Documentation/usb/usbmon.txt

   1 * Introduction
   2
   3 The name "usbmon" in lowercase refers to a facility in kernel which is
   4 used to collect traces of I/O on the USB bus. This function is analogous
   5 to a packet socket used by network monitoring tools such as tcpdump(1)
   6 or Ethereal. Similarly, it is expected that a tool such as usbdump or
   7 USBMon (with uppercase letters) is used to examine raw traces produced
   8 by usbmon.
   9
  10 The usbmon reports requests made by peripheral-specific drivers to Host
  11 Controller Drivers (HCD). So, if HCD is buggy, the traces reported by
  12 usbmon may not correspond to bus transactions precisely. This is the same
  13 situation as with tcpdump.
  14
  15 * How to use usbmon to collect raw text traces
  16
  17 Unlike the packet socket, usbmon has an interface which provides traces
  18 in a text format. This is used for two purposes. First, it serves as a
  19 common trace exchange format for tools while most sophisticated formats
  20 are finalized. Second, humans can read it in case tools are not available.
  21
  22 To collect a raw text trace, execute following steps.
  23
  24 1. Prepare
  25
  26 Mount debugfs (it has to be enabled in your kernel configuration), and
  27 load the usbmon module (if built as module). The second step is skipped
  28 if usbmon is built into the kernel.
  29
  30 # mount -t debugfs none_debugs /sys/kernel/debug
  31 # modprobe usbmon
  32 #
  33
  34 Verify that bus sockets are present.
  35
  36 # ls /sys/kernel/debug/usbmon
  37 1s  1t  2s  2t  3s  3t  4s  4t
  38 #
  39
  40 2. Find which bus connects to the desired device
  41
  42 Run "cat /proc/bus/usb/devices", and find the T-line which corresponds to
  43 the device. Usually you do it by looking for the vendor string. If you have
  44 many similar devices, unplug one and compare two /proc/bus/usb/devices outputs.
  45 The T-line will have a bus number. Example:
  46
  47 T:  Bus=03 Lev=01 Prnt=01 Port=00 Cnt=01 Dev#=  2 Spd=12  MxCh= 0
  48 D:  Ver= 1.10 Cls=00(>ifc ) Sub=00 Prot=00 MxPS= 8 #Cfgs=  1
  49 P:  Vendor=0557 ProdID=2004 Rev= 1.00
  50 S:  Manufacturer=ATEN
  51 S:  Product=UC100KM V2.00
  52
  53 Bus=03 means it's bus 3.
  54
  55 3. Start 'cat'
  56
  57 # cat /sys/kernel/debug/usbmon/3t > /tmp/1.mon.out
  58
  59 This process will be reading until killed. Naturally, the output can be
  60 redirected to a desirable location. This is preferred, because it is going
  61 to be quite long.
  62
  63 4. Perform the desired operation on the USB bus
  64
  65 This is where you do something that creates the traffic: plug in a flash key,
  66 copy files, control a webcam, etc.
  67
  68 5. Kill cat
  69
  70 Usually it's done with a keyboard interrupt (Control-C).
  71
  72 At this point the output file (/tmp/1.mon.out in this example) can be saved,
  73 sent by e-mail, or inspected with a text editor. In the last case make sure
  74 that the file size is not excessive for your favourite editor.
  75
  76 * Raw text data format
  77
  78 The '1t' type data consists of a stream of events, such as URB submission,
  79 URB callback, submission error. Every event is a text line, which consists
  80 of whitespace separated words. The number or position of words may depend
  81 on the event type, but there is a set of words, common for all types.
  82
  83 Here is the list of words, from left to right:
  84 - URB Tag. This is used to identify URBs is normally a kernel mode address
  85  of the URB structure in hexadecimal.
  86 - Timestamp in microseconds, a decimal number. The timestamp's resolution
  87   depends on available clock, and so it can be much worse than a microsecond
  88   (if the implementation uses jiffies, for example).
  89 - Event Type. This type refers to the format of the event, not URB type.
  90   Available types are: S - submission, C - callback, E - submission error.
  91 - "Pipe". The pipe concept is deprecated. This is a composite word, used to
  92   be derived from information in pipes. It consists of three fields, separated
  93   by colons: URB type and direction, Device address, Endpoint number.
  94   Type and direction are encoded with two bytes in the following manner:
  95     Ci Co   Control input and output
  96     Zi Zo   Isochronous input and output
  97     Ii Io   Interrupt input and output
  98     Bi Bo   Bulk input and output
  99   Device address and Endpoint number are 3-digit and 2-digit (respectively)
 100   decimal numbers, with leading zeroes.
 101 - URB Status. In most cases, this field contains a number, sometimes negative,
 102   which represents a "status" field of the URB. This field makes no sense for
 103   submissions, but is present anyway to help scripts with parsing. When an
 104   error occurs, the field contains the error code. In case of a submission of
 105   a Control packet, this field contains a Setup Tag instead of an error code.
 106   It is easy to tell whether the Setup Tag is present because it is never a
 107   number. Thus if scripts find a number in this field, they proceed to read
 108   Data Length. If they find something else, like a letter, they read the setup
 109   packet before reading the Data Length.
 110 - Setup packet, if present, consists of 5 words: one of each for bmRequestType,
 111   bRequest, wValue, wIndex, wLength, as specified by the USB Specification 2.0.
 112   These words are safe to decode if Setup Tag was 's'. Otherwise, the setup
 113   packet was present, but not captured, and the fields contain filler.
 114 - Data Length. For submissions, this is the requested length. For callbacks,
 115   this is the actual length.
 116 - Data tag. The usbmon may not always capture data, even if length is nonzero.
 117   The data words are present only if this tag is '='.
 118 - Data words follow, in big endian hexadecimal format. Notice that they are
 119   not machine words, but really just a byte stream split into words to make
 120   it easier to read. Thus, the last word may contain from one to four bytes.
 121   The length of collected data is limited and can be less than the data length
 122   report in Data Length word.
 123
 124 Here is an example of code to read the data stream in a well known programming
 125 language:
 126
 127 class ParsedLine {
 128         int data_len;           /* Available length of data */
 129         byte data[];
 130
 131         void parseData(StringTokenizer st) {
 132                 int availwords = st.countTokens();
 133                 data = new byte[availwords * 4];
 134                 data_len = 0;
 135                 while (st.hasMoreTokens()) {
 136                         String data_str = st.nextToken();
 137                         int len = data_str.length() / 2;
 138                         int i;
 139                         int b;  // byte is signed, apparently?! XXX
 140                         for (i = 0; i < len; i++) {
 141                                 // data[data_len] = Byte.parseByte(
 142                                 //     data_str.substring(i*2, i*2 + 2),
 143                                 //     16);
 144                                 b = Integer.parseInt(
 145                                      data_str.substring(i*2, i*2 + 2),
 146                                      16);
 147                                 if (b >= 128)
 148                                         b *= -1;
 149                                 data[data_len] = (byte) b;
 150                                 data_len++;
 151                         }
 152                 }
 153         }
 154 }
 155
 156 This format may be changed in the future.
 157
 158 Examples:
 159
 160 An input control transfer to get a port status.
 161
 162 d5ea89a0 3575914555 S Ci:001:00 s a3 00 0000 0003 0004 4 <
 163 d5ea89a0 3575914560 C Ci:001:00 0 4 = 01050000
 164
 165 An output bulk transfer to send a SCSI command 0x5E in a 31-byte Bulk wrapper
 166 to a storage device at address 5:
 167
 168 dd65f0e8 4128379752 S Bo:005:02 -115 31 = 55534243 5e000000 00000000 00000600 00000000 00000000 00000000 000000
 169 dd65f0e8 4128379808 C Bo:005:02 0 31 >
 170
 171 * Raw binary format and API
 172
 173 The overall architecture of the API is about the same as the one above,
 174 only the events are delivered in binary format. Each event is sent in
 175 the following structure (its name is made up, so that we can refer to it):
 176
 177 struct usbmon_packet {
 178         u64 id;                 /*  0: URB ID - from submission to callback */
 179         unsigned char type;     /*  8: Same as text; extensible. */
 180         unsigned char xfer_type; /*    ISO (0), Intr, Control, Bulk (3) */
 181         unsigned char epnum;    /*     Endpoint number and transfer direction */
 182         unsigned char devnum;   /*     Device address */
 183         u16 busnum;             /* 12: Bus number */
 184         char flag_setup;        /* 14: Same as text */
 185         char flag_data;         /* 15: Same as text; Binary zero is OK. */
 186         s64 ts_sec;             /* 16: gettimeofday */
 187         s32 ts_usec;            /* 24: gettimeofday */
 188         int status;             /* 28: */
 189         unsigned int length;    /* 32: Length of data (submitted or actual) */
 190         unsigned int len_cap;   /* 36: Delivered length */
 191         unsigned char setup[8]; /* 40: Only for Control 'S' */
 192 };                              /* 48 bytes total */
 193
 194 These events can be received from a character device by reading with read(2),
 195 with an ioctl(2), or by accessing the buffer with mmap.
 196
 197 The character device is usually called /dev/usbmonN, where N is the USB bus
 198 number. Number zero (/dev/usbmon0) is special and means "all buses".
 199 However, this feature is not implemented yet. Note that specific naming
 200 policy is set by your Linux distribution.
 201
 202 If you create /dev/usbmon0 by hand, make sure that it is owned by root
 203 and has mode 0600. Otherwise, unpriviledged users will be able to snoop
 204 keyboard traffic.
 205
 206 The following ioctl calls are available, with MON_IOC_MAGIC 0x92:
 207
 208  MON_IOCQ_URB_LEN, defined as _IO(MON_IOC_MAGIC, 1)
 209
 210 This call returns the length of data in the next event. Note that majority of
 211 events contain no data, so if this call returns zero, it does not mean that
 212 no events are available.
 213
 214  MON_IOCG_STATS, defined as _IOR(MON_IOC_MAGIC, 3, struct mon_bin_stats)
 215
 216 The argument is a pointer to the following structure:
 217
 218 struct mon_bin_stats {
 219         u32 queued;
 220         u32 dropped;
 221 };
 222
 223 The member "queued" refers to the number of events currently queued in the
 224 buffer (and not to the number of events processed since the last reset).
 225
 226 The member "dropped" is the number of events lost since the last call
 227 to MON_IOCG_STATS.
 228
 229  MON_IOCT_RING_SIZE, defined as _IO(MON_IOC_MAGIC, 4)
 230
 231 This call sets the buffer size. The argument is the size in bytes.
 232 The size may be rounded down to the next chunk (or page). If the requested
 233 size is out of [unspecified] bounds for this kernel, the call fails with
 234 -EINVAL.
 235
 236  MON_IOCQ_RING_SIZE, defined as _IO(MON_IOC_MAGIC, 5)
 237
 238 This call returns the current size of the buffer in bytes.
 239
 240  MON_IOCX_GET, defined as _IOW(MON_IOC_MAGIC, 6, struct mon_get_arg)
 241
 242 This call waits for events to arrive if none were in the kernel buffer,
 243 then returns the first event. Its argument is a pointer to the following
 244 structure:
 245
 246 struct mon_get_arg {
 247         struct usbmon_packet *hdr;
 248         void *data;
 249         size_t alloc;           /* Length of data (can be zero) */
 250 };
 251
 252 Before the call, hdr, data, and alloc should be filled. Upon return, the area
 253 pointed by hdr contains the next event structure, and the data buffer contains
 254 the data, if any. The event is removed from the kernel buffer.
 255
 256  MON_IOCX_MFETCH, defined as _IOWR(MON_IOC_MAGIC, 7, struct mon_mfetch_arg)
 257
 258 This ioctl is primarily used when the application accesses the buffer
 259 with mmap(2). Its argument is a pointer to the following structure:
 260
 261 struct mon_mfetch_arg {
 262         uint32_t *offvec;       /* Vector of events fetched */
 263         uint32_t nfetch;        /* Number of events to fetch (out: fetched) */
 264         uint32_t nflush;        /* Number of events to flush */
 265 };
 266
 267 The ioctl operates in 3 stages.
 268
 269 First, it removes and discards up to nflush events from the kernel buffer.
 270 The actual number of events discarded is returned in nflush.
 271
 272 Second, it waits for an event to be present in the buffer, unless the pseudo-
 273 device is open with O_NONBLOCK.
 274
 275 Third, it extracts up to nfetch offsets into the mmap buffer, and stores
 276 them into the offvec. The actual number of event offsets is stored into
 277 the nfetch.
 278
 279  MON_IOCH_MFLUSH, defined as _IO(MON_IOC_MAGIC, 8)
 280
 281 This call removes a number of events from the kernel buffer. Its argument
 282 is the number of events to remove. If the buffer contains fewer events
 283 than requested, all events present are removed, and no error is reported.
 284 This works when no events are available too.
 285
 286  FIONBIO
 287
 288 The ioctl FIONBIO may be implemented in the future, if there's a need.
 289
 290 In addition to ioctl(2) and read(2), the special file of binary API can
 291 be polled with select(2) and poll(2). But lseek(2) does not work.
 292
 293 * Memory-mapped access of the kernel buffer for the binary API
 294
 295 The basic idea is simple:
 296
 297 To prepare, map the buffer by getting the current size, then using mmap(2).
 298 Then, execute a loop similar to the one written in pseudo-code below:
 299
 300    struct mon_mfetch_arg fetch;
 301    struct usbmon_packet *hdr;
 302    int nflush = 0;
 303    for (;;) {
 304       fetch.offvec = vec; // Has N 32-bit words
 305       fetch.nfetch = N;   // Or less than N
 306       fetch.nflush = nflush;
 307       ioctl(fd, MON_IOCX_MFETCH, &fetch);   // Process errors, too
 308       nflush = fetch.nfetch;       // This many packets to flush when done
 309       for (i = 0; i < nflush; i++) {
 310          hdr = (struct ubsmon_packet *) &mmap_area[vec[i]];
 311          if (hdr->type == '@')     // Filler packet
 312             continue;
 313          caddr_t data = &mmap_area[vec[i]] + 64;
 314          process_packet(hdr, data);
 315       }
 316    }
 317
 318 Thus, the main idea is to execute only one ioctl per N events.
 319
 320 Although the buffer is circular, the returned headers and data do not cross
 321 the end of the buffer, so the above pseudo-code does not need any gathering.