fix: configuration descriptor has incorrectly sizes

update some descriptors

examples/device/video_capture/src/usb_descriptors.c

@@ -102,7 +102,7 @@ uint8_t const * tud_descriptor_device_cb(void)
 uint8_t const desc_fs_configuration[] =
 {
   // Config number, interface count, string index, total length, attribute, power in mA
-  TUD_CONFIG_DESCRIPTOR(1, ITF_NUM_TOTAL, 0, CONFIG_TOTAL_LEN, TUSB_DESC_CONFIG_ATT_REMOTE_WAKEUP, 100),
+  TUD_CONFIG_DESCRIPTOR(1, ITF_NUM_TOTAL, 0, CONFIG_TOTAL_LEN, 0, 500),
   // IAD for Video Control
   TUD_VIDEO_CAPTURE_DESCRIPTOR(4, EPNUM_VIDEO_IN, 128, 96, 15)
 };

examples/device/video_capture/src/usb_descriptors.h

@@ -39,18 +39,27 @@ enum {
   ITF_NUM_TOTAL
 };
 
-#define TUD_VIDEO_CAPTURE_DESC_LEN (TUD_VIDEO_DESC_IAD_LEN\
+#define TUD_VIDEO_CAPTURE_DESC_LEN (\
+    TUD_VIDEO_DESC_IAD_LEN\
     /* control */\
     + TUD_VIDEO_DESC_STD_VC_LEN\
-    + TUD_VIDEO_DESC_CS_VC_LEN\
+    + (TUD_VIDEO_DESC_CS_VC_LEN + 1/*bInCollection*/)\
     + TUD_VIDEO_DESC_INPUT_TERM_LEN\
     + TUD_VIDEO_DESC_OUTPUT_TERM_LEN\
     /* Interface 1, Alternate 0 */\
     + TUD_VIDEO_DESC_STD_VS_LEN\
-    + TUD_VIDEO_DESC_CS_VS_IN_LEN\
+    + (TUD_VIDEO_DESC_CS_VS_IN_LEN + 1/*bNumFormats x bControlSize*/)\
     + TUD_VIDEO_DESC_CS_VS_FMT_UNCOMPR_LEN\
-    + TUD_VIDEO_DESC_CS_VS_FRM_UNCOMPR_CONT_LEN)
+    + TUD_VIDEO_DESC_CS_VS_FRM_UNCOMPR_CONT_LEN\
+    + TUD_VIDEO_DESC_CS_VS_COLOR_MATCHING_LEN\
+  )
 
+#if 0
+    /* Interface 1, Alternate 1 */\
+    + TUD_VIDEO_DESC_STD_VS_LEN\
+    + 7/* Endpoint */\
+
+#endif
 
 #define TUD_VIDEO_DESC_CS_VS_FMT_YUY2(_fmtidx, _numfmtdesc, _frmidx, _asrx, _asry, _interlace, _cp) \
   TUD_VIDEO_DESC_CS_VS_FMT_UNCOMPR(_fmtidx, _numfmtdesc, TUD_VIDEO_GUID_YUY2, 16, _frmidx, _asrx, _asry, _interlace, _cp)
@@ -64,34 +73,40 @@ enum {
 #define TUD_VIDEO_CAPTURE_DESCRIPTOR(_stridx, _epin, _width, _height, _fps) \
   TUD_VIDEO_DESC_IAD(ITF_NUM_VIDEO_CONTROL, ITF_NUM_TOTAL, _stridx), \
     /* Video control 0 */ \
-    TUD_VIDEO_DESC_STD_VC(ITF_NUM_VIDEO_CONTROL, 0, 0), \
+    TUD_VIDEO_DESC_STD_VC(ITF_NUM_VIDEO_CONTROL, 0, _stridx), \
       TUD_VIDEO_DESC_CS_VC( /* UVC 1.5*/ 0x0150, \
-           TUD_VIDEO_DESC_INPUT_TERM_LEN + \
+           /* wTotalLength - bLength */ \
-           TUD_VIDEO_DESC_OUTPUT_TERM_LEN, \
+           TUD_VIDEO_DESC_INPUT_TERM_LEN + TUD_VIDEO_DESC_OUTPUT_TERM_LEN, \
            UVC_CLOCK_FREQUENCY, 1, 1), \
         TUD_VIDEO_DESC_INPUT_TERM(UVC_ENTITY_CAP_INPUT_TERMINAL,\
                                    VIDEO_TT_COMPOSITE_CONNECTOR, 0, 0), \
         TUD_VIDEO_DESC_OUTPUT_TERM(UVC_ENTITY_CAP_OUTPUT_TERMINAL,\
-                                   VIDEO_TT_STREAMING, 0, 1, 0),   \
+                                   VIDEO_TT_STREAMING, 0, 1, 0), \
     /* Video stream alt. 0 */ \
     TUD_VIDEO_DESC_STD_VS( 1, 0, 0, 0), \
       /* Video stream header for without still image capture */ \
-      TUD_VIDEO_DESC_CS_VS_INPUT(1, \
+      TUD_VIDEO_DESC_CS_VS_INPUT( /*bNumFormats*/1, \
-          TUD_VIDEO_DESC_CS_VS_FMT_UNCOMPR_LEN + \
+          /*wTotalLength - bLength */\
-          TUD_VIDEO_DESC_CS_VS_FRM_UNCOMPR_CONT_LEN,\
+          TUD_VIDEO_DESC_CS_VS_FMT_UNCOMPR_LEN\
-          _epin, /*bmInfo*/ 0, UVC_ENTITY_CAP_OUTPUT_TERMINAL, \
+          + TUD_VIDEO_DESC_CS_VS_FRM_UNCOMPR_CONT_LEN\
-          0, 2, 0, 0, 0, 1, 0), \
+          + TUD_VIDEO_DESC_CS_VS_COLOR_MATCHING_LEN,\
+          _epin, /*bmInfo*/0, /*bTerminalLink*/UVC_ENTITY_CAP_OUTPUT_TERMINAL, \
+          /*bStillCaptureMethod*/0, /*bTriggerSupport*/0, /*bTriggerUsage*/0, \
+          /*bControlSize*/1, /*bmaControls(1)*/0), \
         /* Video stream format */ \
-        TUD_VIDEO_DESC_CS_VS_FMT_I420(0, 1, 0, 1, 1, 0, 0), \
+        TUD_VIDEO_DESC_CS_VS_FMT_I420(/*bFormatIndex*/1, /*bNumFrameDescriptors*/1,\
+          /*bDefaultFrameIndex*/1, 1, 1, 0, /*bCopyProtect*/0), \
           /* Video stream frame format */ \
-          TUD_VIDEO_DESC_CS_VS_FRM_UNCOMPR_CONT( 1, 0, _width, _height, \
+          TUD_VIDEO_DESC_CS_VS_FRM_UNCOMPR_CONT(/*bFrameIndex */1, 0, _width, _height, \
-                                                 _width * _height * 12, _width * _height * 12 * _fps, \
+              _width * _height * 12, _width * _height * 12 * _fps, \
-                                                 _width * _height * 12, \
+              _width * _height * 12, \
-                                                 (10000000/_fps), (10000000/1), (10000000/_fps), 166666), \
+              (10000000/_fps), (10000000/1), (10000000/_fps), 166666), \
-    /* VS alt 1 */ \
+        TUD_VIDEO_DESC_CS_VS_COLOR_MATCHING(VIDEO_COLOR_PRIMARIES_BT709,\
-    TUD_VIDEO_DESC_STD_VS( 1, 1, 1, 0), \
+           VIDEO_COLOR_XFER_CH_BT709, VIDEO_COLOR_COEF_SMPTE170M), \
-    /* EP */ \
+    /* VS alt 1 */\
-    TUD_VIDEO_DESC_EP_ISO(_epin, (_width * _height * 12 / 8) * _fps, 1)
+    TUD_VIDEO_DESC_STD_VS(1, 1, 1, 0), \
+      /* EP */ \
+      TUD_VIDEO_DESC_EP_ISO(_epin, (_width * _height * 12 / 8) * _fps, 1)
 
 
 #endif

src/class/video/video.h

@@ -29,6 +29,35 @@
 
 #include "common/tusb_common.h"
 
+typedef enum {
+  VIDEO_COLOR_PRIMARIES_UNDEFINED = 0x00,
+  VIDEO_COLOR_PRIMARIES_BT709,
+  VIDEO_COLOR_PRIMARIES_BT470_2M,
+  VIDEO_COLOR_PRIMARIES_BT470_2BG,
+  VIDEO_COLOR_PRIMARIES_SMPTE170M,
+  VIDEO_COLOR_PRIMARIES_SMPTE240M,
+} video_color_primaries_t;
+
+typedef enum {
+  VIDEO_COLOR_XFER_CH_UNDEFINED = 0x00,
+  VIDEO_COLOR_XFER_CH_BT709,
+  VIDEO_COLOR_XFER_CH_BT470_2M,
+  VIDEO_COLOR_XFER_CH_BT470_2BG,
+  VIDEO_COLOR_XFER_CH_SMPTE170M,
+  VIDEO_COLOR_XFER_CH_SMPTE240M,
+  VIDEO_COLOR_XFER_CH_LINEAR,
+  VIDEO_COLOR_XFER_CH_SRGB,
+} video_color_transfer_characteristics_t;
+
+typedef enum {
+  VIDEO_COLOR_COEF_UNDEFINED = 0x00,
+  VIDEO_COLOR_COEF_BT709,
+  VIDEO_COLOR_COEF_FCC,
+  VIDEO_COLOR_COEF_BT470_2BG,
+  VIDEO_COLOR_COEF_SMPTE170M,
+  VIDEO_COLOR_COEF_SMPTE240M,
+} video_color_matrix_coefficients_t;
+
 /* 4.2.1.2 */
 typedef enum {
   VIDEO_NO_ERROR = 0, /* The request succeeded. */
@@ -69,7 +98,7 @@ typedef enum
   VIDEO_CS_VC_INTERFACE_PROCESSING_UNIT,
   VIDEO_CS_VC_INTERFACE_EXTENSION_UNIT,
   VIDEO_CS_VC_INTERFACE_ENCODING_UNIT,
-} vide_cs_vc_interface_subtype_t;
+} video_cs_vc_interface_subtype_t;
 
 /* A.6 */
 typedef enum
@@ -278,6 +307,7 @@ typedef struct TU_ATTR_PACKED {
 #define TUD_VIDEO_DESC_CS_VS_FMT_UNCOMPR_LEN      27
 #define TUD_VIDEO_DESC_CS_VS_FRM_UNCOMPR_CONT_LEN 38
 #define TUD_VIDEO_DESC_CS_VS_FRM_UNCOMPR_DISC_LEN 26
+#define TUD_VIDEO_DESC_CS_VS_COLOR_MATCHING_LEN   6
 
 /* 2.2 compression formats */
 #define TUD_VIDEO_GUID_YUY2   0x59,0x55,0x59,0x32,0x00,0x00,0x10,0x00,0x00,0x80,0x71,0x9B,0x38,0x00,0xAA,0x00
@@ -296,8 +326,8 @@ typedef struct TU_ATTR_PACKED {
 
 /* 3.7.2 */
 #define TUD_VIDEO_DESC_CS_VC(_bcdUVC, _totallen, _clkfreq, _coll, ...)	\
-  TUD_VIDEO_DESC_CS_VC_LEN + _coll, TUSB_DESC_CS_INTERFACE, VIDEO_CS_VC_INTERFACE_HEADER, \
+  TUD_VIDEO_DESC_CS_VC_LEN + (_coll), TUSB_DESC_CS_INTERFACE, VIDEO_CS_VC_INTERFACE_HEADER, \
-  U16_TO_U8S_LE(_bcdUVC), U16_TO_U8S_LE(_totallen + TUD_VIDEO_DESC_CS_VC_LEN), \
+  U16_TO_U8S_LE(_bcdUVC), U16_TO_U8S_LE((_totallen) + TUD_VIDEO_DESC_CS_VC_LEN + (_coll)), \
   U32_TO_U8S_LE(_clkfreq), _coll, __VA_ARGS__
 
 /* 3.7.2.1 */
@@ -307,30 +337,31 @@ typedef struct TU_ATTR_PACKED {
 
 /* 3.7.2.2 */
 #define TUD_VIDEO_DESC_OUTPUT_TERM(_tid, _tt, _at, _srcid, _stridx) \
-  TUD_VIDEO_DESC_INPUT_TERM_LEN, TUSB_DESC_CS_INTERFACE, VIDEO_CS_VC_INTERFACE_OUTPUT_TERMINAL, \
+  TUD_VIDEO_DESC_OUTPUT_TERM_LEN, TUSB_DESC_CS_INTERFACE, VIDEO_CS_VC_INTERFACE_OUTPUT_TERMINAL, \
-    _tid, U16_TO_U8S_LE(_tt), _at, _stridx
+    _tid, U16_TO_U8S_LE(_tt), _at, _srcid, _stridx
 
 /* 3.9.1 */
-#define TUD_VIDEO_DESC_STD_VS(_itfnum, _alt, _epn, _stridx)   \
+#define TUD_VIDEO_DESC_STD_VS(_itfnum, _alt, _epn, _stridx) \
-  TUD_VIDEO_DESC_STD_VC_LEN, TUSB_DESC_INTERFACE, _itfnum, _alt, \
+  TUD_VIDEO_DESC_STD_VS_LEN, TUSB_DESC_INTERFACE, _itfnum, _alt, \
   _epn, TUSB_CLASS_VIDEO, VIDEO_SUBCLASS_STREAMING, VIDEO_INT_PROTOCOL_CODE_15, _stridx
 
 /* 3.9.2.1 */
-#define TUD_VIDEO_DESC_CS_VS_INPUT(_numfmt, _totlen, _epn, _inf, _termlnk, _sticaptmeth, _trgspt, _trgusg, _ctlsz, ...) \
+#define TUD_VIDEO_DESC_CS_VS_INPUT(_numfmt, _totallen, _ep, _inf, _termlnk, _sticaptmeth, _trgspt, _trgusg, _ctlsz, ...) \
   TUD_VIDEO_DESC_CS_VS_IN_LEN + (_numfmt) * (_ctlsz), TUSB_DESC_CS_INTERFACE, \
   VIDEO_CS_VS_INTERFACE_INPUT_HEADER, _numfmt, \
-  U16_TO_U8S_LE(_totlen + TUD_VIDEO_DESC_CS_VS_IN_LEN + (_numfmt) * (_ctlsz)), \
+  U16_TO_U8S_LE((_totallen) + TUD_VIDEO_DESC_CS_VS_IN_LEN + (_numfmt) * (_ctlsz)), \
-  _epn, _inf, _termlnk, _sticaptmeth, _trgspt, _trgusg, _ctlsz, __VA_ARGS__
+  _ep, _inf, _termlnk, _sticaptmeth, _trgspt, _trgusg, _ctlsz, __VA_ARGS__
 
 /* 3.9.2.2 */
-#define TUD_VIDEO_DESC_CS_VS_OUTPUT(_numfmt, _totlen, _epn, _inf, _termlnk, _ctlsz, ...) \
+#define TUD_VIDEO_DESC_CS_VS_OUTPUT(_numfmt, _totallen, _ep, _inf, _termlnk, _ctlsz, ...) \
   TUD_VIDEO_DESC_CS_VS_OUT_LEN + (_numfmt) * (_ctlsz), TUSB_DESC_CS_INTERFACE, \
   VIDEO_CS_VS_INTERFACE_OUTPUT_HEADER, _numfmt, \
-  U16_TO_U8S_LE(_totlen + TUD_VIDEO_DESC_CS_VS_OUT_LEN + (_numfmt) * (_ctlsz)), \
+  U16_TO_U8S_LE((_totallen) + TUD_VIDEO_DESC_CS_VS_OUT_LEN + (_numfmt) * (_ctlsz)), \
-  _epn, _inf, _termlnk,  _trgusg, _ctlsz, __VA_ARGS__
+  _ep, _inf, _termlnk, _ctlsz, __VA_ARGS__
 
 /* Uncompressed 3.1.1 */
 #define TUD_VIDEO_GUID(_g0,_g1,_g2,_g3,_g4,_g5,_g6,_g7,_g8,_g9,_g10,_g11,_g12,_g13,_g14,_g15) _g0,_g1,_g2,_g3,_g4,_g5,_g6,_g7,_g8,_g9,_g10,_g11,_g12,_g13,_g14,_g15
+
 #define TUD_VIDEO_DESC_CS_VS_FMT_UNCOMPR(_fmtidx, _numfrmdesc, \
   _guid, _bitsperpix, _frmidx, _asrx, _asry, _interlace, _cp) \
   TUD_VIDEO_DESC_CS_VS_FMT_UNCOMPR_LEN, TUSB_DESC_CS_INTERFACE, VIDEO_CS_VS_INTERFACE_FORMAT_UNCOMPRESSED, \
@@ -353,6 +384,12 @@ typedef struct TU_ATTR_PACKED {
   _frmidx, _cap, U16_TO_U8S_LE(_width), U16_TO_U8S_LE(_height), U32_TO_U8S_LE(_minbr), U32_TO_U8S_LE(_maxbr), \
   U32_TO_U8S_LE(_maxfrmbufsz), U32_TO_U8S_LE(_frminterval), _numfrminterval, __VA_ARGS__
 
+/* 3.9.2.6 */
+#define TUD_VIDEO_DESC_CS_VS_COLOR_MATCHING(_color, _trns, _mat) \
+  TUD_VIDEO_DESC_CS_VS_COLOR_MATCHING_LEN, \
+  TUSB_DESC_CS_INTERFACE, VIDEO_CS_VS_INTERFACE_COLORFORMAT, \
+  _color, _trns, _mat
+
 /* 3.10.1.1 */
 #define TUD_VIDEO_DESC_EP_ISO(_ep, _epsize, _ep_interval) \
   7, TUSB_DESC_ENDPOINT, _ep, TUSB_XFER_ISOCHRONOUS | TUSB_ISO_EP_ATT_ASYNCHRONOUS,\

src/class/video/video_device.c

@@ -183,7 +183,7 @@ static void const* _find_desc_itf(void const *beg, void const *end, unsigned itf
 static void const* _find_desc_entity(tusb_desc_vc_itf_t const *vc, unsigned entityid)
 {
   void const *beg = (void const*)vc;
-  void const *end = beg + vc->std.bLength + vc->ctl.bLength + vc->ctl.wTotalLength;
+  void const *end = beg + vc->std.bLength + vc->ctl.wTotalLength;
   for (void const *cur = beg; cur < end; cur = _find_desc(cur, end, TUSB_DESC_CS_INTERFACE)) {
     tusb_desc_cs_video_entity_itf_t const *itf = (tusb_desc_cs_video_entity_itf_t const *)cur;
     if ((VIDEO_CS_VC_INTERFACE_INPUT_TERMINAL  == itf->bDescriptorSubtype ||
@@ -206,7 +206,7 @@ static bool _close_vc_itf(uint8_t rhport, videod_interface_t *self)
   /* The next descriptor after the class-specific VC interface header descriptor. */
   void const *cur = (void const*)vc + vc->std.bLength + vc->ctl.bLength;
   /* The end of the video control interface descriptor. */
-  void const *end = cur + vc->ctl.wTotalLength;
+  void const *end = (void const*)vc + vc->std.bLength + vc->ctl.wTotalLength;
   if (vc->std.bNumEndpoints) {
     /* Find the notification endpoint descriptor. */
     cur = _find_desc(cur, end, TUSB_DESC_ENDPOINT);
@@ -224,21 +224,25 @@ static bool _close_vc_itf(uint8_t rhport, videod_interface_t *self)
  * @param[in]     altnum   The target alternate setting number. */
 static bool _open_vc_itf(uint8_t rhport, videod_interface_t *self, unsigned altnum)
 {
+  TU_LOG2("    open VC %d\r\n", altnum);
   void const *beg = self->beg;
   void const *end = beg + self->len;
   /* The first descriptor is a video control interface descriptor. */
   unsigned itfnum = ((tusb_desc_interface_t const *)beg)->bInterfaceNumber;
   void const *cur = _find_desc_itf(beg, end, itfnum, altnum);
+  TU_LOG2("    cur %ld\r\n", cur - beg);
   TU_VERIFY(cur < end);
 
   tusb_desc_vc_itf_t const *vc = (tusb_desc_vc_itf_t const *)cur;
+  TU_LOG2("    bInCollection %d\r\n", vc->ctl.bInCollection);
   /* Support for up to 2 streaming interfaces only. */
   TU_ASSERT(vc->ctl.bInCollection < 3);
 
+  /* Update to point the end of the video control interface descriptor. */
+  end  = cur + vc->std.bLength + vc->ctl.wTotalLength;
   /* Advance to the next descriptor after the class-specific VC interface header descriptor. */
   cur += vc->std.bLength + vc->ctl.bLength;
-  /* Update to point the end of the video control interface descriptor. */
+  TU_LOG2("    bNumEndpoints %d\r\n", vc->std.bNumEndpoints);
-  end  = cur + vc->ctl.wTotalLength;
   /* Open the notification endpoint if it exist. */
   if (vc->std.bNumEndpoints) {
     /* Support for 1 endpoint only. */
@@ -266,7 +270,7 @@ static bool _close_vs_itf(uint8_t rhport, videod_interface_t *self, unsigned itf
   /* The next of the video streaming interface header descriptor. */
   void const *cur = (void const*)vs + vs->std.bLength + vs->stm.bLength;
   /* The end of the video streaming interface descriptor. */
-  void const *end = cur + vs->stm.wTotalLength;
+  void const *end = (void const*)vs + vs->std.bLength + vs->stm.wTotalLength;
   for (unsigned i = 0; i < vs->std.bNumEndpoints; ++i) {
     cur = _find_desc(cur, end, TUSB_DESC_ENDPOINT);
     TU_ASSERT(cur < end);
@@ -285,6 +289,7 @@ static bool _close_vs_itf(uint8_t rhport, videod_interface_t *self, unsigned itf
  * @param[in]     altnum   The target alternate setting number. */
 static bool _open_vs_itf(uint8_t rhport, videod_interface_t *self, unsigned itfnum, unsigned altnum)
 {
+  TU_LOG2("    open VS %d,%d\r\n", itfnum, altnum);
   uint16_t   *ofs = NULL;
   for (unsigned i = 1; i < sizeof(self->ofs)/sizeof(self->ofs[0]); ++i) {
     if (!self->ofs[i]) {
@@ -297,17 +302,17 @@ static bool _open_vs_itf(uint8_t rhport, videod_interface_t *self, unsigned itfn
   tusb_desc_vc_itf_t const *vc = _get_desc_vc(self);
   void const *end = self->beg + self->len;
   /* Set the end of the video control interface descriptor. */
-  void const *cur = (void const*)vc + vc->std.bLength + vc->ctl.bLength + vc->ctl.wTotalLength;
+  void const *cur = (void const*)vc + vc->std.bLength + vc->ctl.wTotalLength;
 
   cur = _find_desc_itf(cur, end, itfnum, altnum);
   TU_VERIFY(cur < end);
   tusb_desc_vs_itf_t const *vs = (tusb_desc_vs_itf_t const*)cur;
   /* Support for up to 2 endpoint only. */
   TU_ASSERT(vs->std.bNumEndpoints < 3);
+  /* Update to point the end of the video control interface descriptor. */
+  end  = cur + vs->std.bLength + vs->stm.wTotalLength;
   /* Advance to the next descriptor after the class-specific VS interface header descriptor. */
   cur += vs->std.bLength + vs->stm.bLength;
-  /* Update to point the end of the video control interface descriptor. */
-  end  = cur + vs->stm.wTotalLength;
   for (unsigned i = 0; i < vs->std.bNumEndpoints; ++i) {
     cur = _find_desc(cur, end, TUSB_DESC_ENDPOINT);
     TU_VERIFY(cur < end);
@@ -612,7 +617,7 @@ uint16_t videod_open(uint8_t rhport, tusb_desc_interface_t const * itf_desc, uin
     itfnum = vc->ctl.baInterfaceNr[i];
     if (!_open_vs_itf(rhport, self, itfnum, 0)) return 0;
   }
-  return end - cur;
+  return (uintptr_t)cur - (uintptr_t)itf_desc;
 }
 
 // Invoked when a control transfer occurred on an interface of this class