8The name "usbmon" in lowercase refers to a facility in kernel which is
   9used to collect traces of I/O on the USB bus. This function is analogous
  10to a packet socket used by network monitoring tools such as tcpdump(1)
  11or Ethereal. Similarly, it is expected that a tool such as usbdump or
  12USBMon (with uppercase letters) is used to examine raw traces produced
  13by usbmon.
  15The usbmon reports requests made by peripheral-specific drivers to Host
  16Controller Drivers (HCD). So, if HCD is buggy, the traces reported by
  17usbmon may not correspond to bus transactions precisely. This is the same
  18situation as with tcpdump.
  20Two APIs are currently implemented: "text" and "binary". The binary API
  21is available through a character device in /dev namespace and is an ABI.
  22The text API is deprecated since 2.6.35, but available for convenience.
  24How to use usbmon to collect raw text traces
  27Unlike the packet socket, usbmon has an interface which provides traces
  28in a text format. This is used for two purposes. First, it serves as a
  29common trace exchange format for tools while more sophisticated formats
  30are finalized. Second, humans can read it in case tools are not available.
  32To collect a raw text trace, execute following steps.
  341. Prepare
  37Mount debugfs (it has to be enabled in your kernel configuration), and
  38load the usbmon module (if built as module). The second step is skipped
  39if usbmon is built into the kernel::
  41        # mount -t debugfs none_debugs /sys/kernel/debug
  42        # modprobe usbmon
  43        #
  45Verify that bus sockets are present:
  47        # ls /sys/kernel/debug/usb/usbmon
  48        0s  0u  1s  1t  1u  2s  2t  2u  3s  3t  3u  4s  4t  4u
  49        #
  51Now you can choose to either use the socket '0u' (to capture packets on all
  52buses), and skip to step #3, or find the bus used by your device with step #2.
  53This allows to filter away annoying devices that talk continuously.
  552. Find which bus connects to the desired device
  58Run "cat /sys/kernel/debug/usb/devices", and find the T-line which corresponds
  59to the device. Usually you do it by looking for the vendor string. If you have
  60many similar devices, unplug one and compare the two
  61/sys/kernel/debug/usb/devices outputs. The T-line will have a bus number.
  65  T:  Bus=03 Lev=01 Prnt=01 Port=00 Cnt=01 Dev#=  2 Spd=12  MxCh= 0
  66  D:  Ver= 1.10 Cls=00(>ifc ) Sub=00 Prot=00 MxPS= 8 #Cfgs=  1
  67  P:  Vendor=0557 ProdID=2004 Rev= 1.00
  68  S:  Manufacturer=ATEN
  69  S:  Product=UC100KM V2.00
  71"Bus=03" means it's bus 3. Alternatively, you can look at the output from
  72"lsusb" and get the bus number from the appropriate line. Example:
  74Bus 003 Device 002: ID 0557:2004 ATEN UC100KM V2.00
  763. Start 'cat'
  81        # cat /sys/kernel/debug/usb/usbmon/3u > /tmp/1.mon.out
  83to listen on a single bus, otherwise, to listen on all buses, type::
  85        # cat /sys/kernel/debug/usb/usbmon/0u > /tmp/1.mon.out
  87This process will read until it is killed. Naturally, the output can be
  88redirected to a desirable location. This is preferred, because it is going
  89to be quite long.
  914. Perform the desired operation on the USB bus
  94This is where you do something that creates the traffic: plug in a flash key,
  95copy files, control a webcam, etc.
  975. Kill cat
 100Usually it's done with a keyboard interrupt (Control-C).
 102At this point the output file (/tmp/1.mon.out in this example) can be saved,
 103sent by e-mail, or inspected with a text editor. In the last case make sure
 104that the file size is not excessive for your favourite editor.
 106Raw text data format
 109Two formats are supported currently: the original, or '1t' format, and
 110the '1u' format. The '1t' format is deprecated in kernel 2.6.21. The '1u'
 111format adds a few fields, such as ISO frame descriptors, interval, etc.
 112It produces slightly longer lines, but otherwise is a perfect superset
 113of '1t' format.
 115If it is desired to recognize one from the other in a program, look at the
 116"address" word (see below), where '1u' format adds a bus number. If 2 colons
 117are present, it's the '1t' format, otherwise '1u'.
 119Any text format data consists of a stream of events, such as URB submission,
 120URB callback, submission error. Every event is a text line, which consists
 121of whitespace separated words. The number or position of words may depend
 122on the event type, but there is a set of words, common for all types.
 124Here is the list of words, from left to right:
 126- URB Tag. This is used to identify URBs, and is normally an in-kernel address
 127  of the URB structure in hexadecimal, but can be a sequence number or any
 128  other unique string, within reason.
 130- Timestamp in microseconds, a decimal number. The timestamp's resolution
 131  depends on available clock, and so it can be much worse than a microsecond
 132  (if the implementation uses jiffies, for example).
 134- Event Type. This type refers to the format of the event, not URB type.
 135  Available types are: S - submission, C - callback, E - submission error.
 137- "Address" word (formerly a "pipe"). It consists of four fields, separated by
 138  colons: URB type and direction, Bus number, Device address, Endpoint number.
 139  Type and direction are encoded with two bytes in the following manner:
 141    == ==   =============================
 142    Ci Co   Control input and output
 143    Zi Zo   Isochronous input and output
 144    Ii Io   Interrupt input and output
 145    Bi Bo   Bulk input and output
 146    == ==   =============================
 148  Bus number, Device address, and Endpoint are decimal numbers, but they may
 149  have leading zeros, for the sake of human readers.
 151- URB Status word. This is either a letter, or several numbers separated
 152  by colons: URB status, interval, start frame, and error count. Unlike the
 153  "address" word, all fields save the status are optional. Interval is printed
 154  only for interrupt and isochronous URBs. Start frame is printed only for
 155  isochronous URBs. Error count is printed only for isochronous callback
 156  events.
 158  The status field is a decimal number, sometimes negative, which represents
 159  a "status" field of the URB. This field makes no sense for submissions, but
 160  is present anyway to help scripts with parsing. When an error occurs, the
 161  field contains the error code.
 163  In case of a submission of a Control packet, this field contains a Setup Tag
 164  instead of an group of numbers. It is easy to tell whether the Setup Tag is
 165  present because it is never a number. Thus if scripts find a set of numbers
 166  in this word, they proceed to read Data Length (except for isochronous URBs).
 167  If they find something else, like a letter, they read the setup packet before
 168  reading the Data Length or isochronous descriptors.
 170- Setup packet, if present, consists of 5 words: one of each for bmRequestType,
 171  bRequest, wValue, wIndex, wLength, as specified by the USB Specification 2.0.
 172  These words are safe to decode if Setup Tag was 's'. Otherwise, the setup
 173  packet was present, but not captured, and the fields contain filler.
 175- Number of isochronous frame descriptors and descriptors themselves.
 176  If an Isochronous transfer event has a set of descriptors, a total number
 177  of them in an URB is printed first, then a word per descriptor, up to a
 178  total of 5. The word consists of 3 colon-separated decimal numbers for
 179  status, offset, and length respectively. For submissions, initial length
 180  is reported. For callbacks, actual length is reported.
 182- Data Length. For submissions, this is the requested length. For callbacks,
 183  this is the actual length.
 185- Data tag. The usbmon may not always capture data, even if length is nonzero.
 186  The data words are present only if this tag is '='.
 188- Data words follow, in big endian hexadecimal format. Notice that they are
 189  not machine words, but really just a byte stream split into words to make
 190  it easier to read. Thus, the last word may contain from one to four bytes.
 191  The length of collected data is limited and can be less than the data length
 192  reported in the Data Length word. In the case of an Isochronous input (Zi)
 193  completion where the received data is sparse in the buffer, the length of
 194  the collected data can be greater than the Data Length value (because Data
 195  Length counts only the bytes that were received whereas the Data words
 196  contain the entire transfer buffer).
 200An input control transfer to get a port status::
 202  d5ea89a0 3575914555 S Ci:1:001:0 s a3 00 0000 0003 0004 4 <
 203  d5ea89a0 3575914560 C Ci:1:001:0 0 4 = 01050000
 205An output bulk transfer to send a SCSI command 0x28 (READ_10) in a 31-byte
 206Bulk wrapper to a storage device at address 5::
 208  dd65f0e8 4128379752 S Bo:1:005:2 -115 31 = 55534243 ad000000 00800000 80010a28 20000000 20000040 00000000 000000
 209  dd65f0e8 4128379808 C Bo:1:005:2 0 31 >
 211Raw binary format and API
 214The overall architecture of the API is about the same as the one above,
 215only the events are delivered in binary format. Each event is sent in
 216the following structure (its name is made up, so that we can refer to it)::
 218  struct usbmon_packet {
 219        u64 id;                 /*  0: URB ID - from submission to callback */
 220        unsigned char type;     /*  8: Same as text; extensible. */
 221        unsigned char xfer_type; /*    ISO (0), Intr, Control, Bulk (3) */
 222        unsigned char epnum;    /*     Endpoint number and transfer direction */
 223        unsigned char devnum;   /*     Device address */
 224        u16 busnum;             /* 12: Bus number */
 225        char flag_setup;        /* 14: Same as text */
 226        char flag_data;         /* 15: Same as text; Binary zero is OK. */
 227        s64 ts_sec;             /* 16: gettimeofday */
 228        s32 ts_usec;            /* 24: gettimeofday */
 229        int status;             /* 28: */
 230        unsigned int length;    /* 32: Length of data (submitted or actual) */
 231        unsigned int len_cap;   /* 36: Delivered length */
 232        union {                 /* 40: */
 233                unsigned char setup[SETUP_LEN]; /* Only for Control S-type */
 234                struct iso_rec {                /* Only for ISO */
 235                        int error_count;
 236                        int numdesc;
 237                } iso;
 238        } s;
 239        int interval;           /* 48: Only for Interrupt and ISO */
 240        int start_frame;        /* 52: For ISO */
 241        unsigned int xfer_flags; /* 56: copy of URB's transfer_flags */
 242        unsigned int ndesc;     /* 60: Actual number of ISO descriptors */
 243  };                            /* 64 total length */
 245These events can be received from a character device by reading with read(2),
 246with an ioctl(2), or by accessing the buffer with mmap. However, read(2)
 247only returns first 48 bytes for compatibility reasons.
 249The character device is usually called /dev/usbmonN, where N is the USB bus
 250number. Number zero (/dev/usbmon0) is special and means "all buses".
 251Note that specific naming policy is set by your Linux distribution.
 253If you create /dev/usbmon0 by hand, make sure that it is owned by root
 254and has mode 0600. Otherwise, unprivileged users will be able to snoop
 255keyboard traffic.
 257The following ioctl calls are available, with MON_IOC_MAGIC 0x92:
 259 MON_IOCQ_URB_LEN, defined as _IO(MON_IOC_MAGIC, 1)
 261This call returns the length of data in the next event. Note that majority of
 262events contain no data, so if this call returns zero, it does not mean that
 263no events are available.
 265 MON_IOCG_STATS, defined as _IOR(MON_IOC_MAGIC, 3, struct mon_bin_stats)
 267The argument is a pointer to the following structure::
 269  struct mon_bin_stats {
 270        u32 queued;
 271        u32 dropped;
 272  };
 274The member "queued" refers to the number of events currently queued in the
 275buffer (and not to the number of events processed since the last reset).
 277The member "dropped" is the number of events lost since the last call
 280 MON_IOCT_RING_SIZE, defined as _IO(MON_IOC_MAGIC, 4)
 282This call sets the buffer size. The argument is the size in bytes.
 283The size may be rounded down to the next chunk (or page). If the requested
 284size is out of [unspecified] bounds for this kernel, the call fails with
 287 MON_IOCQ_RING_SIZE, defined as _IO(MON_IOC_MAGIC, 5)
 289This call returns the current size of the buffer in bytes.
 291 MON_IOCX_GET, defined as _IOW(MON_IOC_MAGIC, 6, struct mon_get_arg)
 292 MON_IOCX_GETX, defined as _IOW(MON_IOC_MAGIC, 10, struct mon_get_arg)
 294These calls wait for events to arrive if none were in the kernel buffer,
 295then return the first event. The argument is a pointer to the following
 298  struct mon_get_arg {
 299        struct usbmon_packet *hdr;
 300        void *data;
 301        size_t alloc;           /* Length of data (can be zero) */
 302  };
 304Before the call, hdr, data, and alloc should be filled. Upon return, the area
 305pointed by hdr contains the next event structure, and the data buffer contains
 306the data, if any. The event is removed from the kernel buffer.
 308The MON_IOCX_GET copies 48 bytes to hdr area, MON_IOCX_GETX copies 64 bytes.
 310 MON_IOCX_MFETCH, defined as _IOWR(MON_IOC_MAGIC, 7, struct mon_mfetch_arg)
 312This ioctl is primarily used when the application accesses the buffer
 313with mmap(2). Its argument is a pointer to the following structure::
 315  struct mon_mfetch_arg {
 316        uint32_t *offvec;       /* Vector of events fetched */
 317        uint32_t nfetch;        /* Number of events to fetch (out: fetched) */
 318        uint32_t nflush;        /* Number of events to flush */
 319  };
 321The ioctl operates in 3 stages.
 323First, it removes and discards up to nflush events from the kernel buffer.
 324The actual number of events discarded is returned in nflush.
 326Second, it waits for an event to be present in the buffer, unless the pseudo-
 327device is open with O_NONBLOCK.
 329Third, it extracts up to nfetch offsets into the mmap buffer, and stores
 330them into the offvec. The actual number of event offsets is stored into
 331the nfetch.
 333 MON_IOCH_MFLUSH, defined as _IO(MON_IOC_MAGIC, 8)
 335This call removes a number of events from the kernel buffer. Its argument
 336is the number of events to remove. If the buffer contains fewer events
 337than requested, all events present are removed, and no error is reported.
 338This works when no events are available too.
 342The ioctl FIONBIO may be implemented in the future, if there's a need.
 344In addition to ioctl(2) and read(2), the special file of binary API can
 345be polled with select(2) and poll(2). But lseek(2) does not work.
 347* Memory-mapped access of the kernel buffer for the binary API
 349The basic idea is simple:
 351To prepare, map the buffer by getting the current size, then using mmap(2).
 352Then, execute a loop similar to the one written in pseudo-code below::
 354   struct mon_mfetch_arg fetch;
 355   struct usbmon_packet *hdr;
 356   int nflush = 0;
 357   for (;;) {
 358      fetch.offvec = vec; // Has N 32-bit words
 359      fetch.nfetch = N;   // Or less than N
 360      fetch.nflush = nflush;
 361      ioctl(fd, MON_IOCX_MFETCH, &fetch);   // Process errors, too
 362      nflush = fetch.nfetch;       // This many packets to flush when done
 363      for (i = 0; i < nflush; i++) {
 364         hdr = (struct ubsmon_packet *) &mmap_area[vec[i]];
 365         if (hdr->type == '@')     // Filler packet
 366            continue;
 367         caddr_t data = &mmap_area[vec[i]] + 64;
 368         process_packet(hdr, data);
 369      }
 370   }
 372Thus, the main idea is to execute only one ioctl per N events.
 374Although the buffer is circular, the returned headers and data do not cross
 375the end of the buffer, so the above pseudo-code does not need any gathering.