
0738         seq_printf(seq, "Result: %s\n", pkt_dev->result);
0739     else
0740         seq_puts(seq, "Result: Idle\n");
0741
0742     return 0;
0743 }
0744
0745
0746 static int hex32_arg(const char __user *user_buffer, unsigned long maxlen,
0747              __u32 *num)
0748 {
0749     int i = 0;
0750     *num = 0;
0751
0752     for (; i < maxlen; i++) {
0753         int value;
0754         char c;
0755         *num <<= 4;
0756         if (get_user(c, &user_buffer[i]))
0757             return -EFAULT;
0758         value = hex_to_bin(c);
0759         if (value >= 0)
0760             *num |= value;
0761         else
0762             break;
0763     }
0764     return i;
0765 }
0766
0767 static int count_trail_chars(const char __user * user_buffer,
0768                  unsigned int maxlen)
0769 {
0770     int i;
0771
0772     for (i = 0; i < maxlen; i++) {
0773         char c;
0774         if (get_user(c, &user_buffer[i]))
0775             return -EFAULT;
0776         switch (c) {
0777         case '\"':
0778         case '\n':
0779         case '\r':
0780         case '\t':
0781         case ' ':
0782         case '=':
0783             break;
0784         default:
0785             goto done;
0786         }
0787     }
0788 done:
0789     return i;
0790 }
0791
0792 static long num_arg(const char __user *user_buffer, unsigned long maxlen,
0793                 unsigned long *num)
0794 {
0795     int i;
0796     *num = 0;
0797
0798     for (i = 0; i < maxlen; i++) {
0799         char c;
0800         if (get_user(c, &user_buffer[i]))
0801             return -EFAULT;
0802         if ((c >= '0') && (c <= '9')) {
0803             *num *= 10;
0804             *num += c - '0';
0805         } else
0806             break;
0807     }
0808     return i;
0809 }
0810
0811 static int strn_len(const char __user * user_buffer, unsigned int maxlen)
0812 {
0813     int i;
0814
0815     for (i = 0; i < maxlen; i++) {
0816         char c;
0817         if (get_user(c, &user_buffer[i]))
0818             return -EFAULT;
0819         switch (c) {
0820         case '\"':
0821         case '\n':
0822         case '\r':
0823         case '\t':
0824         case ' ':
0825             goto done_str;
0826         default:
0827             break;
0828         }
0829     }
0830 done_str:
0831     return i;
0832 }
0833
0834 /* Parses imix entries from user buffer.
0835  * The user buffer should consist of imix entries separated by spaces
0836  * where each entry consists of size and weight delimited by commas.
0837  * "size1,weight_1 size2,weight_2 ... size_n,weight_n" for example.
0838  */
0839 static ssize_t get_imix_entries(const char __user *buffer,
0840                 struct pktgen_dev *pkt_dev)
0841 {
0842     const int max_digits = 10;
0843     int i = 0;
0844     long len;
0845     char c;
0846
0847     pkt_dev->n_imix_entries = 0;
0848
0849     do {
0850         unsigned long weight;
0851         unsigned long size;
0852
0853         len = num_arg(&buffer[i], max_digits, &size);
0854         if (len < 0)
0855             return len;
0856         i += len;
0857         if (get_user(c, &buffer[i]))
0858             return -EFAULT;
0859         /* Check for comma between size_i and weight_i */
0860         if (c != ',')
0861             return -EINVAL;
0862         i++;
0863
0864         if (size < 14 + 20 + 8)
0865             size = 14 + 20 + 8;
0866
0867         len = num_arg(&buffer[i], max_digits, &weight);
0868         if (len < 0)
0869             return len;
0870         if (weight <= 0)
0871             return -EINVAL;
0872
0873         pkt_dev->imix_entries[pkt_dev->n_imix_entries].size = size;
0874         pkt_dev->imix_entries[pkt_dev->n_imix_entries].weight = weight;
0875
0876         i += len;
0877         if (get_user(c, &buffer[i]))
0878             return -EFAULT;
0879
0880         i++;
0881         pkt_dev->n_imix_entries++;
0882
0883         if (pkt_dev->n_imix_entries > MAX_IMIX_ENTRIES)
0884             return -E2BIG;
0885     } while (c == ' ');
0886
0887     return i;
0888 }
0889
0890 static ssize_t get_labels(const char __user *buffer, struct pktgen_dev *pkt_dev)
0891 {
0892     unsigned int n = 0;
0893     char c;
0894     ssize_t i = 0;
0895     int len;
0896
0897     pkt_dev->nr_labels = 0;
0898     do {
0899         __u32 tmp;
0900         len = hex32_arg(&buffer[i], 8, &tmp);
0901         if (len <= 0)
0902             return len;
0903         pkt_dev->labels[n] = htonl(tmp);
0904         if (pkt_dev->labels[n] & MPLS_STACK_BOTTOM)
0905             pkt_dev->flags |= F_MPLS_RND;
0906         i += len;
0907         if (get_user(c, &buffer[i]))
0908             return -EFAULT;
0909         i++;
0910         n++;
0911         if (n >= MAX_MPLS_LABELS)
0912             return -E2BIG;
0913     } while (c == ',');
0914
0915     pkt_dev->nr_labels = n;
0916     return i;
0917 }
0918
0919 static __u32 pktgen_read_flag(const char *f, bool *disable)
0920 {
0921     __u32 i;
0922
0923     if (f[0] == '!') {
0924         *disable = true;
0925         f++;
0926     }
0927
0928     for (i = 0; i < NR_PKT_FLAGS; i++) {
0929         if (!IS_ENABLED(CONFIG_XFRM) && i == IPSEC_SHIFT)
0930             continue;
0931
0932         /* allow only disabling ipv6 flag */
0933         if (!*disable && i == IPV6_SHIFT)
0934             continue;
0935
0936         if (strcmp(f, pkt_flag_names[i]) == 0)
0937             return 1 << i;
0938     }
0939
0940     if (strcmp(f, "FLOW_RND") == 0) {
0941         *disable = !*disable;
0942         return F_FLOW_SEQ;
0943     }
0944
0945     return 0;
0946 }
0947
0948 static ssize_t pktgen_if_write(struct file *file,
0949                    const char __user * user_buffer, size_t count,
0950                    loff_t * offset)
0951 {
0952     struct seq_file *seq = file->private_data;
0953     struct pktgen_dev *pkt_dev = seq->private;
0954     int i, max, len;
0955     char name[16], valstr[32];
0956     unsigned long value = 0;
0957     char *pg_result = NULL;
0958     int tmp = 0;
0959     char buf[128];
0960
0961     pg_result = &(pkt_dev->result[0]);
0962
0963     if (count < 1) {
0964         pr_warn("wrong command format\n");
0965         return -EINVAL;
0966     }
0967
0968     max = count;
0969     tmp = count_trail_chars(user_buffer, max);
0970     if (tmp < 0) {
0971         pr_warn("illegal format\n");
0972         return tmp;
0973     }
0974     i = tmp;
0975
0976     /* Read variable name */
0977
0978     len = strn_len(&user_buffer[i], sizeof(name) - 1);
0979     if (len < 0)
0980         return len;
0981
0982     memset(name, 0, sizeof(name));
0983     if (copy_from_user(name, &user_buffer[i], len))
0984         return -EFAULT;
0985     i += len;
0986
0987     max = count - i;
0988     len = count_trail_chars(&user_buffer[i], max);
0989     if (len < 0)
0990         return len;
0991
0992     i += len;
0993
0994     if (debug) {
0995         size_t copy = min_t(size_t, count + 1, 1024);
0996         char *tp = strndup_user(user_buffer, copy);
0997
0998         if (IS_ERR(tp))
0999             return PTR_ERR(tp);
1000
1001         pr_debug("%s,%zu  buffer -:%s:-\n", name, count, tp);
1002         kfree(tp);
1003     }
1004
1005     if (!strcmp(name, "min_pkt_size")) {
1006         len = num_arg(&user_buffer[i], 10, &value);
1007         if (len < 0)
1008             return len;
1009
1010         i += len;
1011         if (value < 14 + 20 + 8)
1012             value = 14 + 20 + 8;
1013         if (value != pkt_dev->min_pkt_size) {
1014             pkt_dev->min_pkt_size = value;
1015             pkt_dev->cur_pkt_size = value;
1016         }
1017         sprintf(pg_result, "OK: min_pkt_size=%d",
1018             pkt_dev->min_pkt_size);
1019         return count;
1020     }
1021
1022     if (!strcmp(name, "max_pkt_size")) {
1023         len = num_arg(&user_buffer[i], 10, &value);
1024         if (len < 0)
1025             return len;
1026
1027         i += len;
1028         if (value < 14 + 20 + 8)
1029             value = 14 + 20 + 8;
1030         if (value != pkt_dev->max_pkt_size) {
1031             pkt_dev->max_pkt_size = value;
1032             pkt_dev->cur_pkt_size = value;
1033         }
1034         sprintf(pg_result, "OK: max_pkt_size=%d",
1035             pkt_dev->max_pkt_size);
1036         return count;
1037     }
1038
1039     /* Shortcut for min = max */
1040
1041     if (!strcmp(name, "pkt_size")) {
1042         len = num_arg(&user_buffer[i], 10, &value);
1043         if (len < 0)
1044             return len;
1045
1046         i += len;
1047         if (value < 14 + 20 + 8)
1048             value = 14 + 20 + 8;
1049         if (value != pkt_dev->min_pkt_size) {
1050             pkt_dev->min_pkt_size = value;
1051             pkt_dev->max_pkt_size = value;
1052             pkt_dev->cur_pkt_size = value;
1053         }
1054         sprintf(pg_result, "OK: pkt_size=%d", pkt_dev->min_pkt_size);
1055         return count;
1056     }
1057
1058     if (!strcmp(name, "imix_weights")) {
1059         if (pkt_dev->clone_skb > 0)
1060             return -EINVAL;
1061
1062         len = get_imix_entries(&user_buffer[i], pkt_dev);
1063         if (len < 0)
1064             return len;
1065
1066         fill_imix_distribution(pkt_dev);
1067
1068         i += len;
1069         return count;
1070     }
1071
1072     if (!strcmp(name, "debug")) {
1073         len = num_arg(&user_buffer[i], 10, &value);
1074         if (len < 0)
1075             return len;
1076
1077         i += len;
1078         debug = value;
1079         sprintf(pg_result, "OK: debug=%u", debug);
1080         return count;
1081     }
1082
1083     if (!strcmp(name, "frags")) {
1084         len = num_arg(&user_buffer[i], 10, &value);
1085         if (len < 0)
1086             return len;
1087
1088         i += len;
1089         pkt_dev->nfrags = value;
1090         sprintf(pg_result, "OK: frags=%d", pkt_dev->nfrags);
1091         return count;
1092     }
1093     if (!strcmp(name, "delay")) {
1094         len = num_arg(&user_buffer[i], 10, &value);
1095         if (len < 0)
1096             return len;
1097
1098         i += len;
1099         if (value == 0x7FFFFFFF)
1100             pkt_dev->delay = ULLONG_MAX;
1101         else
1102             pkt_dev->delay = (u64)value;
1103
1104         sprintf(pg_result, "OK: delay=%llu",
1105             (unsigned long long) pkt_dev->delay);
1106         return count;
1107     }
1108     if (!strcmp(name, "rate")) {
1109         len = num_arg(&user_buffer[i], 10, &value);
1110         if (len < 0)
1111             return len;
1112
1113         i += len;
1114         if (!value)
1115             return len;
1116         pkt_dev->delay = pkt_dev->min_pkt_size*8*NSEC_PER_USEC/value;
1117         if (debug)
1118             pr_info("Delay set at: %llu ns\n", pkt_dev->delay);
1119
1120         sprintf(pg_result, "OK: rate=%lu", value);
1121         return count;
1122     }
1123     if (!strcmp(name, "ratep")) {
1124         len = num_arg(&user_buffer[i], 10, &value);
1125         if (len < 0)
1126             return len;
1127
1128         i += len;
1129         if (!value)
1130             return len;
1131         pkt_dev->delay = NSEC_PER_SEC/value;
1132         if (debug)
1133             pr_info("Delay set at: %llu ns\n", pkt_dev->delay);
1134
1135         sprintf(pg_result, "OK: rate=%lu", value);
1136         return count;
1137     }
1138     if (!strcmp(name, "udp_src_min")) {
1139         len = num_arg(&user_buffer[i], 10, &value);
1140         if (len < 0)
1141             return len;
1142
1143         i += len;
1144         if (value != pkt_dev->udp_src_min) {
1145             pkt_dev->udp_src_min = value;
1146             pkt_dev->cur_udp_src = value;
1147         }
1148         sprintf(pg_result, "OK: udp_src_min=%u", pkt_dev->udp_src_min);
1149         return count;
1150     }
1151     if (!strcmp(name, "udp_dst_min")) {
1152         len = num_arg(&user_buffer[i], 10, &value);
1153         if (len < 0)
1154             return len;
1155
1156         i += len;
1157         if (value != pkt_dev->udp_dst_min) {
1158             pkt_dev->udp_dst_min = value;
1159             pkt_dev->cur_udp_dst = value;
1160         }
1161         sprintf(pg_result, "OK: udp_dst_min=%u", pkt_dev->udp_dst_min);
1162         return count;
1163     }
1164     if (!strcmp(name, "udp_src_max")) {
1165         len = num_arg(&user_buffer[i], 10, &value);
1166         if (len < 0)
1167             return len;
1168
1169         i += len;
1170         if (value != pkt_dev->udp_src_max) {
1171             pkt_dev->udp_src_max = value;
1172             pkt_dev->cur_udp_src = value;
1173         }
1174         sprintf(pg_result, "OK: udp_src_max=%u", pkt_dev->udp_src_max);
1175         return count;
1176     }
1177     if (!strcmp(name, "udp_dst_max")) {
1178         len = num_arg(&user_buffer[i], 10, &value);
1179         if (len < 0)
1180             return len;
1181
1182         i += len;
1183         if (value != pkt_dev->udp_dst_max) {
1184             pkt_dev->udp_dst_max = value;
1185             pkt_dev->cur_udp_dst = value;
1186         }
1187         sprintf(pg_result, "OK: udp_dst_max=%u", pkt_dev->udp_dst_max);
1188         return count;
1189     }
1190     if (!strcmp(name, "clone_skb")) {
1191         len = num_arg(&user_buffer[i], 10, &value);
1192         if (len < 0)
1193             return len;
1194         /* clone_skb is not supported for netif_receive xmit_mode and
1195          * IMIX mode.
1196          */
1197         if ((value > 0) &&
1198             ((pkt_dev->xmit_mode == M_NETIF_RECEIVE) ||
1199              !(pkt_dev->odev->priv_flags & IFF_TX_SKB_SHARING)))
1200             return -ENOTSUPP;
1201         if (value > 0 && pkt_dev->n_imix_entries > 0)
1202             return -EINVAL;
1203
1204         i += len;
1205         pkt_dev->clone_skb = value;
1206
1207         sprintf(pg_result, "OK: clone_skb=%d", pkt_dev->clone_skb);
1208         return count;
1209     }
1210     if (!strcmp(name, "count")) {
1211         len = num_arg(&user_buffer[i], 10, &value);
1212         if (len < 0)
1213             return len;
1214
1215         i += len;
1216         pkt_dev->count = value;
1217         sprintf(pg_result, "OK: count=%llu",
1218             (unsigned long long)pkt_dev->count);
1219         return count;
1220     }
1221     if (!strcmp(name, "src_mac_count")) {
1222         len = num_arg(&user_buffer[i], 10, &value);
1223         if (len < 0)
1224             return len;
1225
1226         i += len;
1227         if (pkt_dev->src_mac_count != value) {
1228             pkt_dev->src_mac_count = value;
1229             pkt_dev->cur_src_mac_offset = 0;
1230         }
1231         sprintf(pg_result, "OK: src_mac_count=%d",
1232             pkt_dev->src_mac_count);
1233         return count;
1234     }
1235     if (!strcmp(name, "dst_mac_count")) {
1236         len = num_arg(&user_buffer[i], 10, &value);
1237         if (len < 0)
1238             return len;
1239
1240         i += len;
1241         if (pkt_dev->dst_mac_count != value) {
1242             pkt_dev->dst_mac_count = value;
1243             pkt_dev->cur_dst_mac_offset = 0;
1244         }
1245         sprintf(pg_result, "OK: dst_mac_count=%d",
1246             pkt_dev->dst_mac_count);
1247         return count;
1248     }
1249     if (!strcmp(name, "burst")) {
1250         len = num_arg(&user_buffer[i], 10, &value);
1251         if (len < 0)
1252             return len;
1253
1254         i += len;
1255         if ((value > 1) &&
1256             ((pkt_dev->xmit_mode == M_QUEUE_XMIT) ||
1257              ((pkt_dev->xmit_mode == M_START_XMIT) &&
1258              (!(pkt_dev->odev->priv_flags & IFF_TX_SKB_SHARING)))))
1259             return -ENOTSUPP;
1260         pkt_dev->burst = value < 1 ? 1 : value;
1261         sprintf(pg_result, "OK: burst=%u", pkt_dev->burst);
1262         return count;
1263     }
1264     if (!strcmp(name, "node")) {
1265         len = num_arg(&user_buffer[i], 10, &value);
1266         if (len < 0)
1267             return len;
1268
1269         i += len;
1270
1271         if (node_possible(value)) {
1272             pkt_dev->node = value;
1273             sprintf(pg_result, "OK: node=%d", pkt_dev->node);
1274             if (pkt_dev->page) {
1275                 put_page(pkt_dev->page);
1276                 pkt_dev->page = NULL;
1277             }
1278         }
1279         else
1280             sprintf(pg_result, "ERROR: node not possible");
1281         return count;
1282     }
1283     if (!strcmp(name, "xmit_mode")) {
1284         char f[32];
1285
1286         memset(f, 0, 32);
1287         len = strn_len(&user_buffer[i], sizeof(f) - 1);
1288         if (len < 0)
1289             return len;
1290
1291         if (copy_from_user(f, &user_buffer[i], len))
1292             return -EFAULT;
1293         i += len;
1294
1295         if (strcmp(f, "start_xmit") == 0) {
1296             pkt_dev->xmit_mode = M_START_XMIT;
1297         } else if (strcmp(f, "netif_receive") == 0) {
1298             /* clone_skb set earlier, not supported in this mode */
1299             if (pkt_dev->clone_skb > 0)
1300                 return -ENOTSUPP;
1301
1302             pkt_dev->xmit_mode = M_NETIF_RECEIVE;
1303
1304             /* make sure new packet is allocated every time
1305              * pktgen_xmit() is called
1306              */
1307             pkt_dev->last_ok = 1;
1308         } else if (strcmp(f, "queue_xmit") == 0) {
1309             pkt_dev->xmit_mode = M_QUEUE_XMIT;
1310             pkt_dev->last_ok = 1;
1311         } else {
1312             sprintf(pg_result,
1313                 "xmit_mode -:%s:- unknown\nAvailable modes: %s",
1314                 f, "start_xmit, netif_receive\n");
1315             return count;
1316         }
1317         sprintf(pg_result, "OK: xmit_mode=%s", f);
1318         return count;
1319     }
1320     if (!strcmp(name, "flag")) {
1321         __u32 flag;
1322         char f[32];
1323         bool disable = false;
1324
1325         memset(f, 0, 32);
1326         len = strn_len(&user_buffer[i], sizeof(f) - 1);
1327         if (len < 0)
1328             return len;
1329
1330         if (copy_from_user(f, &user_buffer[i], len))
1331             return -EFAULT;
1332         i += len;
1333
1334         flag = pktgen_read_flag(f, &disable);
1335
1336         if (flag) {
1337             if (disable)
1338                 pkt_dev->flags &= ~flag;
1339             else
1340                 pkt_dev->flags |= flag;
1341         } else {
1342             sprintf(pg_result,
1343                 "Flag -:%s:- unknown\nAvailable flags, (prepend ! to un-set flag):\n%s",
1344                 f,
1345                 "IPSRC_RND, IPDST_RND, UDPSRC_RND, UDPDST_RND, "
1346                 "MACSRC_RND, MACDST_RND, TXSIZE_RND, IPV6, "
1347                 "MPLS_RND, VID_RND, SVID_RND, FLOW_SEQ, "
1348                 "QUEUE_MAP_RND, QUEUE_MAP_CPU, UDPCSUM, "
1349                 "NO_TIMESTAMP, "
1350 #ifdef CONFIG_XFRM
1351                 "IPSEC, "
1352 #endif
1353                 "NODE_ALLOC\n");
1354             return count;
1355         }
1356         sprintf(pg_result, "OK: flags=0x%x", pkt_dev->flags);
1357         return count;
1358     }
1359     if (!strcmp(name, "dst_min") || !strcmp(name, "dst")) {
1360         len = strn_len(&user_buffer[i], sizeof(pkt_dev->dst_min) - 1);
1361         if (len < 0)
1362             return len;
1363
1364         if (copy_from_user(buf, &user_buffer[i], len))
1365             return -EFAULT;
1366         buf[len] = 0;
1367         if (strcmp(buf, pkt_dev->dst_min) != 0) {
1368             memset(pkt_dev->dst_min, 0, sizeof(pkt_dev->dst_min));
1369             strcpy(pkt_dev->dst_min, buf);
1370             pkt_dev->daddr_min = in_aton(pkt_dev->dst_min);
1371             pkt_dev->cur_daddr = pkt_dev->daddr_min;
1372         }
1373         if (debug)
1374             pr_debug("dst_min set to: %s\n", pkt_dev->dst_min);
1375         i += len;
1376         sprintf(pg_result, "OK: dst_min=%s", pkt_dev->dst_min);
1377         return count;
1378     }
1379     if (!strcmp(name, "dst_max")) {
1380         len = strn_len(&user_buffer[i], sizeof(pkt_dev->dst_max) - 1);
1381         if (len < 0)
1382             return len;
1383
1384         if (copy_from_user(buf, &user_buffer[i], len))
1385             return -EFAULT;
1386         buf[len] = 0;
1387         if (strcmp(buf, pkt_dev->dst_max) != 0) {
1388             memset(pkt_dev->dst_max, 0, sizeof(pkt_dev->dst_max));
1389             strcpy(pkt_dev->dst_max, buf);
1390             pkt_dev->daddr_max = in_aton(pkt_dev->dst_max);
1391             pkt_dev->cur_daddr = pkt_dev->daddr_max;
1392         }
1393         if (debug)
1394             pr_debug("dst_max set to: %s\n", pkt_dev->dst_max);
1395         i += len;
1396         sprintf(pg_result, "OK: dst_max=%s", pkt_dev->dst_max);
1397         return count;
1398     }
1399     if (!strcmp(name, "dst6")) {
1400         len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1401         if (len < 0)
1402             return len;
1403
1404         pkt_dev->flags |= F_IPV6;
1405
1406         if (copy_from_user(buf, &user_buffer[i], len))
1407             return -EFAULT;
1408         buf[len] = 0;
1409
1410         in6_pton(buf, -1, pkt_dev->in6_daddr.s6_addr, -1, NULL);
1411         snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->in6_daddr);
1412
1413         pkt_dev->cur_in6_daddr = pkt_dev->in6_daddr;
1414
1415         if (debug)
1416             pr_debug("dst6 set to: %s\n", buf);
1417
1418         i += len;
1419         sprintf(pg_result, "OK: dst6=%s", buf);
1420         return count;
1421     }
1422     if (!strcmp(name, "dst6_min")) {
1423         len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1424         if (len < 0)
1425             return len;
1426
1427         pkt_dev->flags |= F_IPV6;
1428
1429         if (copy_from_user(buf, &user_buffer[i], len))
1430             return -EFAULT;
1431         buf[len] = 0;
1432
1433         in6_pton(buf, -1, pkt_dev->min_in6_daddr.s6_addr, -1, NULL);
1434         snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->min_in6_daddr);
1435
1436         pkt_dev->cur_in6_daddr = pkt_dev->min_in6_daddr;
1437         if (debug)
1438             pr_debug("dst6_min set to: %s\n", buf);
1439
1440         i += len;
1441         sprintf(pg_result, "OK: dst6_min=%s", buf);
1442         return count;
1443     }
1444     if (!strcmp(name, "dst6_max")) {
1445         len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1446         if (len < 0)
1447             return len;
1448
1449         pkt_dev->flags |= F_IPV6;
1450
1451         if (copy_from_user(buf, &user_buffer[i], len))
1452             return -EFAULT;
1453         buf[len] = 0;
1454
1455         in6_pton(buf, -1, pkt_dev->max_in6_daddr.s6_addr, -1, NULL);
1456         snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->max_in6_daddr);
1457
1458         if (debug)
1459             pr_debug("dst6_max set to: %s\n", buf);
1460
1461         i += len;
1462         sprintf(pg_result, "OK: dst6_max=%s", buf);
1463         return count;
1464     }
1465     if (!strcmp(name, "src6")) {
1466         len = strn_len(&user_buffer[i], sizeof(buf) - 1);
1467         if (len < 0)
1468             return len;
1469
1470         pkt_dev->flags |= F_IPV6;
1471
1472         if (copy_from_user(buf, &user_buffer[i], len))
1473             return -EFAULT;
1474         buf[len] = 0;
1475
1476         in6_pton(buf, -1, pkt_dev->in6_saddr.s6_addr, -1, NULL);
1477         snprintf(buf, sizeof(buf), "%pI6c", &pkt_dev->in6_saddr);
1478
1479         pkt_dev->cur_in6_saddr = pkt_dev->in6_saddr;
1480
1481         if (debug)
1482             pr_debug("src6 set to: %s\n", buf);
1483
1484         i += len;
1485         sprintf(pg_result, "OK: src6=%s", buf);
1486         return count;
1487     }
1488     if (!strcmp(name, "src_min")) {
1489         len = strn_len(&user_buffer[i], sizeof(pkt_dev->src_min) - 1);
1490         if (len < 0)
1491             return len;
1492
1493         if (copy_from_user(buf, &user_buffer[i], len))
1494             return -EFAULT;
1495         buf[len] = 0;
1496         if (strcmp(buf, pkt_dev->src_min) != 0) {
1497             memset(pkt_dev->src_min, 0, sizeof(pkt_dev->src_min));
1498             strcpy(pkt_dev->src_min, buf);
1499             pkt_dev->saddr_min = in_aton(pkt_dev->src_min);
1500             pkt_dev->cur_saddr = pkt_dev->saddr_min;
1501         }
1502         if (debug)
1503             pr_debug("src_min set to: %s\n", pkt_dev->src_min);
1504         i += len;
1505         sprintf(pg_result, "OK: src_min=%s", pkt_dev->src_min);
1506         return count;
1507     }
1508     if (!strcmp(name, "src_max")) {
1509         len = strn_len(&user_buffer[i], sizeof(pkt_dev->src_max) - 1);
1510         if (len < 0)
1511             return len;
1512
1513         if (copy_from_user(buf, &user_buffer[i], len))
1514             return -EFAULT;
1515         buf[len] = 0;
1516         if (strcmp(buf, pkt_dev->src_max) != 0) {
1517             memset(pkt_dev->src_max, 0, sizeof(pkt_dev->src_max));
1518             strcpy(pkt_dev->src_max, buf);
1519             pkt_dev->saddr_max = in_aton(pkt_dev->src_max);
1520             pkt_dev->cur_saddr = pkt_dev->saddr_max;
1521         }
1522         if (debug)
1523             pr_debug("src_max set to: %s\n", pkt_dev->src_max);
1524         i += len;
1525         sprintf(pg_result, "OK: src_max=%s", pkt_dev->src_max);
1526         return count;
1527     }
1528     if (!strcmp(name, "dst_mac")) {
1529         len = strn_len(&user_buffer[i], sizeof(valstr) - 1);
1530         if (len < 0)
1531             return len;
1532
1533         memset(valstr, 0, sizeof(valstr));
1534         if (copy_from_user(valstr, &user_buffer[i], len))
1535             return -EFAULT;
1536
1537         if (!mac_pton(valstr, pkt_dev->dst_mac))
1538             return -EINVAL;
1539         /* Set up Dest MAC */
1540         ether_addr_copy(&pkt_dev->hh[0], pkt_dev->dst_mac);
1541
1542         sprintf(pg_result, "OK: dstmac %pM", pkt_dev->dst_mac);
1543         return count;
1544     }
1545     if (!strcmp(name, "src_mac")) {
1546         len = strn_len(&user_buffer[i], sizeof(valstr) - 1);
1547         if (len < 0)
1548             return len;
1549
1550         memset(valstr, 0, sizeof(valstr));
1551         if (copy_from_user(valstr, &user_buffer[i], len))
1552             return -EFAULT;
1553
1554         if (!mac_pton(valstr, pkt_dev->src_mac))
1555             return -EINVAL;
1556         /* Set up Src MAC */
1557         ether_addr_copy(&pkt_dev->hh[6], pkt_dev->src_mac);
1558
1559         sprintf(pg_result, "OK: srcmac %pM", pkt_dev->src_mac);
1560         return count;
1561     }
1562
1563     if (!strcmp(name, "clear_counters")) {
1564         pktgen_clear_counters(pkt_dev);
1565         sprintf(pg_result, "OK: Clearing counters.\n");
1566         return count;
1567     }
1568
1569     if (!strcmp(name, "flows")) {
1570         len = num_arg(&user_buffer[i], 10, &value);
1571         if (len < 0)
1572             return len;
1573
1574         i += len;
1575         if (value > MAX_CFLOWS)
1576             value = MAX_CFLOWS;
1577
1578         pkt_dev->cflows = value;
1579         sprintf(pg_result, "OK: flows=%u", pkt_dev->cflows);
1580         return count;
1581     }
1582 #ifdef CONFIG_XFRM
1583     if (!strcmp(name, "spi")) {
1584         len = num_arg(&user_buffer[i], 10, &value);
1585         if (len < 0)
1586             return len;
1587
1588         i += len;
1589         pkt_dev->spi = value;
1590         sprintf(pg_result, "OK: spi=%u", pkt_dev->spi);
1591         return count;
1592     }
1593 #endif
1594     if (!strcmp(name, "flowlen")) {
1595         len = num_arg(&user_buffer[i], 10, &value);
1596         if (len < 0)
1597             return len;
1598
1599         i += len;
1600         pkt_dev->lflow = value;
1601         sprintf(pg_result, "OK: flowlen=%u", pkt_dev->lflow);
1602         return count;
1603     }
1604
1605     if (!strcmp(name, "queue_map_min")) {
1606         len = num_arg(&user_buffer[i], 5, &value);
1607         if (len < 0)
1608             return len;
1609
1610         i += len;
1611         pkt_dev->queue_map_min = value;
1612         sprintf(pg_result, "OK: queue_map_min=%u", pkt_dev->queue_map_min);
1613         return count;
1614     }
1615
1616     if (!strcmp(name, "queue_map_max")) {
1617         len = num_arg(&user_buffer[i], 5, &value);
1618         if (len < 0)
1619             return len;
1620
1621         i += len;
1622         pkt_dev->queue_map_max = value;
1623         sprintf(pg_result, "OK: queue_map_max=%u", pkt_dev->queue_map_max);
1624         return count;
1625     }
1626
1627     if (!strcmp(name, "mpls")) {
1628         unsigned int n, cnt;
1629
1630         len = get_labels(&user_buffer[i], pkt_dev);
1631         if (len < 0)
1632             return len;
1633         i += len;
1634         cnt = sprintf(pg_result, "OK: mpls=");
1635         for (n = 0; n < pkt_dev->nr_labels; n++)
1636             cnt += sprintf(pg_result + cnt,
1637                        "%08x%s", ntohl(pkt_dev->labels[n]),
1638                        n == pkt_dev->nr_labels-1 ? "" : ",");
1639
1640         if (pkt_dev->nr_labels && pkt_dev->vlan_id != 0xffff) {
1641             pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1642             pkt_dev->svlan_id = 0xffff;
1643
1644             if (debug)
1645                 pr_debug("VLAN/SVLAN auto turned off\n");
1646         }
1647         return count;
1648     }
1649
1650     if (!strcmp(name, "vlan_id")) {
1651         len = num_arg(&user_buffer[i], 4, &value);
1652         if (len < 0)
1653             return len;
1654
1655         i += len;
1656         if (value <= 4095) {
1657             pkt_dev->vlan_id = value;  /* turn on VLAN */
1658
1659             if (debug)
1660                 pr_debug("VLAN turned on\n");
1661
1662             if (debug && pkt_dev->nr_labels)
1663                 pr_debug("MPLS auto turned off\n");
1664
1665             pkt_dev->nr_labels = 0;    /* turn off MPLS */
1666             sprintf(pg_result, "OK: vlan_id=%u", pkt_dev->vlan_id);
1667         } else {
1668             pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1669             pkt_dev->svlan_id = 0xffff;
1670
1671             if (debug)
1672                 pr_debug("VLAN/SVLAN turned off\n");
1673         }
1674         return count;
1675     }
1676
1677     if (!strcmp(name, "vlan_p")) {
1678         len = num_arg(&user_buffer[i], 1, &value);
1679         if (len < 0)
1680             return len;
1681
1682         i += len;
1683         if ((value <= 7) && (pkt_dev->vlan_id != 0xffff)) {
1684             pkt_dev->vlan_p = value;
1685             sprintf(pg_result, "OK: vlan_p=%u", pkt_dev->vlan_p);
1686         } else {
1687             sprintf(pg_result, "ERROR: vlan_p must be 0-7");
1688         }
1689         return count;
1690     }
1691
1692     if (!strcmp(name, "vlan_cfi")) {
1693         len = num_arg(&user_buffer[i], 1, &value);
1694         if (len < 0)
1695             return len;
1696
1697         i += len;
1698         if ((value <= 1) && (pkt_dev->vlan_id != 0xffff)) {
1699             pkt_dev->vlan_cfi = value;
1700             sprintf(pg_result, "OK: vlan_cfi=%u", pkt_dev->vlan_cfi);
1701         } else {
1702             sprintf(pg_result, "ERROR: vlan_cfi must be 0-1");
1703         }
1704         return count;
1705     }
1706
1707     if (!strcmp(name, "svlan_id")) {
1708         len = num_arg(&user_buffer[i], 4, &value);
1709         if (len < 0)
1710             return len;
1711
1712         i += len;
1713         if ((value <= 4095) && ((pkt_dev->vlan_id != 0xffff))) {
1714             pkt_dev->svlan_id = value;  /* turn on SVLAN */
1715
1716             if (debug)
1717                 pr_debug("SVLAN turned on\n");
1718
1719             if (debug && pkt_dev->nr_labels)
1720                 pr_debug("MPLS auto turned off\n");
1721
1722             pkt_dev->nr_labels = 0;    /* turn off MPLS */
1723             sprintf(pg_result, "OK: svlan_id=%u", pkt_dev->svlan_id);
1724         } else {
1725             pkt_dev->vlan_id = 0xffff; /* turn off VLAN/SVLAN */
1726             pkt_dev->svlan_id = 0xffff;
1727
1728             if (debug)
1729                 pr_debug("VLAN/SVLAN turned off\n");
1730         }
1731         return count;
1732     }
1733
1734     if (!strcmp(name, "svlan_p")) {
1735         len = num_arg(&user_buffer[i], 1, &value);
1736         if (len < 0)
1737             return len;
1738
1739         i += len;
1740         if ((value <= 7) && (pkt_dev->svlan_id != 0xffff)) {
1741             pkt_dev->svlan_p = value;
1742             sprintf(pg_result, "OK: svlan_p=%u", pkt_dev->svlan_p);
1743         } else {
1744             sprintf(pg_result, "ERROR: svlan_p must be 0-7");
1745         }
1746         return count;
1747     }
1748
1749     if (!strcmp(name, "svlan_cfi")) {
1750         len = num_arg(&user_buffer[i], 1, &value);
1751         if (len < 0)
1752             return len;
1753
1754         i += len;
1755         if ((value <= 1) && (pkt_dev->svlan_id != 0xffff)) {
1756             pkt_dev->svlan_cfi = value;
1757             sprintf(pg_result, "OK: svlan_cfi=%u", pkt_dev->svlan_cfi);
1758         } else {
1759             sprintf(pg_result, "ERROR: svlan_cfi must be 0-1");
1760         }
1761         return count;
1762     }
1763
1764     if (!strcmp(name, "tos")) {
1765         __u32 tmp_value = 0;
1766         len = hex32_arg(&user_buffer[i], 2, &tmp_value);
1767         if (len < 0)
1768             return len;
1769
1770         i += len;
1771         if (len == 2) {
1772             pkt_dev->tos = tmp_value;
1773             sprintf(pg_result, "OK: tos=0x%02x", pkt_dev->tos);
1774         } else {
1775             sprintf(pg_result, "ERROR: tos must be 00-ff");
1776         }
1777         return count;
1778     }
1779
1780     if (!strcmp(name, "traffic_class")) {
1781         __u32 tmp_value = 0;
1782         len = hex32_arg(&user_buffer[i], 2, &tmp_value);
1783         if (len < 0)
1784             return len;
1785
1786         i += len;
1787         if (len == 2) {
1788             pkt_dev->traffic_class = tmp_value;
1789             sprintf(pg_result, "OK: traffic_class=0x%02x", pkt_dev->traffic_class);
1790         } else {
1791             sprintf(pg_result, "ERROR: traffic_class must be 00-ff");
1792         }
1793         return count;
1794     }
1795
1796     if (!strcmp(name, "skb_priority")) {
1797         len = num_arg(&user_buffer[i], 9, &value);
1798         if (len < 0)
1799             return len;
1800
1801         i += len;
1802         pkt_dev->skb_priority = value;
1803         sprintf(pg_result, "OK: skb_priority=%i",
1804             pkt_dev->skb_priority);
1805         return count;
1806     }
1807
1808     sprintf(pkt_dev->result, "No such parameter \"%s\"", name);
1809     return -EINVAL;
1810 }
1811
1812 static int pktgen_if_open(struct inode *inode, struct file *file)
1813 {
1814     return single_open(file, pktgen_if_show, pde_data(inode));
1815 }
1816
1817 static const struct proc_ops pktgen_if_proc_ops = {
1818     .proc_open  = pktgen_if_open,
1819     .proc_read  = seq_read,
1820     .proc_lseek = seq_lseek,
1821     .proc_write = pktgen_if_write,
1822     .proc_release   = single_release,
1823 };
1824
1825 static int pktgen_thread_show(struct seq_file *seq, void *v)
1826 {
1827     struct pktgen_thread *t = seq->private;
1828     const struct pktgen_dev *pkt_dev;
1829
1830     BUG_ON(!t);
1831
1832     seq_puts(seq, "Running: ");
1833
1834     rcu_read_lock();
1835     list_for_each_entry_rcu(pkt_dev, &t->if_list, list)
1836         if (pkt_dev->running)
1837             seq_printf(seq, "%s ", pkt_dev->odevname);
1838
1839     seq_puts(seq, "\nStopped: ");
1840
1841     list_for_each_entry_rcu(pkt_dev, &t->if_list, list)
1842         if (!pkt_dev->running)
1843             seq_printf(seq, "%s ", pkt_dev->odevname);
1844
1845     if (t->result[0])
1846         seq_printf(seq, "\nResult: %s\n", t->result);
1847     else
1848         seq_puts(seq, "\nResult: NA\n");
1849
1850     rcu_read_unlock();
1851
1852     return 0;
1853 }
1854
1855 static ssize_t pktgen_thread_write(struct file *file,
1856                    const char __user * user_buffer,
1857                    size_t count, loff_t * offset)
1858 {
1859     struct seq_file *seq = file->private_data;
1860     struct pktgen_thread *t = seq->private;
1861     int i, max, len, ret;
1862     char name[40];
1863     char *pg_result;
1864
1865     if (count < 1) {
1866         //      sprintf(pg_result, "Wrong command format");
1867         return -EINVAL;
1868     }
1869
1870     max = count;
1871     len = count_trail_chars(user_buffer, max);
1872     if (len < 0)
1873         return len;
1874
1875     i = len;
1876
1877     /* Read variable name */
1878
1879     len = strn_len(&user_buffer[i], sizeof(name) - 1);
1880     if (len < 0)
1881         return len;
1882
1883     memset(name, 0, sizeof(name));
1884     if (copy_from_user(name, &user_buffer[i], len))
1885         return -EFAULT;
1886     i += len;
1887
1888     max = count - i;
1889     len = count_trail_chars(&user_buffer[i], max);
1890     if (len < 0)
1891         return len;
1892
1893     i += len;
1894
1895     if (debug)
1896         pr_debug("t=%s, count=%lu\n", name, (unsigned long)count);
1897
1898     if (!t) {
1899         pr_err("ERROR: No thread\n");
1900         ret = -EINVAL;
1901         goto out;
1902     }
1903
1904     pg_result = &(t->result[0]);
1905
1906     if (!strcmp(name, "add_device")) {
1907         char f[32];
1908         memset(f, 0, 32);
1909         len = strn_len(&user_buffer[i], sizeof(f) - 1);
1910         if (len < 0) {
1911             ret = len;
1912             goto out;
1913         }
1914         if (copy_from_user(f, &user_buffer[i], len))
1915             return -EFAULT;
1916         i += len;
1917         mutex_lock(&pktgen_thread_lock);
1918         ret = pktgen_add_device(t, f);
1919         mutex_unlock(&pktgen_thread_lock);
1920         if (!ret) {
1921             ret = count;
1922             sprintf(pg_result, "OK: add_device=%s", f);
1923         } else
1924             sprintf(pg_result, "ERROR: can not add device %s", f);
1925         goto out;
1926     }
1927
1928     if (!strcmp(name, "rem_device_all")) {
1929         mutex_lock(&pktgen_thread_lock);
1930         t->control |= T_REMDEVALL;
1931         mutex_unlock(&pktgen_thread_lock);
1932         schedule_timeout_interruptible(msecs_to_jiffies(125));  /* Propagate thread->control  */
1933         ret = count;
1934         sprintf(pg_result, "OK: rem_device_all");
1935         goto out;
1936     }
1937
1938     if (!strcmp(name, "max_before_softirq")) {
1939         sprintf(pg_result, "OK: Note! max_before_softirq is obsoleted -- Do not use");
1940         ret = count;
1941         goto out;
1942     }
1943
1944     ret = -EINVAL;
1945 out:
1946     return ret;
1947 }
1948
1949 static int pktgen_thread_open(struct inode *inode, struct file *file)
1950 {
1951     return single_open(file, pktgen_thread_show, pde_data(inode));
1952 }
1953
1954 static const struct proc_ops pktgen_thread_proc_ops = {
1955     .proc_open  = pktgen_thread_open,
1956     .proc_read  = seq_read,
1957     .proc_lseek = seq_lseek,
1958     .proc_write = pktgen_thread_write,
1959     .proc_release   = single_release,
1960 };
1961
1962 /* Think find or remove for NN */
1963 static struct pktgen_dev *__pktgen_NN_threads(const struct pktgen_net *pn,
1964                           const char *ifname, int remove)
1965 {
1966     struct pktgen_thread *t;
1967     struct pktgen_dev *pkt_dev = NULL;
1968     bool exact = (remove == FIND);
1969
1970     list_for_each_entry(t, &pn->pktgen_threads, th_list) {
1971         pkt_dev = pktgen_find_dev(t, ifname, exact);
1972         if (pkt_dev) {
1973             if (remove) {
1974                 pkt_dev->removal_mark = 1;
1975                 t->control |= T_REMDEV;
1976             }
1977             break;
1978         }
1979     }
1980     return pkt_dev;
1981 }
1982
1983 /*
1984  * mark a device for removal
1985  */
1986 static void pktgen_mark_device(const struct pktgen_net *pn, const char *ifname)
1987 {
1988     struct pktgen_dev *pkt_dev = NULL;
1989     const int max_tries = 10, msec_per_try = 125;
1990     int i = 0;
1991
1992     mutex_lock(&pktgen_thread_lock);
1993     pr_debug("%s: marking %s for removal\n", __func__, ifname);
1994
1995     while (1) {
1996
1997         pkt_dev = __pktgen_NN_threads(pn, ifname, REMOVE);
1998         if (pkt_dev == NULL)
1999             break;  /* success */
2000
2001         mutex_unlock(&pktgen_thread_lock);
2002         pr_debug("%s: waiting for %s to disappear....\n",
2003              __func__, ifname);
2004         schedule_timeout_interruptible(msecs_to_jiffies(msec_per_try));
2005         mutex_lock(&pktgen_thread_lock);
2006
2007         if (++i >= max_tries) {
2008             pr_err("%s: timed out after waiting %d msec for device %s to be removed\n",
2009                    __func__, msec_per_try * i, ifname);
2010             break;
2011         }
2012
2013     }
2014
2015     mutex_unlock(&pktgen_thread_lock);
2016 }
2017
2018 static void pktgen_change_name(const struct pktgen_net *pn, struct net_device *dev)
2019 {
2020     struct pktgen_thread *t;
2021
2022     mutex_lock(&pktgen_thread_lock);
2023
2024     list_for_each_entry(t, &pn->pktgen_threads, th_list) {
2025         struct pktgen_dev *pkt_dev;
2026
2027         if_lock(t);
2028         list_for_each_entry(pkt_dev, &t->if_list, list) {
2029             if (pkt_dev->odev != dev)
2030                 continue;
2031
2032             proc_remove(pkt_dev->entry);
2033
2034             pkt_dev->entry = proc_create_data(dev->name, 0600,
2035                               pn->proc_dir,
2036                               &pktgen_if_proc_ops,
2037                               pkt_dev);
2038             if (!pkt_dev->entry)
2039                 pr_err("can't move proc entry for '%s'\n",
2040                        dev->name);
2041             break;
2042         }
2043         if_unlock(t);
2044     }
2045     mutex_unlock(&pktgen_thread_lock);
2046 }
2047
2048 static int pktgen_device_event(struct notifier_block *unused,
2049                    unsigned long event, void *ptr)
2050 {
2051     struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2052     struct pktgen_net *pn = net_generic(dev_net(dev), pg_net_id);
2053
2054     if (pn->pktgen_exiting)
2055         return NOTIFY_DONE;
2056
2057     /* It is OK that we do not hold the group lock right now,
2058      * as we run under the RTNL lock.
2059      */
2060
2061     switch (event) {
2062     case NETDEV_CHANGENAME:
2063         pktgen_change_name(pn, dev);
2064         break;
2065
2066     case NETDEV_UNREGISTER:
2067         pktgen_mark_device(pn, dev->name);
2068         break;
2069     }
2070
2071     return NOTIFY_DONE;
2072 }
2073
2074 static struct net_device *pktgen_dev_get_by_name(const struct pktgen_net *pn,
2075                          struct pktgen_dev *pkt_dev,
2076                          const char *ifname)
2077 {
2078     char b[IFNAMSIZ+5];
2079     int i;
2080
2081     for (i = 0; ifname[i] != '@'; i++) {
2082         if (i == IFNAMSIZ)
2083             break;
2084
2085         b[i] = ifname[i];
2086     }
2087     b[i] = 0;
2088
2089     return dev_get_by_name(pn->net, b);
2090 }
2091
2092
2093 /* Associate pktgen_dev with a device. */
2094
2095 static int pktgen_setup_dev(const struct pktgen_net *pn,
2096                 struct pktgen_dev *pkt_dev, const char *ifname)
2097 {
2098     struct net_device *odev;
2099     int err;
2100
2101     /* Clean old setups */
2102     if (pkt_dev->odev) {
2103         netdev_put(pkt_dev->odev, &pkt_dev->dev_tracker);
2104         pkt_dev->odev = NULL;
2105     }
2106
2107     odev = pktgen_dev_get_by_name(pn, pkt_dev, ifname);
2108     if (!odev) {
2109         pr_err("no such netdevice: \"%s\"\n", ifname);
2110         return -ENODEV;
2111     }
2112
2113     if (odev->type != ARPHRD_ETHER && odev->type != ARPHRD_LOOPBACK) {
2114         pr_err("not an ethernet or loopback device: \"%s\"\n", ifname);
2115         err = -EINVAL;
2116     } else if (!netif_running(odev)) {
2117         pr_err("device is down: \"%s\"\n", ifname);
2118         err = -ENETDOWN;
2119     } else {
2120         pkt_dev->odev = odev;
2121         netdev_tracker_alloc(odev, &pkt_dev->dev_tracker, GFP_KERNEL);
2122         return 0;
2123     }
2124
2125     dev_put(odev);
2126     return err;
2127 }
2128
2129 /* Read pkt_dev from the interface and set up internal pktgen_dev
2130  * structure to have the right information to create/send packets
2131  */
2132 static void pktgen_setup_inject(struct pktgen_dev *pkt_dev)
2133 {
2134     int ntxq;
2135
2136     if (!pkt_dev->odev) {
2137         pr_err("ERROR: pkt_dev->odev == NULL in setup_inject\n");
2138         sprintf(pkt_dev->result,
2139             "ERROR: pkt_dev->odev == NULL in setup_inject.\n");
2140         return;
2141     }
2142
2143     /* make sure that we don't pick a non-existing transmit queue */
2144     ntxq = pkt_dev->odev->real_num_tx_queues;
2145
2146     if (ntxq <= pkt_dev->queue_map_min) {
2147         pr_warn("WARNING: Requested queue_map_min (zero-based) (%d) exceeds valid range [0 - %d] for (%d) queues on %s, resetting\n",
2148             pkt_dev->queue_map_min, (ntxq ?: 1) - 1, ntxq,
2149             pkt_dev->odevname);
2150         pkt_dev->queue_map_min = (ntxq ?: 1) - 1;
2151     }
2152     if (pkt_dev->queue_map_max >= ntxq) {
2153         pr_warn("WARNING: Requested queue_map_max (zero-based) (%d) exceeds valid range [0 - %d] for (%d) queues on %s, resetting\n",
2154             pkt_dev->queue_map_max, (ntxq ?: 1) - 1, ntxq,
2155             pkt_dev->odevname);
2156         pkt_dev->queue_map_max = (ntxq ?: 1) - 1;
2157     }
2158
2159     /* Default to the interface's mac if not explicitly set. */
2160
2161     if (is_zero_ether_addr(pkt_dev->src_mac))
2162         ether_addr_copy(&(pkt_dev->hh[6]), pkt_dev->odev->dev_addr);
2163
2164     /* Set up Dest MAC */
2165     ether_addr_copy(&(pkt_dev->hh[0]), pkt_dev->dst_mac);
2166
2167     if (pkt_dev->flags & F_IPV6) {
2168         int i, set = 0, err = 1;
2169         struct inet6_dev *idev;
2170
2171         if (pkt_dev->min_pkt_size == 0) {
2172             pkt_dev->min_pkt_size = 14 + sizeof(struct ipv6hdr)
2173                         + sizeof(struct udphdr)
2174                         + sizeof(struct pktgen_hdr)
2175                         + pkt_dev->pkt_overhead;
2176         }
2177
2178         for (i = 0; i < sizeof(struct in6_addr); i++)
2179             if (pkt_dev->cur_in6_saddr.s6_addr[i]) {
2180                 set = 1;
2181                 break;
2182             }
2183
2184         if (!set) {
2185
2186             /*
2187              * Use linklevel address if unconfigured.
2188              *
2189              * use ipv6_get_lladdr if/when it's get exported
2190              */
2191
2192             rcu_read_lock();
2193             idev = __in6_dev_get(pkt_dev->odev);
2194             if (idev) {
2195                 struct inet6_ifaddr *ifp;
2196
2197                 read_lock_bh(&idev->lock);
2198                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2199                     if ((ifp->scope & IFA_LINK) &&
2200                         !(ifp->flags & IFA_F_TENTATIVE)) {
2201                         pkt_dev->cur_in6_saddr = ifp->addr;
2202                         err = 0;
2203                         break;
2204                     }
2205                 }
2206                 read_unlock_bh(&idev->lock);
2207             }
2208             rcu_read_unlock();
2209             if (err)
2210                 pr_err("ERROR: IPv6 link address not available\n");
2211         }
2212     } else {
2213         if (pkt_dev->min_pkt_size == 0) {
2214             pkt_dev->min_pkt_size = 14 + sizeof(struct iphdr)
2215                         + sizeof(struct udphdr)
2216                         + sizeof(struct pktgen_hdr)
2217                         + pkt_dev->pkt_overhead;
2218         }
2219
2220         pkt_dev->saddr_min = 0;
2221         pkt_dev->saddr_max = 0;
2222         if (strlen(pkt_dev->src_min) == 0) {
2223
2224             struct in_device *in_dev;
2225
2226             rcu_read_lock();
2227             in_dev = __in_dev_get_rcu(pkt_dev->odev);
2228             if (in_dev) {
2229                 const struct in_ifaddr *ifa;
2230
2231                 ifa = rcu_dereference(in_dev->ifa_list);
2232                 if (ifa) {
2233                     pkt_dev->saddr_min = ifa->ifa_address;
2234                     pkt_dev->saddr_max = pkt_dev->saddr_min;
2235                 }
2236             }
2237             rcu_read_unlock();
2238         } else {
2239             pkt_dev->saddr_min = in_aton(pkt_dev->src_min);
2240             pkt_dev->saddr_max = in_aton(pkt_dev->src_max);
2241         }
2242
2243         pkt_dev->daddr_min = in_aton(pkt_dev->dst_min);
2244         pkt_dev->daddr_max = in_aton(pkt_dev->dst_max);
2245     }
2246     /* Initialize current values. */
2247     pkt_dev->cur_pkt_size = pkt_dev->min_pkt_size;
2248     if (pkt_dev->min_pkt_size > pkt_dev->max_pkt_size)
2249         pkt_dev->max_pkt_size = pkt_dev->min_pkt_size;
2250
2251     pkt_dev->cur_dst_mac_offset = 0;
2252     pkt_dev->cur_src_mac_offset = 0;
2253     pkt_dev->cur_saddr = pkt_dev->saddr_min;
2254     pkt_dev->cur_daddr = pkt_dev->daddr_min;
2255     pkt_dev->cur_udp_dst = pkt_dev->udp_dst_min;
2256     pkt_dev->cur_udp_src = pkt_dev->udp_src_min;
2257     pkt_dev->nflows = 0;
2258 }
2259
2260
2261 static void spin(struct pktgen_dev *pkt_dev, ktime_t spin_until)
2262 {
2263     ktime_t start_time, end_time;
2264     s64 remaining;
2265     struct hrtimer_sleeper t;
2266
2267     hrtimer_init_sleeper_on_stack(&t, CLOCK_MONOTONIC, HRTIMER_MODE_ABS);
2268     hrtimer_set_expires(&t.timer, spin_until);
2269
2270     remaining = ktime_to_ns(hrtimer_expires_remaining(&t.timer));
2271     if (remaining <= 0)
2272         goto out;
2273
2274     start_time = ktime_get();
2275     if (remaining < 100000) {
2276         /* for small delays (<100us), just loop until limit is reached */
2277         do {
2278             end_time = ktime_get();
2279         } while (ktime_compare(end_time, spin_until) < 0);
2280     } else {
2281         do {
2282             set_current_state(TASK_INTERRUPTIBLE);
2283             hrtimer_sleeper_start_expires(&t, HRTIMER_MODE_ABS);
2284
2285             if (likely(t.task))
2286                 schedule();
2287
2288             hrtimer_cancel(&t.timer);
2289         } while (t.task && pkt_dev->running && !signal_pending(current));
2290         __set_current_state(TASK_RUNNING);
2291         end_time = ktime_get();
2292     }
2293
2294     pkt_dev->idle_acc += ktime_to_ns(ktime_sub(end_time, start_time));
2295 out:
2296     pkt_dev->next_tx = ktime_add_ns(spin_until, pkt_dev->delay);
2297     destroy_hrtimer_on_stack(&t.timer);
2298 }
2299
2300 static inline void set_pkt_overhead(struct pktgen_dev *pkt_dev)
2301 {
2302     pkt_dev->pkt_overhead = 0;
2303     pkt_dev->pkt_overhead += pkt_dev->nr_labels*sizeof(u32);
2304     pkt_dev->pkt_overhead += VLAN_TAG_SIZE(pkt_dev);
2305     pkt_dev->pkt_overhead += SVLAN_TAG_SIZE(pkt_dev);
2306 }
2307
2308 static inline int f_seen(const struct pktgen_dev *pkt_dev, int flow)
2309 {
2310     return !!(pkt_dev->flows[flow].flags & F_INIT);
2311 }
2312
2313 static inline int f_pick(struct pktgen_dev *pkt_dev)
2314 {
2315     int flow = pkt_dev->curfl;
2316
2317     if (pkt_dev->flags & F_FLOW_SEQ) {
2318         if (pkt_dev->flows[flow].count >= pkt_dev->lflow) {
2319             /* reset time */
2320             pkt_dev->flows[flow].count = 0;
2321             pkt_dev->flows[flow].flags = 0;
2322             pkt_dev->curfl += 1;
2323             if (pkt_dev->curfl >= pkt_dev->cflows)
2324                 pkt_dev->curfl = 0; /*reset */
2325         }
2326     } else {
2327         flow = prandom_u32() % pkt_dev->cflows;
2328         pkt_dev->curfl = flow;
2329
2330         if (pkt_dev->flows[flow].count > pkt_dev->lflow) {
2331             pkt_dev->flows[flow].count = 0;
2332             pkt_dev->flows[flow].flags = 0;
2333         }
2334     }
2335
2336     return pkt_dev->curfl;
2337 }
2338
2339
2340 #ifdef CONFIG_XFRM
2341 /* If there was already an IPSEC SA, we keep it as is, else
2342  * we go look for it ...
2343 */
2344 #define DUMMY_MARK 0
2345 static void get_ipsec_sa(struct pktgen_dev *pkt_dev, int flow)
2346 {
2347     struct xfrm_state *x = pkt_dev->flows[flow].x;
2348     struct pktgen_net *pn = net_generic(dev_net(pkt_dev->odev), pg_net_id);
2349     if (!x) {
2350
2351         if (pkt_dev->spi) {
2352             /* We need as quick as possible to find the right SA
2353              * Searching with minimum criteria to archieve this.
2354              */
2355             x = xfrm_state_lookup_byspi(pn->net, htonl(pkt_dev->spi), AF_INET);
2356         } else {
2357             /* slow path: we dont already have xfrm_state */
2358             x = xfrm_stateonly_find(pn->net, DUMMY_MARK, 0,
2359                         (xfrm_address_t *)&pkt_dev->cur_daddr,
2360                         (xfrm_address_t *)&pkt_dev->cur_saddr,
2361                         AF_INET,
2362                         pkt_dev->ipsmode,
2363                         pkt_dev->ipsproto, 0);
2364         }
2365         if (x) {
2366             pkt_dev->flows[flow].x = x;
2367             set_pkt_overhead(pkt_dev);
2368             pkt_dev->pkt_overhead += x->props.header_len;
2369         }
2370
2371     }
2372 }
2373 #endif
2374 static void set_cur_queue_map(struct pktgen_dev *pkt_dev)
2375 {
2376
2377     if (pkt_dev->flags & F_QUEUE_MAP_CPU)
2378         pkt_dev->cur_queue_map = smp_processor_id();
2379
2380     else if (pkt_dev->queue_map_min <= pkt_dev->queue_map_max) {
2381         __u16 t;
2382         if (pkt_dev->flags & F_QUEUE_MAP_RND) {
2383             t = prandom_u32() %
2384                 (pkt_dev->queue_map_max -
2385                  pkt_dev->queue_map_min + 1)
2386                 + pkt_dev->queue_map_min;
2387         } else {
2388             t = pkt_dev->cur_queue_map + 1;
2389             if (t > pkt_dev->queue_map_max)
2390                 t = pkt_dev->queue_map_min;
2391         }
2392         pkt_dev->cur_queue_map = t;
2393     }
2394     pkt_dev->cur_queue_map  = pkt_dev->cur_queue_map % pkt_dev->odev->real_num_tx_queues;
2395 }
2396
2397 /* Increment/randomize headers according to flags and current values
2398  * for IP src/dest, UDP src/dst port, MAC-Addr src/dst
2399  */
2400 static void mod_cur_headers(struct pktgen_dev *pkt_dev)
2401 {
2402     __u32 imn;
2403     __u32 imx;
2404     int flow = 0;
2405
2406     if (pkt_dev->cflows)
2407         flow = f_pick(pkt_dev);
2408
2409     /*  Deal with source MAC */
2410     if (pkt_dev->src_mac_count > 1) {
2411         __u32 mc;
2412         __u32 tmp;
2413
2414         if (pkt_dev->flags & F_MACSRC_RND)
2415             mc = prandom_u32() % pkt_dev->src_mac_count;
2416         else {
2417             mc = pkt_dev->cur_src_mac_offset++;
2418             if (pkt_dev->cur_src_mac_offset >=
2419                 pkt_dev->src_mac_count)
2420                 pkt_dev->cur_src_mac_offset = 0;
2421         }
2422
2423         tmp = pkt_dev->src_mac[5] + (mc & 0xFF);
2424         pkt_dev->hh[11] = tmp;
2425         tmp = (pkt_dev->src_mac[4] + ((mc >> 8) & 0xFF) + (tmp >> 8));
2426         pkt_dev->hh[10] = tmp;
2427         tmp = (pkt_dev->src_mac[3] + ((mc >> 16) & 0xFF) + (tmp >> 8));
2428         pkt_dev->hh[9] = tmp;
2429         tmp = (pkt_dev->src_mac[2] + ((mc >> 24) & 0xFF) + (tmp >> 8));
2430         pkt_dev->hh[8] = tmp;
2431         tmp = (pkt_dev->src_mac[1] + (tmp >> 8));
2432         pkt_dev->hh[7] = tmp;
2433     }
2434
2435     /*  Deal with Destination MAC */
2436     if (pkt_dev->dst_mac_count > 1) {
2437         __u32 mc;
2438         __u32 tmp;
2439
2440         if (pkt_dev->flags & F_MACDST_RND)
2441             mc = prandom_u32() % pkt_dev->dst_mac_count;
2442
2443         else {
2444             mc = pkt_dev->cur_dst_mac_offset++;
2445             if (pkt_dev->cur_dst_mac_offset >=
2446                 pkt_dev->dst_mac_count) {
2447                 pkt_dev->cur_dst_mac_offset = 0;
2448             }
2449         }
2450
2451         tmp = pkt_dev->dst_mac[5] + (mc & 0xFF);
2452         pkt_dev->hh[5] = tmp;
2453         tmp = (pkt_dev->dst_mac[4] + ((mc >> 8) & 0xFF) + (tmp >> 8));
2454         pkt_dev->hh[4] = tmp;
2455         tmp = (pkt_dev->dst_mac[3] + ((mc >> 16) & 0xFF) + (tmp >> 8));
2456         pkt_dev->hh[3] = tmp;
2457         tmp = (pkt_dev->dst_mac[2] + ((mc >> 24) & 0xFF) + (tmp >> 8));
2458         pkt_dev->hh[2] = tmp;
2459         tmp = (pkt_dev->dst_mac[1] + (tmp >> 8));
2460         pkt_dev->hh[1] = tmp;
2461     }
2462
2463     if (pkt_dev->flags & F_MPLS_RND) {
2464         unsigned int i;
2465         for (i = 0; i < pkt_dev->nr_labels; i++)
2466             if (pkt_dev->labels[i] & MPLS_STACK_BOTTOM)
2467                 pkt_dev->labels[i] = MPLS_STACK_BOTTOM |
2468                          ((__force __be32)prandom_u32() &
2469                               htonl(0x000fffff));
2470     }
2471
2472     if ((pkt_dev->flags & F_VID_RND) && (pkt_dev->vlan_id != 0xffff)) {
2473         pkt_dev->vlan_id = prandom_u32() & (4096 - 1);
2474     }
2475
2476     if ((pkt_dev->flags & F_SVID_RND) && (pkt_dev->svlan_id != 0xffff)) {
2477         pkt_dev->svlan_id = prandom_u32() & (4096 - 1);
2478     }
2479
2480     if (pkt_dev->udp_src_min < pkt_dev->udp_src_max) {
2481         if (pkt_dev->flags & F_UDPSRC_RND)
2482             pkt_dev->cur_udp_src = prandom_u32() %
2483                 (pkt_dev->udp_src_max - pkt_dev->udp_src_min)
2484                 + pkt_dev->udp_src_min;
2485
2486         else {
2487             pkt_dev->cur_udp_src++;
2488             if (pkt_dev->cur_udp_src >= pkt_dev->udp_src_max)
2489                 pkt_dev->cur_udp_src = pkt_dev->udp_src_min;
2490         }
2491     }
2492
2493     if (pkt_dev->udp_dst_min < pkt_dev->udp_dst_max) {
2494         if (pkt_dev->flags & F_UDPDST_RND) {
2495             pkt_dev->cur_udp_dst = prandom_u32() %
2496                 (pkt_dev->udp_dst_max - pkt_dev->udp_dst_min)
2497                 + pkt_dev->udp_dst_min;
2498         } else {
2499             pkt_dev->cur_udp_dst++;
2500             if (pkt_dev->cur_udp_dst >= pkt_dev->udp_dst_max)
2501                 pkt_dev->cur_udp_dst = pkt_dev->udp_dst_min;
2502         }
2503     }
2504
2505     if (!(pkt_dev->flags & F_IPV6)) {
2506
2507         imn = ntohl(pkt_dev->saddr_min);
2508         imx = ntohl(pkt_dev->saddr_max);
2509         if (imn < imx) {
2510             __u32 t;
2511             if (pkt_dev->flags & F_IPSRC_RND)
2512                 t = prandom_u32() % (imx - imn) + imn;
2513             else {
2514                 t = ntohl(pkt_dev->cur_saddr);
2515                 t++;
2516                 if (t > imx)
2517                     t = imn;
2518
2519             }
2520             pkt_dev->cur_saddr = htonl(t);
2521         }
2522
2523         if (pkt_dev->cflows && f_seen(pkt_dev, flow)) {
2524             pkt_dev->cur_daddr = pkt_dev->flows[flow].cur_daddr;
2525         } else {
2526             imn = ntohl(pkt_dev->daddr_min);
2527             imx = ntohl(pkt_dev->daddr_max);
2528             if (imn < imx) {
2529                 __u32 t;
2530                 __be32 s;
2531                 if (pkt_dev->flags & F_IPDST_RND) {
2532
2533                     do {
2534                         t = prandom_u32() %
2535                             (imx - imn) + imn;
2536                         s = htonl(t);
2537                     } while (ipv4_is_loopback(s) ||
2538                         ipv4_is_multicast(s) ||
2539                         ipv4_is_lbcast(s) ||
2540                         ipv4_is_zeronet(s) ||
2541                         ipv4_is_local_multicast(s));
2542                     pkt_dev->cur_daddr = s;
2543                 } else {
2544                     t = ntohl(pkt_dev->cur_daddr);
2545                     t++;
2546                     if (t > imx) {
2547                         t = imn;
2548                     }
2549                     pkt_dev->cur_daddr = htonl(t);
2550                 }
2551             }
2552             if (pkt_dev->cflows) {
2553                 pkt_dev->flows[flow].flags |= F_INIT;
2554                 pkt_dev->flows[flow].cur_daddr =
2555                     pkt_dev->cur_daddr;
2556 #ifdef CONFIG_XFRM
2557                 if (pkt_dev->flags & F_IPSEC)
2558                     get_ipsec_sa(pkt_dev, flow);
2559 #endif
2560                 pkt_dev->nflows++;
2561             }
2562         }
2563     } else {        /* IPV6 * */
2564
2565         if (!ipv6_addr_any(&pkt_dev->min_in6_daddr)) {
2566             int i;
2567
2568             /* Only random destinations yet */
2569
2570             for (i = 0; i < 4; i++) {
2571                 pkt_dev->cur_in6_daddr.s6_addr32[i] =
2572                     (((__force __be32)prandom_u32() |
2573                       pkt_dev->min_in6_daddr.s6_addr32[i]) &
2574                      pkt_dev->max_in6_daddr.s6_addr32[i]);
2575             }
2576         }
2577     }
2578
2579     if (pkt_dev->min_pkt_size < pkt_dev->max_pkt_size) {
2580         __u32 t;
2581         if (pkt_dev->flags & F_TXSIZE_RND) {
2582             t = prandom_u32() %
2583                 (pkt_dev->max_pkt_size - pkt_dev->min_pkt_size)
2584                 + pkt_dev->min_pkt_size;
2585         } else {
2586             t = pkt_dev->cur_pkt_size + 1;
2587             if (t > pkt_dev->max_pkt_size)
2588                 t = pkt_dev->min_pkt_size;
2589         }
2590         pkt_dev->cur_pkt_size = t;
2591     } else if (pkt_dev->n_imix_entries > 0) {
2592         struct imix_pkt *entry;
2593         __u32 t = prandom_u32() % IMIX_PRECISION;
2594         __u8 entry_index = pkt_dev->imix_distribution[t];
2595
2596         entry = &pkt_dev->imix_entries[entry_index];
2597         entry->count_so_far++;
2598         pkt_dev->cur_pkt_size = entry->size;
2599     }
2600
2601     set_cur_queue_map(pkt_dev);
2602
2603     pkt_dev->flows[flow].count++;
2604 }
2605
2606 static void fill_imix_distribution(struct pktgen_dev *pkt_dev)
2607 {
2608     int cumulative_probabilites[MAX_IMIX_ENTRIES];
2609     int j = 0;
2610     __u64 cumulative_prob = 0;
2611     __u64 total_weight = 0;
2612     int i = 0;
2613
2614     for (i = 0; i < pkt_dev->n_imix_entries; i++)
2615         total_weight += pkt_dev->imix_entries[i].weight;
2616
2617     /* Fill cumulative_probabilites with sum of normalized probabilities */
2618     for (i = 0; i < pkt_dev->n_imix_entries - 1; i++) {
2619         cumulative_prob += div64_u64(pkt_dev->imix_entries[i].weight *
2620                              IMIX_PRECISION,
2621                          total_weight);
2622         cumulative_probabilites[i] = cumulative_prob;
2623     }
2624     cumulative_probabilites[pkt_dev->n_imix_entries - 1] = 100;
2625
2626     for (i = 0; i < IMIX_PRECISION; i++) {
2627         if (i == cumulative_probabilites[j])
2628             j++;
2629         pkt_dev->imix_distribution[i] = j;
2630     }
2631 }
2632
2633 #ifdef CONFIG_XFRM
2634 static u32 pktgen_dst_metrics[RTAX_MAX + 1] = {
2635
2636     [RTAX_HOPLIMIT] = 0x5, /* Set a static hoplimit */
2637 };
2638
2639 static int pktgen_output_ipsec(struct sk_buff *skb, struct pktgen_dev *pkt_dev)
2640 {
2641     struct xfrm_state *x = pkt_dev->flows[pkt_dev->curfl].x;
2642     int err = 0;
2643     struct net *net = dev_net(pkt_dev->odev);
2644
2645     if (!x)
2646         return 0;
2647     /* XXX: we dont support tunnel mode for now until
2648      * we resolve the dst issue */
2649     if ((x->props.mode != XFRM_MODE_TRANSPORT) && (pkt_dev->spi == 0))
2650         return 0;
2651
2652     /* But when user specify an valid SPI, transformation
2653      * supports both transport/tunnel mode + ESP/AH type.
2654      */
2655     if ((x->props.mode == XFRM_MODE_TUNNEL) && (pkt_dev->spi != 0))
2656         skb->_skb_refdst = (unsigned long)&pkt_dev->xdst.u.dst | SKB_DST_NOREF;
2657
2658     rcu_read_lock_bh();
2659     err = pktgen_xfrm_outer_mode_output(x, skb);
2660     rcu_read_unlock_bh();
2661     if (err) {
2662         XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATEMODEERROR);
2663         goto error;
2664     }
2665     err = x->type->output(x, skb);
2666     if (err) {
2667         XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATEPROTOERROR);
2668         goto error;
2669     }
2670     spin_lock_bh(&x->lock);
2671     x->curlft.bytes += skb->len;
2672     x->curlft.packets++;
2673     spin_unlock_bh(&x->lock);
2674 error:
2675     return err;
2676 }
2677
2678 static void free_SAs(struct pktgen_dev *pkt_dev)
2679 {
2680     if (pkt_dev->cflows) {
2681         /* let go of the SAs if we have them */
2682         int i;
2683         for (i = 0; i < pkt_dev->cflows; i++) {
2684             struct xfrm_state *x = pkt_dev->flows[i].x;
2685             if (x) {
2686                 xfrm_state_put(x);
2687                 pkt_dev->flows[i].x = NULL;
2688             }
2689         }
2690     }
2691 }
2692
2693 static int process_ipsec(struct pktgen_dev *pkt_dev,
2694                   struct sk_buff *skb, __be16 protocol)
2695 {
2696     if (pkt_dev->flags & F_IPSEC) {
2697         struct xfrm_state *x = pkt_dev->flows[pkt_dev->curfl].x;
2698         int nhead = 0;
2699         if (x) {
2700             struct ethhdr *eth;
2701             struct iphdr *iph;
2702             int ret;
2703
2704             nhead = x->props.header_len - skb_headroom(skb);
2705             if (nhead > 0) {
2706                 ret = pskb_expand_head(skb, nhead, 0, GFP_ATOMIC);
2707                 if (ret < 0) {
2708                     pr_err("Error expanding ipsec packet %d\n",
2709                            ret);
2710                     goto err;
2711                 }
2712             }
2713
2714             /* ipsec is not expecting ll header */
2715             skb_pull(skb, ETH_HLEN);
2716             ret = pktgen_output_ipsec(skb, pkt_dev);
2717             if (ret) {
2718                 pr_err("Error creating ipsec packet %d\n", ret);
2719                 goto err;
2720             }
2721             /* restore ll */
2722             eth = skb_push(skb, ETH_HLEN);
2723             memcpy(eth, pkt_dev->hh, 2 * ETH_ALEN);
2724             eth->h_proto = protocol;
2725
2726             /* Update IPv4 header len as well as checksum value */
2727             iph = ip_hdr(skb);
2728             iph->tot_len = htons(skb->len - ETH_HLEN);
2729             ip_send_check(iph);
2730         }
2731     }
2732     return 1;
2733 err:
2734     kfree_skb(skb);
2735     return 0;
2736 }
2737 #endif
2738
2739 static void mpls_push(__be32 *mpls, struct pktgen_dev *pkt_dev)
2740 {
2741     unsigned int i;
2742     for (i = 0; i < pkt_dev->nr_labels; i++)
2743         *mpls++ = pkt_dev->labels[i] & ~MPLS_STACK_BOTTOM;
2744
2745     mpls--;
2746     *mpls |= MPLS_STACK_BOTTOM;
2747 }
2748
2749 static inline __be16 build_tci(unsigned int id, unsigned int cfi,
2750                    unsigned int prio)
2751 {
2752     return htons(id | (cfi << 12) | (prio << 13));
2753 }
2754
2755 static void pktgen_finalize_skb(struct pktgen_dev *pkt_dev, struct sk_buff *skb,
2756                 int datalen)
2757 {
2758     struct timespec64 timestamp;
2759     struct pktgen_hdr *pgh;
2760
2761     pgh = skb_put(skb, sizeof(*pgh));
2762     datalen -= sizeof(*pgh);
2763
2764     if (pkt_dev->nfrags <= 0) {
2765         skb_put_zero(skb, datalen);
2766     } else {
2767         int frags = pkt_dev->nfrags;
2768         int i, len;
2769         int frag_len;
2770
2771
2772         if (frags > MAX_SKB_FRAGS)
2773             frags = MAX_SKB_FRAGS;
2774         len = datalen - frags * PAGE_SIZE;
2775         if (len > 0) {
2776             skb_put_zero(skb, len);
2777             datalen = frags * PAGE_SIZE;
2778         }
2779
2780         i = 0;
2781         frag_len = (datalen/frags) < PAGE_SIZE ?
2782                (datalen/frags) : PAGE_SIZE;
2783         while (datalen > 0) {
2784             if (unlikely(!pkt_dev->page)) {
2785                 int node = numa_node_id();
2786
2787                 if (pkt_dev->node >= 0 && (pkt_dev->flags & F_NODE))
2788                     node = pkt_dev->node;
2789                 pkt_dev->page = alloc_pages_node(node, GFP_KERNEL | __GFP_ZERO, 0);
2790                 if (!pkt_dev->page)
2791                     break;
2792             }
2793             get_page(pkt_dev->page);
2794             skb_frag_set_page(skb, i, pkt_dev->page);
2795             skb_frag_off_set(&skb_shinfo(skb)->frags[i], 0);
2796             /*last fragment, fill rest of data*/
2797             if (i == (frags - 1))
2798                 skb_frag_size_set(&skb_shinfo(skb)->frags[i],
2799                     (datalen < PAGE_SIZE ? datalen : PAGE_SIZE));
2800             else
2801                 skb_frag_size_set(&skb_shinfo(skb)->frags[i], frag_len);
2802             datalen -= skb_frag_size(&skb_shinfo(skb)->frags[i]);
2803             skb->len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2804             skb->data_len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2805             i++;
2806             skb_shinfo(skb)->nr_frags = i;
2807         }
2808     }
2809
2810     /* Stamp the time, and sequence number,
2811      * convert them to network byte order
2812      */
2813     pgh->pgh_magic = htonl(PKTGEN_MAGIC);
2814     pgh->seq_num = htonl(pkt_dev->seq_num);
2815
2816     if (pkt_dev->flags & F_NO_TIMESTAMP) {
2817         pgh->tv_sec = 0;
2818         pgh->tv_usec = 0;
2819     } else {
2820         /*
2821          * pgh->tv_sec wraps in y2106 when interpreted as unsigned
2822          * as done by wireshark, or y2038 when interpreted as signed.
2823          * This is probably harmless, but if anyone wants to improve
2824          * it, we could introduce a variant that puts 64-bit nanoseconds
2825          * into the respective header bytes.
2826          * This would also be slightly faster to read.
2827          */
2828         ktime_get_real_ts64(&timestamp);
2829         pgh->tv_sec = htonl(timestamp.tv_sec);
2830         pgh->tv_usec = htonl(timestamp.tv_nsec / NSEC_PER_USEC);
2831     }
2832 }
2833
2834 static struct sk_buff *pktgen_alloc_skb(struct net_device *dev,
2835                     struct pktgen_dev *pkt_dev)
2836 {
2837     unsigned int extralen = LL_RESERVED_SPACE(dev);
2838     struct sk_buff *skb = NULL;
2839     unsigned int size;
2840
2841     size = pkt_dev->cur_pkt_size + 64 + extralen + pkt_dev->pkt_overhead;
2842     if (pkt_dev->flags & F_NODE) {
2843         int node = pkt_dev->node >= 0 ? pkt_dev->node : numa_node_id();
2844
2845         skb = __alloc_skb(NET_SKB_PAD + size, GFP_NOWAIT, 0, node);
2846         if (likely(skb)) {
2847             skb_reserve(skb, NET_SKB_PAD);
2848             skb->dev = dev;
2849         }
2850     } else {
2851          skb = __netdev_alloc_skb(dev, size, GFP_NOWAIT);
2852     }
2853
2854     /* the caller pre-fetches from skb->data and reserves for the mac hdr */
2855     if (likely(skb))
2856         skb_reserve(skb, extralen - 16);
2857
2858     return skb;
2859 }
2860
2861 static struct sk_buff *fill_packet_ipv4(struct net_device *odev,
2862                     struct pktgen_dev *pkt_dev)
2863 {
2864     struct sk_buff *skb = NULL;
2865     __u8 *eth;
2866     struct udphdr *udph;
2867     int datalen, iplen;
2868     struct iphdr *iph;
2869     __be16 protocol = htons(ETH_P_IP);
2870     __be32 *mpls;
2871     __be16 *vlan_tci = NULL;                 /* Encapsulates priority and VLAN ID */
2872     __be16 *vlan_encapsulated_proto = NULL;  /* packet type ID field (or len) for VLAN tag */
2873     __be16 *svlan_tci = NULL;                /* Encapsulates priority and SVLAN ID */
2874     __be16 *svlan_encapsulated_proto = NULL; /* packet type ID field (or len) for SVLAN tag */
2875     u16 queue_map;
2876
2877     if (pkt_dev->nr_labels)
2878         protocol = htons(ETH_P_MPLS_UC);
2879
2880     if (pkt_dev->vlan_id != 0xffff)
2881         protocol = htons(ETH_P_8021Q);
2882
2883     /* Update any of the values, used when we're incrementing various
2884      * fields.
2885      */
2886     mod_cur_headers(pkt_dev);
2887     queue_map = pkt_dev->cur_queue_map;
2888
2889     skb = pktgen_alloc_skb(odev, pkt_dev);
2890     if (!skb) {
2891         sprintf(pkt_dev->result, "No memory");
2892         return NULL;
2893     }
2894
2895     prefetchw(skb->data);
2896     skb_reserve(skb, 16);
2897
2898     /*  Reserve for ethernet and IP header  */
2899     eth = skb_push(skb, 14);
2900     mpls = skb_put(skb, pkt_dev->nr_labels * sizeof(__u32));
2901     if (pkt_dev->nr_labels)
2902         mpls_push(mpls, pkt_dev);
2903
2904     if (pkt_dev->vlan_id != 0xffff) {
2905         if (pkt_dev->svlan_id != 0xffff) {
2906             svlan_tci = skb_put(skb, sizeof(__be16));
2907             *svlan_tci = build_tci(pkt_dev->svlan_id,
2908                            pkt_dev->svlan_cfi,
2909                            pkt_dev->svlan_p);
2910             svlan_encapsulated_proto = skb_put(skb,
2911                                sizeof(__be16));
2912             *svlan_encapsulated_proto = htons(ETH_P_8021Q);
2913         }
2914         vlan_tci = skb_put(skb, sizeof(__be16));
2915         *vlan_tci = build_tci(pkt_dev->vlan_id,
2916                       pkt_dev->vlan_cfi,
2917                       pkt_dev->vlan_p);
2918         vlan_encapsulated_proto = skb_put(skb, sizeof(__be16));
2919         *vlan_encapsulated_proto = htons(ETH_P_IP);
2920     }
2921
2922     skb_reset_mac_header(skb);
2923     skb_set_network_header(skb, skb->len);
2924     iph = skb_put(skb, sizeof(struct iphdr));
2925
2926     skb_set_transport_header(skb, skb->len);
2927     udph = skb_put(skb, sizeof(struct udphdr));
2928     skb_set_queue_mapping(skb, queue_map);
2929     skb->priority = pkt_dev->skb_priority;
2930
2931     memcpy(eth, pkt_dev->hh, 12);
2932     *(__be16 *) & eth[12] = protocol;
2933
2934     /* Eth + IPh + UDPh + mpls */
2935     datalen = pkt_dev->cur_pkt_size - 14 - 20 - 8 -
2936           pkt_dev->pkt_overhead;
2937     if (datalen < 0 || datalen < sizeof(struct pktgen_hdr))
2938         datalen = sizeof(struct pktgen_hdr);
2939
2940     udph->source = htons(pkt_dev->cur_udp_src);
2941     udph->dest = htons(pkt_dev->cur_udp_dst);
2942     udph->len = htons(datalen + 8); /* DATA + udphdr */
2943     udph->check = 0;
2944
2945     iph->ihl = 5;
2946     iph->version = 4;
2947     iph->ttl = 32;
2948     iph->tos = pkt_dev->tos;
2949     iph->protocol = IPPROTO_UDP;    /* UDP */
2950     iph->saddr = pkt_dev->cur_saddr;
2951     iph->daddr = pkt_dev->cur_daddr;
2952     iph->id = htons(pkt_dev->ip_id);
2953     pkt_dev->ip_id++;
2954     iph->frag_off = 0;
2955     iplen = 20 + 8 + datalen;
2956     iph->tot_len = htons(iplen);
2957     ip_send_check(iph);
2958     skb->protocol = protocol;
2959     skb->dev = odev;
2960     skb->pkt_type = PACKET_HOST;
2961
2962     pktgen_finalize_skb(pkt_dev, skb, datalen);
2963
2964     if (!(pkt_dev->flags & F_UDPCSUM)) {
2965         skb->ip_summed = CHECKSUM_NONE;
2966     } else if (odev->features & (NETIF_F_HW_CSUM | NETIF_F_IP_CSUM)) {
2967         skb->ip_summed = CHECKSUM_PARTIAL;
2968         skb->csum = 0;
2969         udp4_hwcsum(skb, iph->saddr, iph->daddr);
2970     } else {
2971         __wsum csum = skb_checksum(skb, skb_transport_offset(skb), datalen + 8, 0);
2972
2973         /* add protocol-dependent pseudo-header */
2974         udph->check = csum_tcpudp_magic(iph->saddr, iph->daddr,
2975                         datalen + 8, IPPROTO_UDP, csum);
2976
2977         if (udph->check == 0)
2978             udph->check = CSUM_MANGLED_0;
2979     }
2980
2981 #ifdef CONFIG_XFRM
2982     if (!process_ipsec(pkt_dev, skb, protocol))
2983         return NULL;
2984 #endif
2985
2986     return skb;
2987 }
2988
2989 static struct sk_buff *fill_packet_ipv6(struct net_device *odev,
2990                     struct pktgen_dev *pkt_dev)
2991 {
2992     struct sk_buff *skb = NULL;
2993     __u8 *eth;
2994     struct udphdr *udph;
2995     int datalen, udplen;
2996     struct ipv6hdr *iph;
2997     __be16 protocol = htons(ETH_P_IPV6);
2998     __be32 *mpls;
2999     __be16 *vlan_tci = NULL;                 /* Encapsulates priority and VLAN ID */
3000     __be16 *vlan_encapsulated_proto = NULL;  /* packet type ID field (or len) for VLAN tag */
3001     __be16 *svlan_tci = NULL;                /* Encapsulates priority and SVLAN ID */
3002     __be16 *svlan_encapsulated_proto = NULL; /* packet type ID field (or len) for SVLAN tag */
3003     u16 queue_map;
3004
3005     if (pkt_dev->nr_labels)
3006         protocol = htons(ETH_P_MPLS_UC);
3007
3008     if (pkt_dev->vlan_id != 0xffff)
3009         protocol = htons(ETH_P_8021Q);
3010
3011     /* Update any of the values, used when we're incrementing various
3012      * fields.
3013      */
3014     mod_cur_headers(pkt_dev);
3015     queue_map = pkt_dev->cur_queue_map;
3016
3017     skb = pktgen_alloc_skb(odev, pkt_dev);
3018     if (!skb) {
3019         sprintf(pkt_dev->result, "No memory");
3020         return NULL;
3021     }
3022
3023     prefetchw(skb->data);
3024     skb_reserve(skb, 16);
3025
3026     /*  Reserve for ethernet and IP header  */
3027     eth = skb_push(skb, 14);
3028     mpls = skb_put(skb, pkt_dev->nr_labels * sizeof(__u32));
3029     if (pkt_dev->nr_labels)
3030         mpls_push(mpls, pkt_dev);
3031
3032     if (pkt_dev->vlan_id != 0xffff) {
3033         if (pkt_dev->svlan_id != 0xffff) {
3034             svlan_tci = skb_put(skb, sizeof(__be16));
3035             *svlan_tci = build_tci(pkt_dev->svlan_id,
3036                            pkt_dev->svlan_cfi,
3037                            pkt_dev->svlan_p);
3038             svlan_encapsulated_proto = skb_put(skb,
3039                                sizeof(__be16));
3040             *svlan_encapsulated_proto = htons(ETH_P_8021Q);
3041         }
3042         vlan_tci = skb_put(skb, sizeof(__be16));
3043         *vlan_tci = build_tci(pkt_dev->vlan_id,
3044                       pkt_dev->vlan_cfi,
3045                       pkt_dev->vlan_p);
3046         vlan_encapsulated_proto = skb_put(skb, sizeof(__be16));
3047         *vlan_encapsulated_proto = htons(ETH_P_IPV6);
3048     }
3049
3050     skb_reset_mac_header(skb);
3051     skb_set_network_header(skb, skb->len);
3052     iph = skb_put(skb, sizeof(struct ipv6hdr));
3053
3054     skb_set_transport_header(skb, skb->len);
3055     udph = skb_put(skb, sizeof(struct udphdr));
3056     skb_set_queue_mapping(skb, queue_map);
3057     skb->priority = pkt_dev->skb_priority;
3058
3059     memcpy(eth, pkt_dev->hh, 12);
3060     *(__be16 *) &eth[12] = protocol;
3061
3062     /* Eth + IPh + UDPh + mpls */
3063     datalen = pkt_dev->cur_pkt_size - 14 -
3064           sizeof(struct ipv6hdr) - sizeof(struct udphdr) -
3065           pkt_dev->pkt_overhead;
3066
3067     if (datalen < 0 || datalen < sizeof(struct pktgen_hdr)) {
3068         datalen = sizeof(struct pktgen_hdr);
3069         net_info_ratelimited("increased datalen to %d\n", datalen);
3070     }
3071
3072     udplen = datalen + sizeof(struct udphdr);
3073     udph->source = htons(pkt_dev->cur_udp_src);
3074     udph->dest = htons(pkt_dev->cur_udp_dst);
3075     udph->len = htons(udplen);
3076     udph->check = 0;
3077
3078     *(__be32 *) iph = htonl(0x60000000);    /* Version + flow */
3079
3080     if (pkt_dev->traffic_class) {
3081         /* Version + traffic class + flow (0) */
3082         *(__be32 *)iph |= htonl(0x60000000 | (pkt_dev->traffic_class << 20));
3083     }
3084
3085     iph->hop_limit = 32;
3086
3087     iph->payload_len = htons(udplen);
3088     iph->nexthdr = IPPROTO_UDP;
3089
3090     iph->daddr = pkt_dev->cur_in6_daddr;
3091     iph->saddr = pkt_dev->cur_in6_saddr;
3092
3093     skb->protocol = protocol;
3094     skb->dev = odev;
3095     skb->pkt_type = PACKET_HOST;
3096
3097     pktgen_finalize_skb(pkt_dev, skb, datalen);
3098
3099     if (!(pkt_dev->flags & F_UDPCSUM)) {
3100         skb->ip_summed = CHECKSUM_NONE;
3101     } else if (odev->features & (NETIF_F_HW_CSUM | NETIF_F_IPV6_CSUM)) {
3102         skb->ip_summed = CHECKSUM_PARTIAL;
3103         skb->csum_start = skb_transport_header(skb) - skb->head;
3104         skb->csum_offset = offsetof(struct udphdr, check);
3105         udph->check = ~csum_ipv6_magic(&iph->saddr, &iph->daddr, udplen, IPPROTO_UDP, 0);
3106     } else {
3107         __wsum csum = skb_checksum(skb, skb_transport_offset(skb), udplen, 0);
3108
3109         /* add protocol-dependent pseudo-header */
3110         udph->check = csum_ipv6_magic(&iph->saddr, &iph->daddr, udplen, IPPROTO_UDP, csum);
3111
3112         if (udph->check == 0)
3113             udph->check = CSUM_MANGLED_0;
3114     }
3115
3116     return skb;
3117 }
3118
3119 static struct sk_buff *fill_packet(struct net_device *odev,
3120                    struct pktgen_dev *pkt_dev)
3121 {
3122     if (pkt_dev->flags & F_IPV6)
3123         return fill_packet_ipv6(odev, pkt_dev);
3124     else
3125         return fill_packet_ipv4(odev, pkt_dev);
3126 }
3127
3128 static void pktgen_clear_counters(struct pktgen_dev *pkt_dev)
3129 {
3130     pkt_dev->seq_num = 1;
3131     pkt_dev->idle_acc = 0;
3132     pkt_dev->sofar = 0;
3133     pkt_dev->tx_bytes = 0;
3134     pkt_dev->errors = 0;
3135 }
3136
3137 /* Set up structure for sending pkts, clear counters */
3138
3139 static void pktgen_run(struct pktgen_thread *t)
3140 {
3141     struct pktgen_dev *pkt_dev;
3142     int started = 0;
3143
3144     func_enter();
3145
3146     rcu_read_lock();
3147     list_for_each_entry_rcu(pkt_dev, &t->if_list, list) {
3148
3149         /*
3150          * setup odev and create initial packet.
3151          */
3152         pktgen_setup_inject(pkt_dev);
3153
3154         if (pkt_dev->odev) {
3155             pktgen_clear_counters(pkt_dev);
3156             pkt_dev->skb = NULL;
3157             pkt_dev->started_at = pkt_dev->next_tx = ktime_get();
3158
3159             set_pkt_overhead(pkt_dev);
3160
3161             strcpy(pkt_dev->result, "Starting");
3162             pkt_dev->running = 1;   /* Cranke yeself! */
3163             started++;
3164         } else
3165             strcpy(pkt_dev->result, "Error starting");
3166     }
3167     rcu_read_unlock();
3168     if (started)
3169         t->control &= ~(T_STOP);
3170 }
3171
3172 static void pktgen_handle_all_threads(struct pktgen_net *pn, u32 flags)
3173 {
3174     struct pktgen_thread *t;
3175
3176     mutex_lock(&pktgen_thread_lock);
3177
3178     list_for_each_entry(t, &pn->pktgen_threads, th_list)
3179         t->control |= (flags);
3180
3181     mutex_unlock(&pktgen_thread_lock);
3182 }
3183
3184 static void pktgen_stop_all_threads(struct pktgen_net *pn)
3185 {
3186     func_enter();
3187
3188     pktgen_handle_all_threads(pn, T_STOP);
3189 }
3190
3191 static int thread_is_running(const struct pktgen_thread *t)
3192 {
3193     const struct pktgen_dev *pkt_dev;
3194
3195     rcu_read_lock();
3196     list_for_each_entry_rcu(pkt_dev, &t->if_list, list)
3197         if (pkt_dev->running) {
3198             rcu_read_unlock();
3199             return 1;
3200         }
3201     rcu_read_unlock();
3202     return 0;
3203 }
3204
3205 static int pktgen_wait_thread_run(struct pktgen_thread *t)
3206 {
3207     while (thread_is_running(t)) {
3208
3209         /* note: 't' will still be around even after the unlock/lock
3210          * cycle because pktgen_thread threads are only cleared at
3211          * net exit
3212          */
3213         mutex_unlock(&pktgen_thread_lock);
3214         msleep_interruptible(100);
3215         mutex_lock(&pktgen_thread_lock);
3216
3217         if (signal_pending(current))
3218             goto signal;
3219     }
3220     return 1;
3221 signal:
3222     return 0;
3223 }
3224
3225 static int pktgen_wait_all_threads_run(struct pktgen_net *pn)
3226 {
3227     struct pktgen_thread *t;
3228     int sig = 1;
3229
3230     /* prevent from racing with rmmod */
3231     if (!try_module_get(THIS_MODULE))
3232         return sig;
3233
3234     mutex_lock(&pktgen_thread_lock);
3235
3236     list_for_each_entry(t, &pn->pktgen_threads, th_list) {
3237         sig = pktgen_wait_thread_run(t);
3238         if (sig == 0)
3239             break;
3240     }
3241
3242     if (sig == 0)
3243         list_for_each_entry(t, &pn->pktgen_threads, th_list)
3244             t->control |= (T_STOP);
3245
3246     mutex_unlock(&pktgen_thread_lock);
3247     module_put(THIS_MODULE);
3248     return sig;
3249 }
3250
3251 static void pktgen_run_all_threads(struct pktgen_net *pn)
3252 {
3253     func_enter();
3254
3255     pktgen_handle_all_threads(pn, T_RUN);
3256
3257     /* Propagate thread->control  */
3258     schedule_timeout_interruptible(msecs_to_jiffies(125));
3259
3260     pktgen_wait_all_threads_run(pn);
3261 }
3262
3263 static void pktgen_reset_all_threads(struct pktgen_net *pn)
3264 {
3265     func_enter();
3266
3267     pktgen_handle_all_threads(pn, T_REMDEVALL);
3268
3269     /* Propagate thread->control  */
3270     schedule_timeout_interruptible(msecs_to_jiffies(125));
3271
3272     pktgen_wait_all_threads_run(pn);
3273 }
3274
3275 static void show_results(struct pktgen_dev *pkt_dev, int nr_frags)
3276 {
3277     __u64 bps, mbps, pps;
3278     char *p = pkt_dev->result;
3279     ktime_t elapsed = ktime_sub(pkt_dev->stopped_at,
3280                     pkt_dev->started_at);
3281     ktime_t idle = ns_to_ktime(pkt_dev->idle_acc);
3282
3283     p += sprintf(p, "OK: %llu(c%llu+d%llu) usec, %llu (%dbyte,%dfrags)\n",
3284              (unsigned long long)ktime_to_us(elapsed),
3285              (unsigned long long)ktime_to_us(ktime_sub(elapsed, idle)),
3286              (unsigned long long)ktime_to_us(idle),
3287              (unsigned long long)pkt_dev->sofar,
3288              pkt_dev->cur_pkt_size, nr_frags);
3289
3290     pps = div64_u64(pkt_dev->sofar * NSEC_PER_SEC,
3291             ktime_to_ns(elapsed));
3292
3293     if (pkt_dev->n_imix_entries > 0) {
3294         int i;
3295         struct imix_pkt *entry;
3296
3297         bps = 0;
3298         for (i = 0; i < pkt_dev->n_imix_entries; i++) {
3299             entry = &pkt_dev->imix_entries[i];
3300             bps += entry->size * entry->count_so_far;
3301         }
3302         bps = div64_u64(bps * 8 * NSEC_PER_SEC, ktime_to_ns(elapsed));
3303     } else {
3304         bps = pps * 8 * pkt_dev->cur_pkt_size;
3305     }
3306
3307     mbps = bps;
3308     do_div(mbps, 1000000);
3309     p += sprintf(p, "  %llupps %lluMb/sec (%llubps) errors: %llu",
3310              (unsigned long long)pps,
3311              (unsigned long long)mbps,
3312              (unsigned long long)bps,
3313              (unsigned long long)pkt_dev->errors);
3314 }
3315
3316 /* Set stopped-at timer, remove from running list, do counters & statistics */
3317 static int pktgen_stop_device(struct pktgen_dev *pkt_dev)
3318 {
3319     int nr_frags = pkt_dev->skb ? skb_shinfo(pkt_dev->skb)->nr_frags : -1;
3320
3321     if (!pkt_dev->running) {
3322         pr_warn("interface: %s is already stopped\n",
3323             pkt_dev->odevname);
3324         return -EINVAL;
3325     }
3326
3327     pkt_dev->running = 0;
3328     kfree_skb(pkt_dev->skb);
3329     pkt_dev->skb = NULL;
3330     pkt_dev->stopped_at = ktime_get();
3331
3332     show_results(pkt_dev, nr_frags);
3333
3334     return 0;
3335 }
3336
3337 static struct pktgen_dev *next_to_run(struct pktgen_thread *t)
3338 {
3339     struct pktgen_dev *pkt_dev, *best = NULL;
3340
3341     rcu_read_lock();
3342     list_for_each_entry_rcu(pkt_dev, &t->if_list, list) {
3343         if (!pkt_dev->running)
3344             continue;
3345         if (best == NULL)
3346             best = pkt_dev;
3347         else if (ktime_compare(pkt_dev->next_tx, best->next_tx) < 0)
3348             best = pkt_dev;
3349     }
3350     rcu_read_unlock();
3351
3352     return best;
3353 }
3354
3355 static void pktgen_stop(struct pktgen_thread *t)
3356 {
3357     struct pktgen_dev *pkt_dev;
3358
3359     func_enter();
3360
3361     rcu_read_lock();
3362
3363     list_for_each_entry_rcu(pkt_dev, &t->if_list, list) {
3364         pktgen_stop_device(pkt_dev);
3365     }
3366
3367     rcu_read_unlock();
3368 }
3369
3370 /*
3371  * one of our devices needs to be removed - find it
3372  * and remove it
3373  */
3374 static void pktgen_rem_one_if(struct pktgen_thread *t)
3375 {
3376     struct list_head *q, *n;
3377     struct pktgen_dev *cur;
3378
3379     func_enter();
3380
3381     list_for_each_safe(q, n, &t->if_list) {
3382         cur = list_entry(q, struct pktgen_dev, list);
3383
3384         if (!cur->removal_mark)
3385             continue;
3386
3387         kfree_skb(cur->skb);
3388         cur->skb = NULL;
3389
3390         pktgen_remove_device(t, cur);
3391
3392         break;
3393     }
3394 }
3395
3396 static void pktgen_rem_all_ifs(struct pktgen_thread *t)
3397 {
3398     struct list_head *q, *n;
3399     struct pktgen_dev *cur;
3400
3401     func_enter();
3402
3403     /* Remove all devices, free mem */
3404
3405     list_for_each_safe(q, n, &t->if_list) {
3406         cur = list_entry(q, struct pktgen_dev, list);
3407
3408         kfree_skb(cur->skb);
3409         cur->skb = NULL;
3410
3411         pktgen_remove_device(t, cur);
3412     }
3413 }
3414
3415 static void pktgen_rem_thread(struct pktgen_thread *t)
3416 {
3417     /* Remove from the thread list */
3418     remove_proc_entry(t->tsk->comm, t->net->proc_dir);
3419 }
3420
3421 static void pktgen_resched(struct pktgen_dev *pkt_dev)
3422 {
3423     ktime_t idle_start = ktime_get();
3424     schedule();
3425     pkt_dev->idle_acc += ktime_to_ns(ktime_sub(ktime_get(), idle_start));
3426 }
3427
3428 static void pktgen_wait_for_skb(struct pktgen_dev *pkt_dev)
3429 {
3430     ktime_t idle_start = ktime_get();
3431
3432     while (refcount_read(&(pkt_dev->skb->users)) != 1) {
3433         if (signal_pending(current))
3434             break;
3435
3436         if (need_resched())
3437             pktgen_resched(pkt_dev);
3438         else
3439             cpu_relax();
3440     }
3441     pkt_dev->idle_acc += ktime_to_ns(ktime_sub(ktime_get(), idle_start));
3442 }
3443
3444 static void pktgen_xmit(struct pktgen_dev *pkt_dev)
3445 {
3446     unsigned int burst = READ_ONCE(pkt_dev->burst);
3447     struct net_device *odev = pkt_dev->odev;
3448     struct netdev_queue *txq;
3449     struct sk_buff *skb;
3450     int ret;
3451
3452     /* If device is offline, then don't send */
3453     if (unlikely(!netif_running(odev) || !netif_carrier_ok(odev))) {
3454         pktgen_stop_device(pkt_dev);
3455         return;
3456     }
3457
3458     /* This is max DELAY, this has special meaning of
3459      * "never transmit"
3460      */
3461     if (unlikely(pkt_dev->delay == ULLONG_MAX)) {
3462         pkt_dev->next_tx = ktime_add_ns(ktime_get(), ULONG_MAX);
3463         return;
3464     }
3465
3466     /* If no skb or clone count exhausted then get new one */
3467     if (!pkt_dev->skb || (pkt_dev->last_ok &&
3468                   ++pkt_dev->clone_count >= pkt_dev->clone_skb)) {
3469         /* build a new pkt */
3470         kfree_skb(pkt_dev->skb);
3471
3472         pkt_dev->skb = fill_packet(odev, pkt_dev);
3473         if (pkt_dev->skb == NULL) {
3474             pr_err("ERROR: couldn't allocate skb in fill_packet\n");
3475             schedule();
3476             pkt_dev->clone_count--; /* back out increment, OOM */
3477             return;
3478         }
3479         pkt_dev->last_pkt_size = pkt_dev->skb->len;
3480         pkt_dev->clone_count = 0;   /* reset counter */
3481     }
3482
3483     if (pkt_dev->delay && pkt_dev->last_ok)
3484         spin(pkt_dev, pkt_dev->next_tx);
3485
3486     if (pkt_dev->xmit_mode == M_NETIF_RECEIVE) {
3487         skb = pkt_dev->skb;
3488         skb->protocol = eth_type_trans(skb, skb->dev);
3489         refcount_add(burst, &skb->users);
3490         local_bh_disable();
3491         do {
3492             ret = netif_receive_skb(skb);
3493             if (ret == NET_RX_DROP)
3494                 pkt_dev->errors++;
3495             pkt_dev->sofar++;
3496             pkt_dev->seq_num++;
3497             if (refcount_read(&skb->users) != burst) {
3498                 /* skb was queued by rps/rfs or taps,
3499                  * so cannot reuse this skb
3500                  */
3501                 WARN_ON(refcount_sub_and_test(burst - 1, &skb->users));
3502                 /* get out of the loop and wait
3503                  * until skb is consumed
3504                  */
3505                 break;
3506             }
3507             /* skb was 'freed' by stack, so clean few
3508              * bits and reuse it
3509              */
3510             skb_reset_redirect(skb);
3511         } while (--burst > 0);
3512         goto out; /* Skips xmit_mode M_START_XMIT */
3513     } else if (pkt_dev->xmit_mode == M_QUEUE_XMIT) {
3514         local_bh_disable();
3515         refcount_inc(&pkt_dev->skb->users);
3516
3517         ret = dev_queue_xmit(pkt_dev->skb);
3518         switch (ret) {
3519         case NET_XMIT_SUCCESS:
3520             pkt_dev->sofar++;
3521             pkt_dev->seq_num++;
3522             pkt_dev->tx_bytes += pkt_dev->last_pkt_size;
3523             break;
3524         case NET_XMIT_DROP:
3525         case NET_XMIT_CN:
3526         /* These are all valid return codes for a qdisc but
3527          * indicate packets are being dropped or will likely
3528          * be dropped soon.
3529          */
3530         case NETDEV_TX_BUSY:
3531         /* qdisc may call dev_hard_start_xmit directly in cases
3532          * where no queues exist e.g. loopback device, virtual
3533          * devices, etc. In this case we need to handle
3534          * NETDEV_TX_ codes.
3535          */
3536         default:
3537             pkt_dev->errors++;
3538             net_info_ratelimited("%s xmit error: %d\n",
3539                          pkt_dev->odevname, ret);
3540             break;
3541         }
3542         goto out;
3543     }
3544
3545     txq = skb_get_tx_queue(odev, pkt_dev->skb);
3546
3547     local_bh_disable();
3548
3549     HARD_TX_LOCK(odev, txq, smp_processor_id());
3550
3551     if (unlikely(netif_xmit_frozen_or_drv_stopped(txq))) {
3552         pkt_dev->last_ok = 0;
3553         goto unlock;
3554     }
3555     refcount_add(burst, &pkt_dev->skb->users);
3556
3557 xmit_more:
3558     ret = netdev_start_xmit(pkt_dev->skb, odev, txq, --burst > 0);
3559
3560     switch (ret) {
3561     case NETDEV_TX_OK:
3562         pkt_dev->last_ok = 1;
3563         pkt_dev->sofar++;
3564         pkt_dev->seq_num++;
3565         pkt_dev->tx_bytes += pkt_dev->last_pkt_size;
3566         if (burst > 0 && !netif_xmit_frozen_or_drv_stopped(txq))
3567             goto xmit_more;
3568         break;
3569     case NET_XMIT_DROP:
3570     case NET_XMIT_CN:
3571         /* skb has been consumed */
3572         pkt_dev->errors++;
3573         break;
3574     default: /* Drivers are not supposed to return other values! */
3575         net_info_ratelimited("%s xmit error: %d\n",
3576                      pkt_dev->odevname, ret);
3577         pkt_dev->errors++;
3578         fallthrough;
3579     case NETDEV_TX_BUSY:
3580         /* Retry it next time */
3581         refcount_dec(&(pkt_dev->skb->users));
3582         pkt_dev->last_ok = 0;
3583     }
3584     if (unlikely(burst))
3585         WARN_ON(refcount_sub_and_test(burst, &pkt_dev->skb->users));
3586 unlock:
3587     HARD_TX_UNLOCK(odev, txq);
3588
3589 out:
3590     local_bh_enable();
3591
3592     /* If pkt_dev->count is zero, then run forever */
3593     if ((pkt_dev->count != 0) && (pkt_dev->sofar >= pkt_dev->count)) {
3594         pktgen_wait_for_skb(pkt_dev);
3595
3596         /* Done with this */
3597         pktgen_stop_device(pkt_dev);
3598     }
3599 }
3600
3601 /*
3602  * Main loop of the thread goes here
3603  */
3604
3605 static int pktgen_thread_worker(void *arg)
3606 {
3607     struct pktgen_thread *t = arg;
3608     struct pktgen_dev *pkt_dev = NULL;
3609     int cpu = t->cpu;
3610
3611     WARN_ON(smp_processor_id() != cpu);
3612
3613     init_waitqueue_head(&t->queue);
3614     complete(&t->start_done);
3615
3616     pr_debug("starting pktgen/%d:  pid=%d\n", cpu, task_pid_nr(current));
3617
3618     set_freezable();
3619
3620     while (!kthread_should_stop()) {
3621         pkt_dev = next_to_run(t);
3622
3623         if (unlikely(!pkt_dev && t->control == 0)) {
3624             if (t->net->pktgen_exiting)
3625                 break;
3626             wait_event_interruptible_timeout(t->queue,
3627                              t->control != 0,
3628                              HZ/10);
3629             try_to_freeze();
3630             continue;
3631         }
3632
3633         if (likely(pkt_dev)) {
3634             pktgen_xmit(pkt_dev);
3635
3636             if (need_resched())
3637                 pktgen_resched(pkt_dev);
3638             else
3639                 cpu_relax();
3640         }
3641
3642         if (t->control & T_STOP) {
3643             pktgen_stop(t);
3644             t->control &= ~(T_STOP);
3645         }
3646
3647         if (t->control & T_RUN) {
3648             pktgen_run(t);
3649             t->control &= ~(T_RUN);
3650         }
3651
3652         if (t->control & T_REMDEVALL) {
3653             pktgen_rem_all_ifs(t);
3654             t->control &= ~(T_REMDEVALL);
3655         }
3656
3657         if (t->control & T_REMDEV) {
3658             pktgen_rem_one_if(t);
3659             t->control &= ~(T_REMDEV);
3660         }
3661
3662         try_to_freeze();
3663     }
3664
3665     pr_debug("%s stopping all device\n", t->tsk->comm);
3666     pktgen_stop(t);
3667
3668     pr_debug("%s removing all device\n", t->tsk->comm);
3669     pktgen_rem_all_ifs(t);
3670
3671     pr_debug("%s removing thread\n", t->tsk->comm);
3672     pktgen_rem_thread(t);
3673
3674     return 0;
3675 }
3676
3677 static struct pktgen_dev *pktgen_find_dev(struct pktgen_thread *t,
3678                       const char *ifname, bool exact)
3679 {
3680     struct pktgen_dev *p, *pkt_dev = NULL;
3681     size_t len = strlen(ifname);
3682
3683     rcu_read_lock();
3684     list_for_each_entry_rcu(p, &t->if_list, list)
3685         if (strncmp(p->odevname, ifname, len) == 0) {
3686             if (p->odevname[len]) {
3687                 if (exact || p->odevname[len] != '@')
3688                     continue;
3689             }
3690             pkt_dev = p;
3691             break;
3692         }
3693
3694     rcu_read_unlock();
3695     pr_debug("find_dev(%s) returning %p\n", ifname, pkt_dev);
3696     return pkt_dev;
3697 }
3698
3699 /*
3700  * Adds a dev at front of if_list.
3701  */
3702
3703 static int add_dev_to_thread(struct pktgen_thread *t,
3704                  struct pktgen_dev *pkt_dev)
3705 {
3706     int rv = 0;
3707
3708     /* This function cannot be called concurrently, as its called
3709      * under pktgen_thread_lock mutex, but it can run from
3710      * userspace on another CPU than the kthread.  The if_lock()
3711      * is used here to sync with concurrent instances of
3712      * _rem_dev_from_if_list() invoked via kthread, which is also
3713      * updating the if_list */
3714     if_lock(t);
3715
3716     if (pkt_dev->pg_thread) {
3717         pr_err("ERROR: already assigned to a thread\n");
3718         rv = -EBUSY;
3719         goto out;
3720     }
3721
3722     pkt_dev->running = 0;
3723     pkt_dev->pg_thread = t;
3724     list_add_rcu(&pkt_dev->list, &t->if_list);
3725
3726 out:
3727     if_unlock(t);
3728     return rv;
3729 }
3730
3731 /* Called under thread lock */
3732
3733 static int pktgen_add_device(struct pktgen_thread *t, const char *ifname)
3734 {
3735     struct pktgen_dev *pkt_dev;
3736     int err;
3737     int node = cpu_to_node(t->cpu);
3738
3739     /* We don't allow a device to be on several threads */
3740
3741     pkt_dev = __pktgen_NN_threads(t->net, ifname, FIND);
3742     if (pkt_dev) {
3743         pr_err("ERROR: interface already used\n");
3744         return -EBUSY;
3745     }
3746
3747     pkt_dev = kzalloc_node(sizeof(struct pktgen_dev), GFP_KERNEL, node);
3748     if (!pkt_dev)
3749         return -ENOMEM;
3750
3751     strcpy(pkt_dev->odevname, ifname);
3752     pkt_dev->flows = vzalloc_node(array_size(MAX_CFLOWS,
3753                          sizeof(struct flow_state)),
3754                       node);
3755     if (pkt_dev->flows == NULL) {
3756         kfree(pkt_dev);
3757         return -ENOMEM;
3758     }
3759
3760     pkt_dev->removal_mark = 0;
3761     pkt_dev->nfrags = 0;
3762     pkt_dev->delay = pg_delay_d;
3763     pkt_dev->count = pg_count_d;
3764     pkt_dev->sofar = 0;
3765     pkt_dev->udp_src_min = 9;   /* sink port */
3766     pkt_dev->udp_src_max = 9;
3767     pkt_dev->udp_dst_min = 9;
3768     pkt_dev->udp_dst_max = 9;
3769     pkt_dev->vlan_p = 0;
3770     pkt_dev->vlan_cfi = 0;
3771     pkt_dev->vlan_id = 0xffff;
3772     pkt_dev->svlan_p = 0;
3773     pkt_dev->svlan_cfi = 0;
3774     pkt_dev->svlan_id = 0xffff;
3775     pkt_dev->burst = 1;
3776     pkt_dev->node = NUMA_NO_NODE;
3777
3778     err = pktgen_setup_dev(t->net, pkt_dev, ifname);
3779     if (err)
3780         goto out1;
3781     if (pkt_dev->odev->priv_flags & IFF_TX_SKB_SHARING)
3782         pkt_dev->clone_skb = pg_clone_skb_d;
3783
3784     pkt_dev->entry = proc_create_data(ifname, 0600, t->net->proc_dir,
3785                       &pktgen_if_proc_ops, pkt_dev);
3786     if (!pkt_dev->entry) {
3787         pr_err("cannot create %s/%s procfs entry\n",
3788                PG_PROC_DIR, ifname);
3789         err = -EINVAL;
3790         goto out2;
3791     }
3792 #ifdef CONFIG_XFRM
3793     pkt_dev->ipsmode = XFRM_MODE_TRANSPORT;
3794     pkt_dev->ipsproto = IPPROTO_ESP;
3795
3796     /* xfrm tunnel mode needs additional dst to extract outter
3797      * ip header protocol/ttl/id field, here creat a phony one.
3798      * instead of looking for a valid rt, which definitely hurting
3799      * performance under such circumstance.
3800      */
3801     pkt_dev->dstops.family = AF_INET;
3802     pkt_dev->xdst.u.dst.dev = pkt_dev->odev;
3803     dst_init_metrics(&pkt_dev->xdst.u.dst, pktgen_dst_metrics, false);
3804     pkt_dev->xdst.child = &pkt_dev->xdst.u.dst;
3805     pkt_dev->xdst.u.dst.ops = &pkt_dev->dstops;
3806 #endif
3807
3808     return add_dev_to_thread(t, pkt_dev);
3809 out2:
3810     netdev_put(pkt_dev->odev, &pkt_dev->dev_tracker);
3811 out1:
3812 #ifdef CONFIG_XFRM
3813     free_SAs(pkt_dev);
3814 #endif
3815     vfree(pkt_dev->flows);
3816     kfree(pkt_dev);
3817     return err;
3818 }
3819
3820 static int __net_init pktgen_create_thread(int cpu, struct pktgen_net *pn)
3821 {
3822     struct pktgen_thread *t;
3823     struct proc_dir_entry *pe;
3824     struct task_struct *p;
3825
3826     t = kzalloc_node(sizeof(struct pktgen_thread), GFP_KERNEL,
3827              cpu_to_node(cpu));
3828     if (!t) {
3829         pr_err("ERROR: out of memory, can't create new thread\n");
3830         return -ENOMEM;
3831     }
3832
3833     mutex_init(&t->if_lock);
3834     t->cpu = cpu;
3835
3836     INIT_LIST_HEAD(&t->if_list);
3837
3838     list_add_tail(&t->th_list, &pn->pktgen_threads);
3839     init_completion(&t->start_done);
3840
3841     p = kthread_create_on_node(pktgen_thread_worker,
3842                    t,
3843                    cpu_to_node(cpu),
3844                    "kpktgend_%d", cpu);
3845     if (IS_ERR(p)) {
3846         pr_err("kthread_create_on_node() failed for cpu %d\n", t->cpu);
3847         list_del(&t->th_list);
3848         kfree(t);
3849         return PTR_ERR(p);
3850     }
3851     kthread_bind(p, cpu);
3852     t->tsk = p;
3853
3854     pe = proc_create_data(t->tsk->comm, 0600, pn->proc_dir,
3855                   &pktgen_thread_proc_ops, t);
3856     if (!pe) {
3857         pr_err("cannot create %s/%s procfs entry\n",
3858                PG_PROC_DIR, t->tsk->comm);
3859         kthread_stop(p);
3860         list_del(&t->th_list);
3861         kfree(t);
3862         return -EINVAL;
3863     }
3864
3865     t->net = pn;
3866     get_task_struct(p);
3867     wake_up_process(p);
3868     wait_for_completion(&t->start_done);
3869
3870     return 0;
3871 }
3872
3873 /*
3874  * Removes a device from the thread if_list.
3875  */
3876 static void _rem_dev_from_if_list(struct pktgen_thread *t,
3877                   struct pktgen_dev *pkt_dev)
3878 {
3879     struct list_head *q, *n;
3880     struct pktgen_dev *p;
3881
3882     if_lock(t);
3883     list_for_each_safe(q, n, &t->if_list) {
3884         p = list_entry(q, struct pktgen_dev, list);
3885         if (p == pkt_dev)
3886             list_del_rcu(&p->list);
3887     }
3888     if_unlock(t);
3889 }
3890
3891 static int pktgen_remove_device(struct pktgen_thread *t,
3892                 struct pktgen_dev *pkt_dev)
3893 {
3894     pr_debug("remove_device pkt_dev=%p\n", pkt_dev);
3895
3896     if (pkt_dev->running) {
3897         pr_warn("WARNING: trying to remove a running interface, stopping it now\n");
3898         pktgen_stop_device(pkt_dev);
3899     }
3900
3901     /* Dis-associate from the interface */
3902
3903     if (pkt_dev->odev) {
3904         netdev_put(pkt_dev->odev, &pkt_dev->dev_tracker);
3905         pkt_dev->odev = NULL;
3906     }
3907
3908     /* Remove proc before if_list entry, because add_device uses
3909      * list to determine if interface already exist, avoid race
3910      * with proc_create_data() */
3911     proc_remove(pkt_dev->entry);
3912
3913     /* And update the thread if_list */
3914     _rem_dev_from_if_list(t, pkt_dev);
3915
3916 #ifdef CONFIG_XFRM
3917     free_SAs(pkt_dev);
3918 #endif
3919     vfree(pkt_dev->flows);
3920     if (pkt_dev->page)
3921         put_page(pkt_dev->page);
3922     kfree_rcu(pkt_dev, rcu);
3923     return 0;
3924 }
3925
3926 static int __net_init pg_net_init(struct net *net)
3927 {
3928     struct pktgen_net *pn = net_generic(net, pg_net_id);
3929     struct proc_dir_entry *pe;
3930     int cpu, ret = 0;
3931
3932     pn->net = net;
3933     INIT_LIST_HEAD(&pn->pktgen_threads);
3934     pn->pktgen_exiting = false;
3935     pn->proc_dir = proc_mkdir(PG_PROC_DIR, pn->net->proc_net);
3936     if (!pn->proc_dir) {
3937         pr_warn("cannot create /proc/net/%s\n", PG_PROC_DIR);
3938         return -ENODEV;
3939     }
3940     pe = proc_create(PGCTRL, 0600, pn->proc_dir, &pktgen_proc_ops);
3941     if (pe == NULL) {
3942         pr_err("cannot create %s procfs entry\n", PGCTRL);
3943         ret = -EINVAL;
3944         goto remove;
3945     }
3946
3947     for_each_online_cpu(cpu) {
3948         int err;
3949
3950         err = pktgen_create_thread(cpu, pn);
3951         if (err)
3952             pr_warn("Cannot create thread for cpu %d (%d)\n",
3953                    cpu, err);
3954     }
3955
3956     if (list_empty(&pn->pktgen_threads)) {
3957         pr_err("Initialization failed for all threads\n");
3958         ret = -ENODEV;
3959         goto remove_entry;
3960     }
3961
3962     return 0;
3963
3964 remove_entry:
3965     remove_proc_entry(PGCTRL, pn->proc_dir);
3966 remove:
3967     remove_proc_entry(PG_PROC_DIR, pn->net->proc_net);
3968     return ret;
3969 }
3970
3971 static void __net_exit pg_net_exit(struct net *net)
3972 {
3973     struct pktgen_net *pn = net_generic(net, pg_net_id);
3974     struct pktgen_thread *t;
3975     struct list_head *q, *n;
3976     LIST_HEAD(list);
3977
3978     /* Stop all interfaces & threads */
3979     pn->pktgen_exiting = true;
3980
3981     mutex_lock(&pktgen_thread_lock);
3982     list_splice_init(&pn->pktgen_threads, &list);
3983     mutex_unlock(&pktgen_thread_lock);
3984
3985     list_for_each_safe(q, n, &list) {
3986         t = list_entry(q, struct pktgen_thread, th_list);
3987         list_del(&t->th_list);
3988         kthread_stop(t->tsk);
3989         put_task_struct(t->tsk);
3990         kfree(t);
3991     }
3992
3993     remove_proc_entry(PGCTRL, pn->proc_dir);
3994     remove_proc_entry(PG_PROC_DIR, pn->net->proc_net);
3995 }
3996
3997 static struct pernet_operations pg_net_ops = {
3998     .init = pg_net_init,
3999     .exit = pg_net_exit,
4000     .id   = &pg_net_id,
4001     .size = sizeof(struct pktgen_net),
4002 };
4003
4004 static int __init pg_init(void)
4005 {
4006     int ret = 0;
4007
4008     pr_info("%s", version);
4009     ret = register_pernet_subsys(&pg_net_ops);
4010     if (ret)
4011         return ret;
4012     ret = register_netdevice_notifier(&pktgen_notifier_block);
4013     if (ret)
4014         unregister_pernet_subsys(&pg_net_ops);
4015
4016     return ret;
4017 }
4018
4019 static void __exit pg_cleanup(void)
4020 {
4021     unregister_netdevice_notifier(&pktgen_notifier_block);
4022     unregister_pernet_subsys(&pg_net_ops);
4023     /* Don't need rcu_barrier() due to use of kfree_rcu() */
4024 }
4025
4026 module_init(pg_init);
4027 module_exit(pg_cleanup);
4028
4029 MODULE_AUTHOR("Robert Olsson <robert.olsson@its.uu.se>");
4030 MODULE_DESCRIPTION("Packet Generator tool");
4031 MODULE_LICENSE("GPL");
4032 MODULE_VERSION(VERSION);
4033 module_param(pg_count_d, int, 0);
4034 MODULE_PARM_DESC(pg_count_d, "Default number of packets to inject");
4035 module_param(pg_delay_d, int, 0);
4036 MODULE_PARM_DESC(pg_delay_d, "Default delay between packets (nanoseconds)");
4037 module_param(pg_clone_skb_d, int, 0);
4038 MODULE_PARM_DESC(pg_clone_skb_d, "Default number of copies of the same packet");
4039 module_param(debug, int, 0);
4040 MODULE_PARM_DESC(debug, "Enable debugging of pktgen module");