update novnc commit

.gitmodules

@@ -1,4 +1,4 @@
 [submodule "kasmweb"]
 	path = kasmweb
 	url = https://github.com/kasmtech/noVNC.git
-	branch = master
+	branch = video

builder/dockerfile.debian_bullseye.build

@@ -13,7 +13,7 @@ RUN apt-get update && \
 RUN DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends tzdata
 RUN apt-get update && apt-get -y build-dep xorg-server libxfont-dev
 RUN apt-get update && apt-get -y install cmake git libjpeg-dev libgnutls28-dev vim wget tightvncserver
-RUN apt-get update && apt-get -y install libjpeg-dev libpng-dev libtiff-dev libgif-dev libavcodec-dev libssl-dev
+RUN apt-get update && apt-get -y install libjpeg-dev libpng-dev libtiff-dev libgif-dev libavcodec-dev libssl-dev libx264-dev
 
 # Additions for webp
 RUN cd /tmp && wget https://storage.googleapis.com/downloads.webmproject.org/releases/webp/libwebp-1.0.2.tar.gz

builder/dockerfile.debian_buster.build

@@ -13,7 +13,7 @@ RUN apt-get update && \
 RUN DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends tzdata
 RUN apt-get update && apt-get -y build-dep xorg-server libxfont-dev
 RUN apt-get update && apt-get -y install cmake git libjpeg-dev libgnutls28-dev vim wget tightvncserver
-RUN apt-get update && apt-get -y install libjpeg-dev libpng-dev libtiff-dev libgif-dev libavcodec-dev libssl-dev
+RUN apt-get update && apt-get -y install libjpeg-dev libpng-dev libtiff-dev libgif-dev libavcodec-dev libssl-dev libx264-dev
 
 # Additions for webp
 RUN cd /tmp && wget https://storage.googleapis.com/downloads.webmproject.org/releases/webp/libwebp-1.0.2.tar.gz

builder/dockerfile.kali_kali-rolling.build

@@ -13,7 +13,7 @@ RUN apt-get update && \
 RUN DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends tzdata
 RUN apt-get update && apt-get -y build-dep xorg-server libxfont-dev
 RUN apt-get update && apt-get -y install cmake git libjpeg-dev libgnutls28-dev vim wget tightvncserver
-RUN apt-get update && apt-get -y install libjpeg-dev libpng-dev libtiff-dev libgif-dev libavcodec-dev libssl-dev
+RUN apt-get update && apt-get -y install libjpeg-dev libpng-dev libtiff-dev libgif-dev libavcodec-dev libssl-dev libx264-dev
 
 # Additions for webp
 RUN cd /tmp && wget https://storage.googleapis.com/downloads.webmproject.org/releases/webp/libwebp-1.0.2.tar.gz

builder/dockerfile.ubuntu_bionic.build

@@ -11,7 +11,7 @@ RUN apt-get update && \
 
 RUN apt-get update && apt-get -y build-dep xorg-server libxfont-dev
 RUN apt-get update && apt-get -y install cmake git libjpeg-dev libgnutls28-dev vim wget tightvncserver
-RUN apt-get update && apt-get -y install libjpeg-dev libpng-dev libtiff-dev libgif-dev libavcodec-dev libssl-dev
+RUN apt-get update && apt-get -y install libjpeg-dev libpng-dev libtiff-dev libgif-dev libavcodec-dev libssl-dev libx264-dev
 
 # Additions for webp
 RUN cd /tmp && wget https://storage.googleapis.com/downloads.webmproject.org/releases/webp/libwebp-1.0.2.tar.gz

builder/dockerfile.ubuntu_focal.build

@@ -13,7 +13,7 @@ RUN apt-get update && \
 RUN apt-get update && apt-get install -y --no-install-recommends tzdata
 RUN apt-get update && apt-get -y build-dep xorg-server libxfont-dev
 RUN apt-get update && apt-get -y install cmake git libjpeg-dev libgnutls28-dev vim wget tightvncserver
-RUN apt-get update && apt-get -y install libjpeg-dev libpng-dev libtiff-dev libgif-dev libavcodec-dev libssl-dev
+RUN apt-get update && apt-get -y install libjpeg-dev libpng-dev libtiff-dev libgif-dev libavcodec-dev libssl-dev libx264-dev
 
 # Additions for webp
 RUN cd /tmp && wget https://storage.googleapis.com/downloads.webmproject.org/releases/webp/libwebp-1.0.2.tar.gz

common/network/websocket.c

@@ -566,7 +566,7 @@ int parse_handshake(ws_ctx_t *ws_ctx, char *handshake) {
     headers->key3[0] = '\0';
 
     if ((strlen(handshake) < 92) || (bcmp(handshake, "GET ", 4) != 0) ||
-        (!strstr(handshake, "Upgrade: websocket"))) {
+        (!strcasestr(handshake, "Upgrade: websocket"))) {
         return 0;
     }
     start = handshake+4;
@@ -587,7 +587,7 @@ int parse_handshake(ws_ctx_t *ws_ctx, char *handshake) {
     if (start) {
         start += 10;
     } else {
-        start = strstr(handshake, "\r\nSec-WebSocket-Origin: ");
+        start = strcasestr(handshake, "\r\nSec-WebSocket-Origin: ");
         if (!start) { return 0; }
         start += 24;
     }
@@ -595,7 +595,7 @@ int parse_handshake(ws_ctx_t *ws_ctx, char *handshake) {
     strncpy(headers->origin, start, end-start);
     headers->origin[end-start] = '\0';
 
-    start = strstr(handshake, "\r\nSec-WebSocket-Version: ");
+    start = strcasestr(handshake, "\r\nSec-WebSocket-Version: ");
     if (start) {
         // HyBi/RFC 6455
         start += 25;
@@ -605,7 +605,7 @@ int parse_handshake(ws_ctx_t *ws_ctx, char *handshake) {
         ws_ctx->hixie = 0;
         ws_ctx->hybi = strtol(headers->version, NULL, 10);
 
-        start = strstr(handshake, "\r\nSec-WebSocket-Key: ");
+        start = strcasestr(handshake, "\r\nSec-WebSocket-Key: ");
         if (!start) { return 0; }
         start += 21;
         end = strstr(start, "\r\n");
@@ -619,7 +619,7 @@ int parse_handshake(ws_ctx_t *ws_ctx, char *handshake) {
         strncpy(headers->connection, start, end-start);
         headers->connection[end-start] = '\0';
 
-        start = strstr(handshake, "\r\nSec-WebSocket-Protocol: ");
+        start = strcasestr(handshake, "\r\nSec-WebSocket-Protocol: ");
         if (!start) { return 0; }
         start += 26;
         end = strstr(start, "\r\n");
@@ -637,14 +637,14 @@ int parse_handshake(ws_ctx_t *ws_ctx, char *handshake) {
             strncpy(headers->key3, start, 8);
             headers->key3[8] = '\0';
 
-            start = strstr(handshake, "\r\nSec-WebSocket-Key1: ");
+            start = strcasestr(handshake, "\r\nSec-WebSocket-Key1: ");
             if (!start) { return 0; }
             start += 22;
             end = strstr(start, "\r\n");
             strncpy(headers->key1, start, end-start);
             headers->key1[end-start] = '\0';
 
-            start = strstr(handshake, "\r\nSec-WebSocket-Key2: ");
+            start = strcasestr(handshake, "\r\nSec-WebSocket-Key2: ");
             if (!start) { return 0; }
             start += 22;
             end = strstr(start, "\r\n");

common/rfb/CMakeLists.txt

@@ -58,6 +58,7 @@ set(RFB_SOURCES
   TightEncoder.cxx
   TightJPEGEncoder.cxx
   TightWEBPEncoder.cxx
+  TightX264Encoder.cxx
   UpdateTracker.cxx
   VNCSConnectionST.cxx
   VNCServerST.cxx
@@ -65,6 +66,11 @@ set(RFB_SOURCES
   ZRLEDecoder.cxx
   encodings.cxx
   util.cxx
+  mp4.c
+  mp4_bitbuf.c
+  mp4_moov.c
+  mp4_moof.c
+  nvidia.cxx
   xxhash.c)
 
 if(UNIX)

common/rfb/CMsgHandler.h

@@ -50,7 +50,7 @@ namespace rfb {
                                         int w, int h,
                                         const ScreenSet& layout);
     virtual void setCursor(int width, int height, const Point& hotspot,
-                           const rdr::U8* data) = 0;
+                           const rdr::U8* data, const bool resizing = false) = 0;
     virtual void setPixelFormat(const PixelFormat& pf);
     virtual void setName(const char* name);
     virtual void fence(rdr::U32 flags, unsigned len, const char data[]);

common/rfb/EncodeManager.cxx

@@ -40,6 +40,7 @@
 #include <rfb/TightEncoder.h>
 #include <rfb/TightJPEGEncoder.h>
 #include <rfb/TightWEBPEncoder.h>
+#include <rfb/TightX264Encoder.h>
 
 using namespace rfb;
 
@@ -68,6 +69,7 @@ enum EncoderClass {
   encoderTight,
   encoderTightJPEG,
   encoderTightWEBP,
+  encoderTightX264,
   encoderZRLE,
   encoderClassMax,
 };
@@ -110,6 +112,8 @@ static const char *encoderClassName(EncoderClass klass)
     return "Tight (JPEG)";
   case encoderTightWEBP:
     return "Tight (WEBP)";
+  case encoderTightX264:
+    return "Tight (X264)";
   case encoderZRLE:
     return "ZRLE";
   case encoderClassMax:
@@ -170,6 +174,7 @@ EncodeManager::EncodeManager(SConnection* conn_, EncCache *encCache_) : conn(con
   encoders[encoderTight] = new TightEncoder(conn);
   encoders[encoderTightJPEG] = new TightJPEGEncoder(conn);
   encoders[encoderTightWEBP] = new TightWEBPEncoder(conn);
+  encoders[encoderTightX264] = new TightX264Encoder(conn, encCache, encoderTightX264);
   encoders[encoderZRLE] = new ZRLEEncoder(conn);
 
   webpBenchResult = ((TightWEBPEncoder *) encoders[encoderTightWEBP])->benchmark();
@@ -393,20 +398,33 @@ void EncodeManager::doUpdate(bool allowLossy, const Region& changed_,
 
     conn->writer()->writeFramebufferUpdateStart(nRects);
 
-    writeCopyRects(copied, copyDelta);
-    writeCopyPassRects(copypassed);
-
     /*
-     * We start by searching for solid rects, which are then removed
-     * from the changed region.
+     * In extra-low-quality mode, if x264 is enabled, send entire screen frames
      */
-    if (conn->cp.supportsLastRect)
-      writeSolidRects(&changed, pb);
-
-    writeRects(changed, pb,
-               &start, true);
-    if (!videoDetected) // In case detection happened between the calls
-      writeRects(cursorRegion, renderedCursor);
+    if (rfb::Server::x264Bitrate && videoDetected &&
+        ((TightX264Encoder *) encoders[encoderTightX264])->tryInit(pb)) {
+      std::vector<Rect> rects;
+      changed.get_rects(&rects);
+      updateVideoStats(rects, pb);
+
+      writeSubRect(pb->getRect(), pb, encoderFullColour, Palette(), std::vector<uint8_t>(),
+                   false);
+    } else {
+      writeCopyRects(copied, copyDelta);
+      writeCopyPassRects(copypassed);
+
+      /*
+       * We start by searching for solid rects, which are then removed
+       * from the changed region.
+       */
+      if (conn->cp.supportsLastRect)
+        writeSolidRects(&changed, pb);
+
+      writeRects(changed, pb,
+                 &start, true);
+      if (!videoDetected) // In case detection happened between the calls
+        writeRects(cursorRegion, renderedCursor);
+    }
 
     updateQualities();
 
@@ -617,6 +635,8 @@ Encoder *EncodeManager::startRect(const Rect& rect, int type, const bool trackQu
   klass = activeEncoders[activeType];
   if (isWebp)
     klass = encoderTightWEBP;
+  else if (rfb::Server::x264Bitrate && videoDetected) // if x264 enabled
+    klass = encoderTightX264;
 
   beforeLength = conn->getOutStream()->length();
 
@@ -656,6 +676,8 @@ void EncodeManager::endRect(const uint8_t isWebp)
   klass = activeEncoders[activeType];
   if (isWebp)
     klass = encoderTightWEBP;
+  else if (rfb::Server::x264Bitrate && videoDetected) // if x264 enabled
+    klass = encoderTightX264;
   stats[klass][activeType].bytes += length;
 }
 
@@ -861,6 +883,8 @@ void EncodeManager::updateVideoStats(const std::vector<Rect> &rects, const Pixel
   uint32_t i;
 
   if (!rfb::Server::videoTime) {
+      if (!videoDetected)
+        ((TightX264Encoder *) encoders[encoderTightX264])->setKeyframe();
     videoDetected = true;
     return;
   }
@@ -888,6 +912,8 @@ void EncodeManager::updateVideoStats(const std::vector<Rect> &rects, const Pixel
 
   if (area > (unsigned) rfb::Server::videoArea) {
     // Initiate low-quality video mode
+    if (!videoDetected)
+      ((TightX264Encoder *) encoders[encoderTightX264])->setKeyframe();
     videoDetected = true;
     videoTimer.start(1000 * rfb::Server::videoOutTime);
   }
@@ -1022,11 +1048,6 @@ void EncodeManager::writeRects(const Region& changed, const PixelBuffer* pb,
   webpTookTooLong = false;
   changed.get_rects(&rects);
 
-  // Update stats
-  if (mainScreen) {
-    updateVideoStats(rects, pb);
-  }
-
   if (videoDetected) {
     rects.clear();
     rects.push_back(pb->getRect());
@@ -1171,6 +1192,11 @@ void EncodeManager::writeRects(const Region& changed, const PixelBuffer* pb,
     writeSubRect(subrects[i], pb, encoderTypes[i], palettes[i], compresseds[i], isWebp[i]);
   }
 
+  // Update stats
+  if (mainScreen) {
+    updateVideoStats(rects, pb);
+  }
+
   if (scaledpb)
     delete scaledpb;
 }

common/rfb/ServerCore.cxx

@@ -199,6 +199,10 @@ rfb::IntParameter rfb::Server::videoScaling
 ("VideoScaling",
  "Scaling method to use when in downscaled video mode. 0 = nearest, 1 = bilinear, 2 = prog bilinear",
  2, 0, 2);
+rfb::IntParameter rfb::Server::x264Bitrate
+("x264Bitrate",
+ "Enable x264 encoding for full-screen video, in kbps. Default 0 (off)",
+ 0, 0, 50000);
 rfb::BoolParameter rfb::Server::printVideoArea
 ("PrintVideoArea",
  "Print the detected video area % value.",

common/rfb/ServerCore.h

@@ -59,6 +59,7 @@ namespace rfb {
     static IntParameter videoOutTime;
     static IntParameter videoArea;
     static IntParameter videoScaling;
+    static IntParameter x264Bitrate;
     static StringParameter kasmPasswordFile;
     static BoolParameter printVideoArea;
     static BoolParameter protocol3_3;

common/rfb/TightConstants.h

@@ -24,7 +24,8 @@ namespace rfb {
   const unsigned int tightJpeg = 0x09;
   const unsigned int tightPng = 0x0a;
   const unsigned int tightWebp = 0x0b;
-  const unsigned int tightMaxSubencoding = 0x0b;
+  const unsigned int tightX264 = 0x0c;
+  const unsigned int tightMaxSubencoding = 0x0c;
 
   // Filters to improve compression efficiency
   const unsigned int tightFilterCopy = 0x00;

common/rfb/TightX264Encoder.cxx

@@ -0,0 +1,507 @@
+/* Copyright (C) 2000-2003 Constantin Kaplinsky.  All Rights Reserved.
+ * Copyright (C) 2011 D. R. Commander.  All Rights Reserved.
+ * Copyright 2014 Pierre Ossman for Cendio AB
+ *
+ * This is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This software is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this software; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111-1307,
+ * USA.
+ */
+#include <rdr/OutStream.h>
+#include <rfb/EncCache.h>
+#include <rfb/encodings.h>
+#include <rfb/LogWriter.h>
+#include <rfb/SConnection.h>
+#include <rfb/ServerCore.h>
+#include <rfb/PixelBuffer.h>
+#include <rfb/TightX264Encoder.h>
+#include <rfb/TightConstants.h>
+
+#include <webp/encode.h>
+#include <x264.h>
+#include "nvidia.h"
+#include "mp4.h"
+
+#define MAX_FRAMELEN (1024 * 1024)
+
+using namespace rfb;
+
+static LogWriter vlog("x264");
+
+static const PixelFormat pfRGBX(32, 24, false, true, 255, 255, 255, 0, 8, 16);
+static const PixelFormat pfBGRX(32, 24, false, true, 255, 255, 255, 16, 8, 0);
+
+bool TightX264Encoder::skip_nvidia = false;
+
+TightX264Encoder::TightX264Encoder(SConnection* conn, EncCache *cache_, uint8_t cacheType_) :
+  Encoder(conn, encodingTight, (EncoderFlags)(EncoderUseNativePF | EncoderLossy), -1),
+  keyframe(true), enc(NULL), params(NULL), mux(NULL), muxstate(NULL), framectr(0),
+  nvidia_init_done(false), using_nvidia(true),
+  encCache(cache_), cacheType(cacheType_),
+  framebuf(NULL), framelen(0), bitbuf(NULL), myw(0), myh(0)
+{
+  params = new x264_param_t;
+  x264_param_default_preset(params, "veryfast", "zerolatency");
+
+  params->i_threads = X264_THREADS_AUTO;
+  params->i_fps_num = params->i_keyint_max = rfb::Server::frameRate;
+  params->i_fps_den = 1;
+  params->rc.i_rc_method = X264_RC_ABR;
+  params->rc.i_bitrate = rfb::Server::x264Bitrate;
+  params->i_csp = X264_CSP_I420;
+  params->i_log_level = X264_LOG_WARNING;
+  params->b_annexb = 0;
+
+  framebuf = new uint8_t[MAX_FRAMELEN];
+  bitbuf = new uint8_t[MAX_FRAMELEN];
+  mux = new Mp4Context;
+  memset(mux, 0, sizeof(Mp4Context));
+  muxstate = new Mp4State;
+  memset(muxstate, 0, sizeof(Mp4State));
+}
+
+TightX264Encoder::~TightX264Encoder()
+{
+  delete params;
+  delete [] framebuf;
+  delete [] bitbuf;
+  delete mux;
+  delete muxstate;
+}
+
+bool TightX264Encoder::isSupported()
+{
+  if (!conn->cp.supportsEncoding(encodingTight))
+    return false;
+
+  // Unconditional support if enabled
+  return rfb::Server::x264Bitrate != 0;
+}
+
+void TightX264Encoder::mp4_write_callback(const void *buffer, size_t size)
+{
+  if (framelen + size > MAX_FRAMELEN)
+    vlog.error("Tried to write too large a frame, %lu bytes", framelen + size);
+
+  memcpy(&framebuf[framelen], buffer, size);
+  framelen += size;
+}
+
+void TightX264Encoder::writeRect(const PixelBuffer* pb, const Palette& palette)
+{
+  const rdr::U8* buffer;
+  int stride;
+
+  rdr::OutStream* os;
+
+  if (pb->width() < 320)
+    return; // Sometimes we get sent an 1x1 frame, or a cursor
+
+  uint32_t w, h;
+  w = pb->width();
+  h = pb->height();
+
+  os = conn->getOutStream();
+
+  if (using_nvidia) {
+
+    if (w != myw || h != myh) {
+      if (nvidia_init_done)
+        nvidia_unload();
+      nvidia_init_done = false;
+    }
+
+    if (!nvidia_init_done) {
+      if (nvidia_init(w, h, rfb::Server::x264Bitrate,
+                  rfb::Server::frameRate) != 0) {
+        vlog.error("nvidia init failed, disabling h264");
+        rfb::Server::x264Bitrate.setParam(0);
+        return;
+      }
+      nvidia_init_done = true;
+      myw = w;
+      myh = h;
+    }
+
+    uint32_t cachelen;
+    const void *cachedata;
+    if (encCache->enabled &&
+        (cachedata = encCache->get(cacheType, framectr, 0, w, h, cachelen))) {
+      os->writeU8(tightX264 << 4);
+      writeCompact(cachelen, os);
+      os->writeBytes(cachedata, cachelen);
+      framectr++;
+      return;
+    }
+
+    if (keyframe) {
+      framectr = 0;
+      keyframe = false;
+
+      free(mux->buf_header.buf);
+      free(mux->buf_mdat.buf);
+      free(mux->buf_moof.buf);
+      memset(mux, 0, sizeof(Mp4Context));
+      memset(muxstate, 0, sizeof(Mp4State));
+    }
+
+    mux->framerate = rfb::Server::frameRate;
+    mux->w = w;
+    mux->h = h;
+
+    buffer = pb->getBuffer(pb->getRect(), &stride);
+
+    if (!pfBGRX.equal(pb->getPF())) {
+      vlog.error("unsupported pixel format");
+      return;
+    }
+
+    // Encode
+    uint32_t bitlen;
+    if (nvenc_frame(buffer, framectr++, bitbuf, bitlen) != 0) {
+      vlog.error("encoding failed");
+      return;
+    }
+
+    // Need to parse NALs out of the stream
+    const uint8_t prefix[3] = { 0, 0, 1 };
+    const uint8_t *nalptr = bitbuf;
+    int i_nals = 0;
+    const uint8_t *nalstarts[32] = { NULL };
+    uint32_t nallens[32] = { 0 };
+    uint32_t remlen = bitlen;
+
+    while (1) {
+      const uint8_t *next = (uint8_t *) memmem(nalptr, remlen, prefix, 3);
+      if (!next)
+        break;
+
+      remlen -= (next + 3) - nalptr;
+      nalptr = nalstarts[i_nals] = next + 3;
+
+      i_nals++;
+    };
+
+    // Lens
+    int i;
+    for (i = 0; i < i_nals; i++) {
+      if (i == i_nals - 1) {
+        nallens[i] = bitbuf + bitlen - nalstarts[i];
+      } else {
+        nallens[i] = nalstarts[i + 1] - nalstarts[i] - 3;
+      }
+    }
+
+    // Mux
+    framelen = 0;
+    os->writeU8(tightX264 << 4);
+
+    for (i = 0; i < i_nals; i++) {
+      uint32_t pack_len = nallens[i];
+      const uint8_t *pack_data = nalstarts[i];
+
+      struct NAL nal; nal_parse_header(&nal, pack_data[0]);
+
+      switch (nal.unit_type) {
+        case NalUnitType_SPS: { set_sps(mux, pack_data, pack_len); break; }
+        case NalUnitType_PPS: { set_pps(mux, pack_data, pack_len); break; }
+        case NalUnitType_CodedSliceIdr:
+        case NalUnitType_CodedSliceNonIdr: {
+            // Write all remaining NALs under the assumption they are the same type.
+            const uint32_t origlen = pack_len;
+            pack_len = bitbuf + bitlen - pack_data;
+            set_slice(mux, pack_data, origlen, pack_len, nal.unit_type);
+            break;
+        }
+        default: break;
+      }
+
+      if (nal.unit_type != NalUnitType_CodedSliceIdr &&
+          nal.unit_type != NalUnitType_CodedSliceNonIdr)
+          continue;
+
+      enum BufError err;
+      if (!muxstate->header_sent) {
+        struct BitBuf header_buf;
+        err = get_header(mux, &header_buf); chk_err_continue
+        mp4_write_callback(header_buf.buf, header_buf.offset);
+
+        muxstate->sequence_number = 1;
+        muxstate->base_data_offset = header_buf.offset;
+        muxstate->base_media_decode_time = 0;
+        muxstate->header_sent = true;
+        muxstate->nals_count = 0;
+        muxstate->default_sample_duration = default_sample_size;
+      }
+
+      err = set_mp4_state(mux, muxstate); chk_err_continue
+      {
+        struct BitBuf moof_buf;
+        err = get_moof(mux, &moof_buf); chk_err_continue
+        mp4_write_callback(moof_buf.buf, moof_buf.offset);
+      }
+      {
+        struct BitBuf mdat_buf;
+        err = get_mdat(mux, &mdat_buf); chk_err_continue
+        mp4_write_callback(mdat_buf.buf, mdat_buf.offset);
+      }
+
+      break;
+    }
+
+    if (encCache->enabled) {
+      void *tmp = malloc(framelen);
+      memcpy(tmp, framebuf, framelen);
+      encCache->add(cacheType, framectr, 0, w, h, framelen, tmp);
+    }
+
+    writeCompact(framelen, os);
+    os->writeBytes(framebuf, framelen);
+  } else {
+    w += w & 1;
+    h += h & 1;
+
+    params->i_width = w;
+    params->i_height = h;
+
+    x264_param_apply_profile(params, "baseline");
+
+    uint32_t cachelen;
+    const void *cachedata;
+    if (encCache->enabled &&
+        (cachedata = encCache->get(cacheType, framectr, 0, w, h, cachelen))) {
+      os->writeU8(tightX264 << 4);
+      writeCompact(cachelen, os);
+      os->writeBytes(cachedata, cachelen);
+      framectr++;
+      return;
+    }
+
+    if (keyframe) {
+      framectr = 0;
+      keyframe = false;
+
+      free(mux->buf_header.buf);
+      free(mux->buf_mdat.buf);
+      free(mux->buf_moof.buf);
+      memset(mux, 0, sizeof(Mp4Context));
+      memset(muxstate, 0, sizeof(Mp4State));
+    }
+
+    mux->framerate = rfb::Server::frameRate;
+    mux->w = params->i_width;
+    mux->h = params->i_height;
+
+    if (!enc) {
+      enc = x264_encoder_open(params);
+    }
+
+    buffer = pb->getBuffer(pb->getRect(), &stride);
+
+    // Convert it to yuv420 using libwebp's helper functions
+    WebPPicture pic;
+
+    WebPPictureInit(&pic);
+    pic.width = pb->getRect().width();
+    pic.height = pb->getRect().height();
+
+    bool freebuffer = false;
+    if (pic.width & 1 || pic.height & 1) {
+      // Expand to divisible-by-2 for x264
+      freebuffer = true;
+      const uint32_t oldw = pic.width;
+      const uint32_t oldh = pic.height;
+      pic.width += pic.width & 1;
+      pic.height += pic.height & 1;
+      stride = pic.width;
+      const rdr::U8 *oldbuffer = buffer;
+      buffer = (const rdr::U8*) calloc(pic.width * pic.height, 4);
+
+      uint32_t y;
+      for (y = 0; y < oldh; y++)
+        memcpy((void *) &buffer[y * stride * 4], &oldbuffer[y * oldw * 4], oldw * 4);
+    }
+
+    if (pfRGBX.equal(pb->getPF())) {
+      WebPPictureImportRGBX(&pic, buffer, stride * 4);
+    } else if (pfBGRX.equal(pb->getPF())) {
+      WebPPictureImportBGRX(&pic, buffer, stride * 4);
+    } else {
+      rdr::U8* tmpbuf = new rdr::U8[pic.width * pic.height * 3];
+      pb->getPF().rgbFromBuffer(tmpbuf, (const rdr::U8 *) buffer, pic.width, stride, pic.height);
+      stride = pic.width * 3;
+
+      WebPPictureImportRGB(&pic, tmpbuf, stride);
+      delete [] tmpbuf;
+    }
+
+    if (freebuffer)
+      free((void *) buffer);
+
+    // Wrap
+    x264_picture_t pic_in, pic_out;
+    x264_picture_init(&pic_in);
+
+    pic_in.img.i_csp = X264_CSP_I420;
+    pic_in.img.i_plane = 3;
+
+    pic_in.img.plane[0] = pic.y;
+    pic_in.img.plane[1] = pic.u;
+    pic_in.img.plane[2] = pic.v;
+
+    pic_in.img.i_stride[0] = pic.y_stride;
+    pic_in.img.i_stride[1] = pic_in.img.i_stride[2] = pic.uv_stride;
+
+    pic_in.i_pts = framectr++;
+
+    // Encode
+    int i_nals;
+    x264_nal_t *nals;
+    const int len = x264_encoder_encode(enc, &nals, &i_nals, &pic_in, &pic_out);
+
+    if (len <= 0 || i_nals <= 0)
+      vlog.info("encoding error");
+
+    // Mux
+    framelen = 0;
+    os->writeU8(tightX264 << 4);
+
+    int i;
+    for (i = 0; i < i_nals; i++) {
+      uint32_t pack_len = nals[i].i_payload - 4;
+      const uint8_t *pack_data = nals[i].p_payload;
+
+      pack_data += 4; // Skip size
+
+      struct NAL nal; nal_parse_header(&nal, pack_data[0]);
+
+      switch (nal.unit_type) {
+        case NalUnitType_SPS: { set_sps(mux, pack_data, pack_len); break; }
+        case NalUnitType_PPS: { set_pps(mux, pack_data, pack_len); break; }
+        case NalUnitType_CodedSliceIdr:
+        case NalUnitType_CodedSliceNonIdr: {
+            // Write all remaining NALs under the assumption they are the same type.
+            const uint32_t origlen = pack_len;
+            int j;
+            for (j = i + 1; j < i_nals; j++)
+                pack_len += nals[j].i_payload;
+            set_slice(mux, pack_data, origlen, pack_len, nal.unit_type);
+            break;
+        }
+        default: break;
+      }
+
+      if (nal.unit_type != NalUnitType_CodedSliceIdr &&
+          nal.unit_type != NalUnitType_CodedSliceNonIdr)
+          continue;
+
+      enum BufError err;
+      if (!muxstate->header_sent) {
+        struct BitBuf header_buf;
+        err = get_header(mux, &header_buf); chk_err_continue
+        mp4_write_callback(header_buf.buf, header_buf.offset);
+
+        muxstate->sequence_number = 1;
+        muxstate->base_data_offset = header_buf.offset;
+        muxstate->base_media_decode_time = 0;
+        muxstate->header_sent = true;
+        muxstate->nals_count = 0;
+        muxstate->default_sample_duration = default_sample_size;
+      }
+
+      err = set_mp4_state(mux, muxstate); chk_err_continue
+      {
+        struct BitBuf moof_buf;
+        err = get_moof(mux, &moof_buf); chk_err_continue
+        mp4_write_callback(moof_buf.buf, moof_buf.offset);
+      }
+      {
+        struct BitBuf mdat_buf;
+        err = get_mdat(mux, &mdat_buf); chk_err_continue
+        mp4_write_callback(mdat_buf.buf, mdat_buf.offset);
+      }
+
+      break;
+    }
+
+    if (encCache->enabled) {
+      void *tmp = malloc(framelen);
+      memcpy(tmp, framebuf, framelen);
+      encCache->add(cacheType, framectr, 0, w, h, framelen, tmp);
+    }
+
+    writeCompact(framelen, os);
+    os->writeBytes(framebuf, framelen);
+
+    // Cleanup
+    WebPPictureFree(&pic);
+    x264_encoder_close(enc);
+    enc = NULL;
+  }
+}
+
+void TightX264Encoder::writeSolidRect(int width, int height,
+                                      const PixelFormat& pf,
+                                      const rdr::U8* colour)
+{
+  // FIXME: Add a shortcut in the X264 compressor to handle this case
+  //        without having to use the default fallback which is very slow.
+  Encoder::writeSolidRect(width, height, pf, colour);
+}
+
+void TightX264Encoder::writeCompact(rdr::U32 value, rdr::OutStream* os) const
+{
+  // Copied from TightEncoder as it's overkill to inherit just for this
+  rdr::U8 b;
+
+  b = value & 0x7F;
+  if (value <= 0x7F) {
+    os->writeU8(b);
+  } else {
+    os->writeU8(b | 0x80);
+    b = value >> 7 & 0x7F;
+    if (value <= 0x3FFF) {
+      os->writeU8(b);
+    } else {
+      os->writeU8(b | 0x80);
+      os->writeU8(value >> 14 & 0xFF);
+    }
+  }
+}
+
+bool TightX264Encoder::tryInit(const PixelBuffer* pb) {
+  if (nvidia_init_done)
+    return true;
+
+  uint32_t w, h;
+  w = pb->width();
+  h = pb->height();
+
+  if (skip_nvidia || nvidia_init(w, h, rfb::Server::x264Bitrate,
+              rfb::Server::frameRate) != 0) {
+    if (!skip_nvidia)
+      vlog.error("nvidia init failed, falling back to x264");
+    using_nvidia = false;
+    nvidia_init_done = true;
+    skip_nvidia = true;
+    myw = w;
+    myh = h;
+    return true;
+  }
+
+  nvidia_init_done = true;
+  myw = w;
+  myh = h;
+
+  return true;
+}

common/rfb/TightX264Encoder.h

@@ -0,0 +1,82 @@
+/* Copyright (C) 2000-2003 Constantin Kaplinsky.  All Rights Reserved.
+ * Copyright (C) 2011 D. R. Commander
+ * Copyright 2014 Pierre Ossman for Cendio AB
+ *
+ * This is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This software is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this software; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111-1307,
+ * USA.
+ */
+#ifndef __RFB_TIGHTX264ENCODER_H__
+#define __RFB_TIGHTX264ENCODER_H__
+
+#include <rfb/Encoder.h>
+#include <stdint.h>
+#include <vector>
+
+struct x264_t;
+struct x264_param_t;
+struct Mp4Context;
+struct Mp4State;
+
+namespace rfb {
+
+  class EncCache;
+
+  class TightX264Encoder : public Encoder {
+  public:
+    TightX264Encoder(SConnection* conn, EncCache *encCache, uint8_t cacheType);
+    virtual ~TightX264Encoder();
+
+    virtual bool isSupported();
+
+    virtual void setQualityLevel(int level) {}
+    virtual void setFineQualityLevel(int quality, int subsampling) {}
+
+    virtual void writeRect(const PixelBuffer* pb, const Palette& palette);
+    virtual void writeSolidRect(int width, int height,
+                                const PixelFormat& pf,
+                                const rdr::U8* colour);
+
+    virtual void setKeyframe() { keyframe = true; }
+
+    bool tryInit(const PixelBuffer* pb);
+
+  protected:
+    void writeCompact(rdr::U32 value, rdr::OutStream* os) const;
+    void mp4_write_callback(const void *buffer, size_t size);
+
+  protected:
+    bool keyframe;
+    x264_t *enc;
+    x264_param_t *params;
+    Mp4Context *mux;
+    Mp4State *muxstate;
+    unsigned framectr;
+
+    bool nvidia_init_done;
+    bool using_nvidia;
+    static bool skip_nvidia;
+
+    EncCache *encCache;
+    uint8_t cacheType;
+  public:
+    uint8_t *framebuf;
+    uint32_t framelen;
+
+    uint8_t *bitbuf;
+
+    uint16_t myw, myh;
+  };
+}
+#endif

common/rfb/VNCServer.h

@@ -79,7 +79,7 @@ namespace rfb {
     // cursorData argument contains width*height rgba quadruplets with
     // non-premultiplied alpha.
     virtual void setCursor(int width, int height, const Point& hotspot,
-                           const rdr::U8* cursorData) = 0;
+                           const rdr::U8* cursorData, const bool resizing = false) = 0;
 
     // setCursorPos() tells the server the current position of the cursor, and
     // whether the server initiated that change (e.g. through another X11

common/rfb/VNCServerST.cxx

@@ -579,7 +579,7 @@ void VNCServerST::add_copied(const Region& dest, const Point& delta)
 }
 
 void VNCServerST::setCursor(int width, int height, const Point& newHotspot,
-                            const rdr::U8* data)
+                            const rdr::U8* data, const bool resizing)
 {
   delete cursor;
   cursor = new Cursor(width, height, newHotspot, data);
@@ -587,6 +587,13 @@ void VNCServerST::setCursor(int width, int height, const Point& newHotspot,
 
   renderedCursorInvalid = true;
 
+  // If an app has an animated cursor on the resized edge, X internals
+  // will call for it to be rendered. Unlucky for us, the VNC screen
+  // is currently pointing to freed memory, and a cursor change
+  // would want to send a screen update. So, don't do that.
+  if (resizing)
+    return;
+
   std::list<VNCSConnectionST*>::iterator ci, ci_next;
   for (ci = clients.begin(); ci != clients.end(); ci = ci_next) {
     ci_next = ci; ci_next++;

common/rfb/VNCServerST.h

@@ -102,7 +102,7 @@ namespace rfb {
     virtual void add_changed(const Region &region);
     virtual void add_copied(const Region &dest, const Point &delta);
     virtual void setCursor(int width, int height, const Point& hotspot,
-                           const rdr::U8* data);
+                           const rdr::U8* data, const bool resizing = false);
     virtual void setCursorPos(const Point& p, bool warped);
     virtual void setLEDState(unsigned state);
 

common/rfb/dynlink_cuda.h

@@ -0,0 +1,423 @@
+/*
+ * This copyright notice applies to this header file only:
+ *
+ * Copyright (c) 2016
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use,
+ * copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the software, and to permit persons to whom the
+ * software is furnished to do so, subject to the following
+ * conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+ * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#if !defined(FFNV_DYNLINK_CUDA_H) && !defined(CUDA_VERSION)
+#define FFNV_DYNLINK_CUDA_H
+
+#include <stddef.h>
+
+#define CUDA_VERSION 7050
+
+#if defined(_WIN32) || defined(__CYGWIN__)
+#define CUDAAPI __stdcall
+#else
+#define CUDAAPI
+#endif
+
+#define CU_CTX_SCHED_BLOCKING_SYNC 4
+
+typedef int CUdevice;
+#if defined(__x86_64) || defined(AMD64) || defined(_M_AMD64) || defined(__LP64__) || defined(__aarch64__)
+typedef unsigned long long CUdeviceptr;
+#else
+typedef unsigned int CUdeviceptr;
+#endif
+typedef unsigned long long CUtexObject;
+
+typedef struct CUarray_st            *CUarray;
+typedef struct CUctx_st              *CUcontext;
+typedef struct CUstream_st           *CUstream;
+typedef struct CUevent_st            *CUevent;
+typedef struct CUfunc_st             *CUfunction;
+typedef struct CUmod_st              *CUmodule;
+typedef struct CUmipmappedArray_st   *CUmipmappedArray;
+typedef struct CUgraphicsResource_st *CUgraphicsResource;
+typedef struct CUextMemory_st        *CUexternalMemory;
+typedef struct CUextSemaphore_st     *CUexternalSemaphore;
+
+typedef struct CUlinkState_st *CUlinkState;
+
+typedef enum cudaError_enum {
+    CUDA_SUCCESS = 0,
+    CUDA_ERROR_NOT_READY = 600
+} CUresult;
+
+/**
+ * Device properties (subset)
+ */
+typedef enum CUdevice_attribute_enum {
+    CU_DEVICE_ATTRIBUTE_CLOCK_RATE = 13,
+    CU_DEVICE_ATTRIBUTE_TEXTURE_ALIGNMENT = 14,
+    CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT = 16,
+    CU_DEVICE_ATTRIBUTE_INTEGRATED = 18,
+    CU_DEVICE_ATTRIBUTE_CAN_MAP_HOST_MEMORY = 19,
+    CU_DEVICE_ATTRIBUTE_COMPUTE_MODE = 20,
+    CU_DEVICE_ATTRIBUTE_CONCURRENT_KERNELS = 31,
+    CU_DEVICE_ATTRIBUTE_PCI_BUS_ID = 33,
+    CU_DEVICE_ATTRIBUTE_PCI_DEVICE_ID = 34,
+    CU_DEVICE_ATTRIBUTE_TCC_DRIVER = 35,
+    CU_DEVICE_ATTRIBUTE_MEMORY_CLOCK_RATE = 36,
+    CU_DEVICE_ATTRIBUTE_GLOBAL_MEMORY_BUS_WIDTH = 37,
+    CU_DEVICE_ATTRIBUTE_ASYNC_ENGINE_COUNT = 40,
+    CU_DEVICE_ATTRIBUTE_UNIFIED_ADDRESSING = 41,
+    CU_DEVICE_ATTRIBUTE_PCI_DOMAIN_ID = 50,
+    CU_DEVICE_ATTRIBUTE_TEXTURE_PITCH_ALIGNMENT = 51,
+    CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MAJOR = 75,
+    CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MINOR = 76,
+    CU_DEVICE_ATTRIBUTE_MANAGED_MEMORY = 83,
+    CU_DEVICE_ATTRIBUTE_MULTI_GPU_BOARD = 84,
+    CU_DEVICE_ATTRIBUTE_MULTI_GPU_BOARD_GROUP_ID = 85,
+} CUdevice_attribute;
+
+typedef enum CUarray_format_enum {
+    CU_AD_FORMAT_UNSIGNED_INT8  = 0x01,
+    CU_AD_FORMAT_UNSIGNED_INT16 = 0x02,
+    CU_AD_FORMAT_UNSIGNED_INT32 = 0x03,
+    CU_AD_FORMAT_SIGNED_INT8    = 0x08,
+    CU_AD_FORMAT_SIGNED_INT16   = 0x09,
+    CU_AD_FORMAT_SIGNED_INT32   = 0x0a,
+    CU_AD_FORMAT_HALF           = 0x10,
+    CU_AD_FORMAT_FLOAT          = 0x20
+} CUarray_format;
+
+typedef enum CUmemorytype_enum {
+    CU_MEMORYTYPE_HOST = 1,
+    CU_MEMORYTYPE_DEVICE = 2,
+    CU_MEMORYTYPE_ARRAY = 3
+} CUmemorytype;
+
+typedef enum CUlimit_enum {
+    CU_LIMIT_STACK_SIZE = 0,
+    CU_LIMIT_PRINTF_FIFO_SIZE = 1,
+    CU_LIMIT_MALLOC_HEAP_SIZE = 2,
+    CU_LIMIT_DEV_RUNTIME_SYNC_DEPTH = 3,
+    CU_LIMIT_DEV_RUNTIME_PENDING_LAUNCH_COUNT = 4
+} CUlimit;
+
+typedef enum CUresourcetype_enum {
+    CU_RESOURCE_TYPE_ARRAY           = 0x00,
+    CU_RESOURCE_TYPE_MIPMAPPED_ARRAY = 0x01,
+    CU_RESOURCE_TYPE_LINEAR          = 0x02,
+    CU_RESOURCE_TYPE_PITCH2D         = 0x03
+} CUresourcetype;
+
+typedef enum CUaddress_mode_enum {
+    CU_TR_ADDRESS_MODE_WRAP = 0,
+    CU_TR_ADDRESS_MODE_CLAMP = 1,
+    CU_TR_ADDRESS_MODE_MIRROR = 2,
+    CU_TR_ADDRESS_MODE_BORDER = 3
+} CUaddress_mode;
+
+typedef enum CUfilter_mode_enum {
+    CU_TR_FILTER_MODE_POINT = 0,
+    CU_TR_FILTER_MODE_LINEAR = 1
+} CUfilter_mode;
+
+typedef enum CUgraphicsRegisterFlags_enum {
+    CU_GRAPHICS_REGISTER_FLAGS_NONE = 0,
+    CU_GRAPHICS_REGISTER_FLAGS_READ_ONLY = 1,
+    CU_GRAPHICS_REGISTER_FLAGS_WRITE_DISCARD = 2,
+    CU_GRAPHICS_REGISTER_FLAGS_SURFACE_LDST = 4,
+    CU_GRAPHICS_REGISTER_FLAGS_TEXTURE_GATHER = 8
+} CUgraphicsRegisterFlags;
+
+typedef enum CUexternalMemoryHandleType_enum {
+    CU_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_FD        = 1,
+    CU_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32     = 2,
+    CU_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32_KMT = 3,
+    CU_EXTERNAL_MEMORY_HANDLE_TYPE_D3D12_HEAP       = 4,
+    CU_EXTERNAL_MEMORY_HANDLE_TYPE_D3D12_RESOURCE   = 5,
+} CUexternalMemoryHandleType;
+
+typedef enum CUexternalSemaphoreHandleType_enum {
+    CU_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD        = 1,
+    CU_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_WIN32     = 2,
+    CU_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_WIN32_KMT = 3,
+    CU_EXTERNAL_SEMAPHORE_HANDLE_TYPE_D3D12_FENCE      = 4
+} CUexternalSemaphoreHandleType;
+
+typedef enum CUjit_option_enum
+{
+    CU_JIT_MAX_REGISTERS               = 0,
+    CU_JIT_THREADS_PER_BLOCK           = 1,
+    CU_JIT_WALL_TIME                   = 2,
+    CU_JIT_INFO_LOG_BUFFER             = 3,
+    CU_JIT_INFO_LOG_BUFFER_SIZE_BYTES  = 4,
+    CU_JIT_ERROR_LOG_BUFFER            = 5,
+    CU_JIT_ERROR_LOG_BUFFER_SIZE_BYTES = 6,
+    CU_JIT_OPTIMIZATION_LEVEL          = 7,
+    CU_JIT_TARGET_FROM_CUCONTEXT       = 8,
+    CU_JIT_TARGET                      = 9,
+    CU_JIT_FALLBACK_STRATEGY           = 10,
+    CU_JIT_GENERATE_DEBUG_INFO         = 11,
+    CU_JIT_LOG_VERBOSE                 = 12,
+    CU_JIT_GENERATE_LINE_INFO          = 13,
+    CU_JIT_CACHE_MODE                  = 14,
+    CU_JIT_NEW_SM3X_OPT                = 15,
+    CU_JIT_FAST_COMPILE                = 16,
+    CU_JIT_GLOBAL_SYMBOL_NAMES         = 17,
+    CU_JIT_GLOBAL_SYMBOL_ADDRESSES     = 18,
+    CU_JIT_GLOBAL_SYMBOL_COUNT         = 19,
+    CU_JIT_NUM_OPTIONS
+} CUjit_option;
+
+typedef enum CUjitInputType_enum
+{
+    CU_JIT_INPUT_CUBIN     = 0,
+    CU_JIT_INPUT_PTX       = 1,
+    CU_JIT_INPUT_FATBINARY = 2,
+    CU_JIT_INPUT_OBJECT    = 3,
+    CU_JIT_INPUT_LIBRARY   = 4,
+    CU_JIT_NUM_INPUT_TYPES
+} CUjitInputType;
+
+#ifndef CU_UUID_HAS_BEEN_DEFINED
+#define CU_UUID_HAS_BEEN_DEFINED
+typedef struct CUuuid_st {
+    char bytes[16];
+} CUuuid;
+#endif
+
+typedef struct CUDA_MEMCPY2D_st {
+    size_t srcXInBytes;
+    size_t srcY;
+    CUmemorytype srcMemoryType;
+    const void *srcHost;
+    CUdeviceptr srcDevice;
+    CUarray srcArray;
+    size_t srcPitch;
+
+    size_t dstXInBytes;
+    size_t dstY;
+    CUmemorytype dstMemoryType;
+    void *dstHost;
+    CUdeviceptr dstDevice;
+    CUarray dstArray;
+    size_t dstPitch;
+
+    size_t WidthInBytes;
+    size_t Height;
+} CUDA_MEMCPY2D;
+
+typedef struct CUDA_RESOURCE_DESC_st {
+    CUresourcetype resType;
+    union {
+        struct {
+            CUarray hArray;
+        } array;
+        struct {
+            CUmipmappedArray hMipmappedArray;
+        } mipmap;
+        struct {
+            CUdeviceptr devPtr;
+            CUarray_format format;
+            unsigned int numChannels;
+            size_t sizeInBytes;
+        } linear;
+        struct {
+            CUdeviceptr devPtr;
+            CUarray_format format;
+            unsigned int numChannels;
+            size_t width;
+            size_t height;
+            size_t pitchInBytes;
+        } pitch2D;
+        struct {
+            int reserved[32];
+        } reserved;
+    } res;
+    unsigned int flags;
+} CUDA_RESOURCE_DESC;
+
+typedef struct CUDA_TEXTURE_DESC_st {
+    CUaddress_mode addressMode[3];
+    CUfilter_mode filterMode;
+    unsigned int flags;
+    unsigned int maxAnisotropy;
+    CUfilter_mode mipmapFilterMode;
+    float mipmapLevelBias;
+    float minMipmapLevelClamp;
+    float maxMipmapLevelClamp;
+    float borderColor[4];
+    int reserved[12];
+} CUDA_TEXTURE_DESC;
+
+/* Unused type */
+typedef struct CUDA_RESOURCE_VIEW_DESC_st CUDA_RESOURCE_VIEW_DESC;
+
+typedef unsigned int GLenum;
+typedef unsigned int GLuint;
+
+typedef enum CUGLDeviceList_enum {
+    CU_GL_DEVICE_LIST_ALL = 1,
+    CU_GL_DEVICE_LIST_CURRENT_FRAME = 2,
+    CU_GL_DEVICE_LIST_NEXT_FRAME = 3,
+} CUGLDeviceList;
+
+typedef struct CUDA_EXTERNAL_MEMORY_HANDLE_DESC_st {
+    CUexternalMemoryHandleType type;
+    union {
+        int fd;
+        struct {
+            void *handle;
+            const void *name;
+        } win32;
+    } handle;
+    unsigned long long size;
+    unsigned int flags;
+    unsigned int reserved[16];
+} CUDA_EXTERNAL_MEMORY_HANDLE_DESC;
+
+typedef struct CUDA_EXTERNAL_MEMORY_BUFFER_DESC_st {
+    unsigned long long offset;
+    unsigned long long size;
+    unsigned int flags;
+    unsigned int reserved[16];
+} CUDA_EXTERNAL_MEMORY_BUFFER_DESC;
+
+typedef struct CUDA_EXTERNAL_SEMAPHORE_HANDLE_DESC_st {
+    CUexternalSemaphoreHandleType type;
+    union {
+        int fd;
+        struct {
+            void *handle;
+            const void *name;
+        } win32;
+    } handle;
+    unsigned int flags;
+    unsigned int reserved[16];
+} CUDA_EXTERNAL_SEMAPHORE_HANDLE_DESC;
+
+typedef struct CUDA_EXTERNAL_SEMAPHORE_SIGNAL_PARAMS_st {
+    struct {
+        struct {
+            unsigned long long value;
+        } fence;
+        unsigned int reserved[16];
+    } params;
+    unsigned int flags;
+    unsigned int reserved[16];
+} CUDA_EXTERNAL_SEMAPHORE_SIGNAL_PARAMS;
+
+typedef CUDA_EXTERNAL_SEMAPHORE_SIGNAL_PARAMS CUDA_EXTERNAL_SEMAPHORE_WAIT_PARAMS;
+
+typedef struct CUDA_ARRAY3D_DESCRIPTOR_st {
+    size_t Width;
+    size_t Height;
+    size_t Depth;
+
+    CUarray_format Format;
+    unsigned int NumChannels;
+    unsigned int Flags;
+} CUDA_ARRAY3D_DESCRIPTOR;
+
+typedef struct CUDA_EXTERNAL_MEMORY_MIPMAPPED_ARRAY_DESC_st {
+    unsigned long long offset;
+    CUDA_ARRAY3D_DESCRIPTOR arrayDesc;
+    unsigned int numLevels;
+    unsigned int reserved[16];
+} CUDA_EXTERNAL_MEMORY_MIPMAPPED_ARRAY_DESC;
+
+#define CU_STREAM_NON_BLOCKING 1
+#define CU_EVENT_BLOCKING_SYNC 1
+#define CU_EVENT_DISABLE_TIMING 2
+#define CU_TRSF_READ_AS_INTEGER 1
+
+typedef void CUDAAPI CUstreamCallback(CUstream hStream, CUresult status, void *userdata);
+
+typedef CUresult CUDAAPI tcuInit(unsigned int Flags);
+typedef CUresult CUDAAPI tcuDeviceGetCount(int *count);
+typedef CUresult CUDAAPI tcuDeviceGet(CUdevice *device, int ordinal);
+typedef CUresult CUDAAPI tcuDeviceGetAttribute(int *pi, CUdevice_attribute attrib, CUdevice dev);
+typedef CUresult CUDAAPI tcuDeviceGetName(char *name, int len, CUdevice dev);
+typedef CUresult CUDAAPI tcuDeviceGetUuid(CUuuid *uuid, CUdevice dev);
+typedef CUresult CUDAAPI tcuDeviceComputeCapability(int *major, int *minor, CUdevice dev);
+typedef CUresult CUDAAPI tcuCtxCreate_v2(CUcontext *pctx, unsigned int flags, CUdevice dev);
+typedef CUresult CUDAAPI tcuCtxSetLimit(CUlimit limit, size_t value);
+typedef CUresult CUDAAPI tcuCtxPushCurrent_v2(CUcontext pctx);
+typedef CUresult CUDAAPI tcuCtxPopCurrent_v2(CUcontext *pctx);
+typedef CUresult CUDAAPI tcuCtxDestroy_v2(CUcontext ctx);
+typedef CUresult CUDAAPI tcuMemAlloc_v2(CUdeviceptr *dptr, size_t bytesize);
+typedef CUresult CUDAAPI tcuMemAllocPitch_v2(CUdeviceptr *dptr, size_t *pPitch, size_t WidthInBytes, size_t Height, unsigned int ElementSizeBytes);
+typedef CUresult CUDAAPI tcuMemsetD8Async(CUdeviceptr dstDevice, unsigned char uc, size_t N, CUstream hStream);
+typedef CUresult CUDAAPI tcuMemFree_v2(CUdeviceptr dptr);
+typedef CUresult CUDAAPI tcuMemcpy(CUdeviceptr dst, CUdeviceptr src, size_t bytesize);
+typedef CUresult CUDAAPI tcuMemcpyAsync(CUdeviceptr dst, CUdeviceptr src, size_t bytesize, CUstream hStream);
+typedef CUresult CUDAAPI tcuMemcpy2D_v2(const CUDA_MEMCPY2D *pcopy);
+typedef CUresult CUDAAPI tcuMemcpy2DAsync_v2(const CUDA_MEMCPY2D *pcopy, CUstream hStream);
+typedef CUresult CUDAAPI tcuGetErrorName(CUresult error, const char** pstr);
+typedef CUresult CUDAAPI tcuGetErrorString(CUresult error, const char** pstr);
+typedef CUresult CUDAAPI tcuCtxGetDevice(CUdevice *device);
+
+typedef CUresult CUDAAPI tcuDevicePrimaryCtxRetain(CUcontext *pctx, CUdevice dev);
+typedef CUresult CUDAAPI tcuDevicePrimaryCtxRelease(CUdevice dev);
+typedef CUresult CUDAAPI tcuDevicePrimaryCtxSetFlags(CUdevice dev, unsigned int flags);
+typedef CUresult CUDAAPI tcuDevicePrimaryCtxGetState(CUdevice dev, unsigned int *flags, int *active);
+typedef CUresult CUDAAPI tcuDevicePrimaryCtxReset(CUdevice dev);
+
+typedef CUresult CUDAAPI tcuStreamCreate(CUstream *phStream, unsigned int flags);
+typedef CUresult CUDAAPI tcuStreamQuery(CUstream hStream);
+typedef CUresult CUDAAPI tcuStreamSynchronize(CUstream hStream);
+typedef CUresult CUDAAPI tcuStreamDestroy_v2(CUstream hStream);
+typedef CUresult CUDAAPI tcuStreamAddCallback(CUstream hStream, CUstreamCallback *callback, void *userdata, unsigned int flags);
+typedef CUresult CUDAAPI tcuEventCreate(CUevent *phEvent, unsigned int flags);
+typedef CUresult CUDAAPI tcuEventDestroy_v2(CUevent hEvent);
+typedef CUresult CUDAAPI tcuEventSynchronize(CUevent hEvent);
+typedef CUresult CUDAAPI tcuEventQuery(CUevent hEvent);
+typedef CUresult CUDAAPI tcuEventRecord(CUevent hEvent, CUstream hStream);
+
+typedef CUresult CUDAAPI tcuLaunchKernel(CUfunction f, unsigned int gridDimX, unsigned int gridDimY, unsigned int gridDimZ, unsigned int blockDimX, unsigned int blockDimY, unsigned int blockDimZ, unsigned int sharedMemBytes, CUstream hStream, void** kernelParams, void** extra);
+typedef CUresult CUDAAPI tcuLinkCreate(unsigned int  numOptions, CUjit_option* options, void** optionValues, CUlinkState* stateOut);
+typedef CUresult CUDAAPI tcuLinkAddData(CUlinkState state, CUjitInputType type, void* data, size_t size, const char* name, unsigned int numOptions, CUjit_option* options, void** optionValues);
+typedef CUresult CUDAAPI tcuLinkComplete(CUlinkState state, void** cubinOut, size_t* sizeOut);
+typedef CUresult CUDAAPI tcuLinkDestroy(CUlinkState state);
+typedef CUresult CUDAAPI tcuModuleLoadData(CUmodule* module, const void* image);
+typedef CUresult CUDAAPI tcuModuleUnload(CUmodule hmod);
+typedef CUresult CUDAAPI tcuModuleGetFunction(CUfunction* hfunc, CUmodule hmod, const char* name);
+typedef CUresult CUDAAPI tcuModuleGetGlobal(CUdeviceptr *dptr, size_t *bytes, CUmodule hmod, const char* name);
+typedef CUresult CUDAAPI tcuTexObjectCreate(CUtexObject* pTexObject, const CUDA_RESOURCE_DESC* pResDesc, const CUDA_TEXTURE_DESC* pTexDesc, const CUDA_RESOURCE_VIEW_DESC* pResViewDesc);
+typedef CUresult CUDAAPI tcuTexObjectDestroy(CUtexObject texObject);
+
+typedef CUresult CUDAAPI tcuGLGetDevices_v2(unsigned int* pCudaDeviceCount, CUdevice* pCudaDevices, unsigned int cudaDeviceCount, CUGLDeviceList deviceList);
+typedef CUresult CUDAAPI tcuGraphicsGLRegisterImage(CUgraphicsResource* pCudaResource, GLuint image, GLenum target, unsigned int Flags);
+typedef CUresult CUDAAPI tcuGraphicsUnregisterResource(CUgraphicsResource resource);
+typedef CUresult CUDAAPI tcuGraphicsMapResources(unsigned int count, CUgraphicsResource* resources, CUstream hStream);
+typedef CUresult CUDAAPI tcuGraphicsUnmapResources(unsigned int count, CUgraphicsResource* resources, CUstream hStream);
+typedef CUresult CUDAAPI tcuGraphicsSubResourceGetMappedArray(CUarray* pArray, CUgraphicsResource resource, unsigned int arrayIndex, unsigned int mipLevel);
+
+typedef CUresult CUDAAPI tcuImportExternalMemory(CUexternalMemory* extMem_out, const CUDA_EXTERNAL_MEMORY_HANDLE_DESC* memHandleDesc);
+typedef CUresult CUDAAPI tcuDestroyExternalMemory(CUexternalMemory extMem);
+typedef CUresult CUDAAPI tcuExternalMemoryGetMappedBuffer(CUdeviceptr* devPtr, CUexternalMemory extMem, const CUDA_EXTERNAL_MEMORY_BUFFER_DESC* bufferDesc);
+typedef CUresult CUDAAPI tcuExternalMemoryGetMappedMipmappedArray(CUmipmappedArray* mipmap, CUexternalMemory extMem, const CUDA_EXTERNAL_MEMORY_MIPMAPPED_ARRAY_DESC* mipmapDesc);
+typedef CUresult CUDAAPI tcuMipmappedArrayGetLevel(CUarray* pLevelArray, CUmipmappedArray hMipmappedArray, unsigned int level);
+typedef CUresult CUDAAPI tcuMipmappedArrayDestroy(CUmipmappedArray hMipmappedArray);
+
+typedef CUresult CUDAAPI tcuImportExternalSemaphore(CUexternalSemaphore* extSem_out, const CUDA_EXTERNAL_SEMAPHORE_HANDLE_DESC* semHandleDesc);
+typedef CUresult CUDAAPI tcuDestroyExternalSemaphore(CUexternalSemaphore extSem);
+typedef CUresult CUDAAPI tcuSignalExternalSemaphoresAsync(const CUexternalSemaphore* extSemArray, const CUDA_EXTERNAL_SEMAPHORE_SIGNAL_PARAMS* paramsArray, unsigned int numExtSems, CUstream stream);
+typedef CUresult CUDAAPI tcuWaitExternalSemaphoresAsync(const CUexternalSemaphore* extSemArray, const CUDA_EXTERNAL_SEMAPHORE_WAIT_PARAMS* paramsArray, unsigned int numExtSems, CUstream stream);
+#endif

common/rfb/dynlink_loader.h

@@ -0,0 +1,340 @@
+/*
+ * This copyright notice applies to this header file only:
+ *
+ * Copyright (c) 2016
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use,
+ * copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the software, and to permit persons to whom the
+ * software is furnished to do so, subject to the following
+ * conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+ * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef FFNV_CUDA_DYNLINK_LOADER_H
+#define FFNV_CUDA_DYNLINK_LOADER_H
+
+#include <stdlib.h>
+
+#include "dynlink_cuda.h"
+#include "nvEncodeAPI.h"
+
+#if defined(_WIN32) && (!defined(FFNV_LOAD_FUNC) || !defined(FFNV_SYM_FUNC) || !defined(FFNV_LIB_HANDLE))
+# include <windows.h>
+#endif
+
+#ifndef FFNV_LIB_HANDLE
+# if defined(_WIN32)
+#  define FFNV_LIB_HANDLE HMODULE
+# else
+#  define FFNV_LIB_HANDLE void*
+# endif
+#endif
+
+#if defined(_WIN32) || defined(__CYGWIN__)
+# define CUDA_LIBNAME "nvcuda.dll"
+# define NVCUVID_LIBNAME "nvcuvid.dll"
+# if defined(_WIN64) || defined(__CYGWIN64__)
+#  define NVENC_LIBNAME "nvEncodeAPI64.dll"
+# else
+#  define NVENC_LIBNAME "nvEncodeAPI.dll"
+# endif
+#else
+# define CUDA_LIBNAME "libcuda.so.1"
+# define NVCUVID_LIBNAME "libnvcuvid.so.1"
+# define NVENC_LIBNAME "libnvidia-encode.so.1"
+#endif
+
+#if !defined(FFNV_LOAD_FUNC) || !defined(FFNV_SYM_FUNC)
+# ifdef _WIN32
+#  define FFNV_LOAD_FUNC(path) LoadLibrary(TEXT(path))
+#  define FFNV_SYM_FUNC(lib, sym) GetProcAddress((lib), (sym))
+#  define FFNV_FREE_FUNC(lib) FreeLibrary(lib)
+# else
+#  include <dlfcn.h>
+#  define FFNV_LOAD_FUNC(path) dlopen((path), RTLD_LAZY)
+#  define FFNV_SYM_FUNC(lib, sym) dlsym((lib), (sym))
+#  define FFNV_FREE_FUNC(lib) dlclose(lib)
+# endif
+#endif
+
+#if !defined(FFNV_LOG_FUNC) || !defined(FFNV_DEBUG_LOG_FUNC)
+# include <stdio.h>
+# define FFNV_LOG_FUNC(logctx, msg, ...) fprintf(stderr, (msg), __VA_ARGS__)
+# define FFNV_DEBUG_LOG_FUNC(logctx, msg, ...)
+#endif
+
+#define LOAD_LIBRARY(l, path)                                  \
+    do {                                                       \
+        if (!((l) = FFNV_LOAD_FUNC(path))) {                   \
+            FFNV_LOG_FUNC(logctx, "Cannot load %s\n", path);   \
+            ret = -1;                                          \
+            goto error;                                        \
+        }                                                      \
+        FFNV_DEBUG_LOG_FUNC(logctx, "Loaded lib: %s\n", path); \
+    } while (0)
+
+#define LOAD_SYMBOL(fun, tp, symbol)                             \
+    do {                                                         \
+        if (!((f->fun) = (tp*)FFNV_SYM_FUNC(f->lib, symbol))) {  \
+            FFNV_LOG_FUNC(logctx, "Cannot load %s\n", symbol);   \
+            ret = -1;                                            \
+            goto error;                                          \
+        }                                                        \
+        FFNV_DEBUG_LOG_FUNC(logctx, "Loaded sym: %s\n", symbol); \
+    } while (0)
+
+#define LOAD_SYMBOL_OPT(fun, tp, symbol)                                      \
+    do {                                                                      \
+        if (!((f->fun) = (tp*)FFNV_SYM_FUNC(f->lib, symbol))) {               \
+            FFNV_DEBUG_LOG_FUNC(logctx, "Cannot load optional %s\n", symbol); \
+        } else {                                                              \
+            FFNV_DEBUG_LOG_FUNC(logctx, "Loaded sym: %s\n", symbol);          \
+        }                                                                     \
+    } while (0)
+
+#define GENERIC_LOAD_FUNC_PREAMBLE(T, n, N)     \
+    T *f;                                       \
+    int ret;                                    \
+                                                \
+    n##_free_functions(functions);              \
+                                                \
+    f = *functions = (T*)calloc(1, sizeof(*f)); \
+    if (!f)                                     \
+        return -1;                              \
+                                                \
+    LOAD_LIBRARY(f->lib, N);
+
+#define GENERIC_LOAD_FUNC_FINALE(n) \
+    return 0;                       \
+error:                              \
+    n##_free_functions(functions);  \
+    return ret;
+
+#define GENERIC_FREE_FUNC()                \
+    if (!functions)                        \
+        return;                            \
+    if (*functions && (*functions)->lib)   \
+        FFNV_FREE_FUNC((*functions)->lib); \
+    free(*functions);                      \
+    *functions = NULL;
+
+#ifdef FFNV_DYNLINK_CUDA_H
+typedef struct CudaFunctions {
+    tcuInit *cuInit;
+    tcuDeviceGetCount *cuDeviceGetCount;
+    tcuDeviceGet *cuDeviceGet;
+    tcuDeviceGetAttribute *cuDeviceGetAttribute;
+    tcuDeviceGetName *cuDeviceGetName;
+    tcuDeviceGetUuid *cuDeviceGetUuid;
+    tcuDeviceComputeCapability *cuDeviceComputeCapability;
+    tcuCtxCreate_v2 *cuCtxCreate;
+    tcuCtxSetLimit *cuCtxSetLimit;
+    tcuCtxPushCurrent_v2 *cuCtxPushCurrent;
+    tcuCtxPopCurrent_v2 *cuCtxPopCurrent;
+    tcuCtxDestroy_v2 *cuCtxDestroy;
+    tcuMemAlloc_v2 *cuMemAlloc;
+    tcuMemAllocPitch_v2 *cuMemAllocPitch;
+    tcuMemsetD8Async *cuMemsetD8Async;
+    tcuMemFree_v2 *cuMemFree;
+    tcuMemcpy *cuMemcpy;
+    tcuMemcpyAsync *cuMemcpyAsync;
+    tcuMemcpy2D_v2 *cuMemcpy2D;
+    tcuMemcpy2DAsync_v2 *cuMemcpy2DAsync;
+    tcuGetErrorName *cuGetErrorName;
+    tcuGetErrorString *cuGetErrorString;
+    tcuCtxGetDevice *cuCtxGetDevice;
+
+    tcuDevicePrimaryCtxRetain *cuDevicePrimaryCtxRetain;
+    tcuDevicePrimaryCtxRelease *cuDevicePrimaryCtxRelease;
+    tcuDevicePrimaryCtxSetFlags *cuDevicePrimaryCtxSetFlags;
+    tcuDevicePrimaryCtxGetState *cuDevicePrimaryCtxGetState;
+    tcuDevicePrimaryCtxReset *cuDevicePrimaryCtxReset;
+
+    tcuStreamCreate *cuStreamCreate;
+    tcuStreamQuery *cuStreamQuery;
+    tcuStreamSynchronize *cuStreamSynchronize;
+    tcuStreamDestroy_v2 *cuStreamDestroy;
+    tcuStreamAddCallback *cuStreamAddCallback;
+    tcuEventCreate *cuEventCreate;
+    tcuEventDestroy_v2 *cuEventDestroy;
+    tcuEventSynchronize *cuEventSynchronize;
+    tcuEventQuery *cuEventQuery;
+    tcuEventRecord *cuEventRecord;
+
+    tcuLaunchKernel *cuLaunchKernel;
+    tcuLinkCreate *cuLinkCreate;
+    tcuLinkAddData *cuLinkAddData;
+    tcuLinkComplete *cuLinkComplete;
+    tcuLinkDestroy *cuLinkDestroy;
+    tcuModuleLoadData *cuModuleLoadData;
+    tcuModuleUnload *cuModuleUnload;
+    tcuModuleGetFunction *cuModuleGetFunction;
+    tcuModuleGetGlobal *cuModuleGetGlobal;
+    tcuTexObjectCreate *cuTexObjectCreate;
+    tcuTexObjectDestroy *cuTexObjectDestroy;
+
+    tcuGLGetDevices_v2 *cuGLGetDevices;
+    tcuGraphicsGLRegisterImage *cuGraphicsGLRegisterImage;
+    tcuGraphicsUnregisterResource *cuGraphicsUnregisterResource;
+    tcuGraphicsMapResources *cuGraphicsMapResources;
+    tcuGraphicsUnmapResources *cuGraphicsUnmapResources;
+    tcuGraphicsSubResourceGetMappedArray *cuGraphicsSubResourceGetMappedArray;
+
+    tcuImportExternalMemory *cuImportExternalMemory;
+    tcuDestroyExternalMemory *cuDestroyExternalMemory;
+    tcuExternalMemoryGetMappedBuffer *cuExternalMemoryGetMappedBuffer;
+    tcuExternalMemoryGetMappedMipmappedArray *cuExternalMemoryGetMappedMipmappedArray;
+    tcuMipmappedArrayDestroy *cuMipmappedArrayDestroy;
+
+    tcuMipmappedArrayGetLevel *cuMipmappedArrayGetLevel;
+
+    tcuImportExternalSemaphore *cuImportExternalSemaphore;
+    tcuDestroyExternalSemaphore *cuDestroyExternalSemaphore;
+    tcuSignalExternalSemaphoresAsync *cuSignalExternalSemaphoresAsync;
+    tcuWaitExternalSemaphoresAsync *cuWaitExternalSemaphoresAsync;
+
+    FFNV_LIB_HANDLE lib;
+} CudaFunctions;
+#else
+typedef struct CudaFunctions CudaFunctions;
+#endif
+
+typedef NVENCSTATUS NVENCAPI tNvEncodeAPICreateInstance(NV_ENCODE_API_FUNCTION_LIST *functionList);
+typedef NVENCSTATUS NVENCAPI tNvEncodeAPIGetMaxSupportedVersion(uint32_t* version);
+
+typedef struct NvencFunctions {
+    tNvEncodeAPICreateInstance *NvEncodeAPICreateInstance;
+    tNvEncodeAPIGetMaxSupportedVersion *NvEncodeAPIGetMaxSupportedVersion;
+
+    FFNV_LIB_HANDLE lib;
+} NvencFunctions;
+
+#ifdef FFNV_DYNLINK_CUDA_H
+static inline void cuda_free_functions(CudaFunctions **functions)
+{
+    GENERIC_FREE_FUNC();
+}
+#endif
+
+static inline void nvenc_free_functions(NvencFunctions **functions)
+{
+    GENERIC_FREE_FUNC();
+}
+
+#ifdef FFNV_DYNLINK_CUDA_H
+static inline int cuda_load_functions(CudaFunctions **functions)
+{
+    GENERIC_LOAD_FUNC_PREAMBLE(CudaFunctions, cuda, CUDA_LIBNAME);
+
+    LOAD_SYMBOL(cuInit, tcuInit, "cuInit");
+    LOAD_SYMBOL(cuDeviceGetCount, tcuDeviceGetCount, "cuDeviceGetCount");
+    LOAD_SYMBOL(cuDeviceGet, tcuDeviceGet, "cuDeviceGet");
+    LOAD_SYMBOL(cuDeviceGetAttribute, tcuDeviceGetAttribute, "cuDeviceGetAttribute");
+    LOAD_SYMBOL(cuDeviceGetName, tcuDeviceGetName, "cuDeviceGetName");
+    LOAD_SYMBOL(cuDeviceComputeCapability, tcuDeviceComputeCapability, "cuDeviceComputeCapability");
+    LOAD_SYMBOL(cuCtxCreate, tcuCtxCreate_v2, "cuCtxCreate_v2");
+    LOAD_SYMBOL(cuCtxSetLimit, tcuCtxSetLimit, "cuCtxSetLimit");
+    LOAD_SYMBOL(cuCtxPushCurrent, tcuCtxPushCurrent_v2, "cuCtxPushCurrent_v2");
+    LOAD_SYMBOL(cuCtxPopCurrent, tcuCtxPopCurrent_v2, "cuCtxPopCurrent_v2");
+    LOAD_SYMBOL(cuCtxDestroy, tcuCtxDestroy_v2, "cuCtxDestroy_v2");
+    LOAD_SYMBOL(cuMemAlloc, tcuMemAlloc_v2, "cuMemAlloc_v2");
+    LOAD_SYMBOL(cuMemAllocPitch, tcuMemAllocPitch_v2, "cuMemAllocPitch_v2");
+    LOAD_SYMBOL(cuMemsetD8Async, tcuMemsetD8Async, "cuMemsetD8Async");
+    LOAD_SYMBOL(cuMemFree, tcuMemFree_v2, "cuMemFree_v2");
+    LOAD_SYMBOL(cuMemcpy, tcuMemcpy, "cuMemcpy");
+    LOAD_SYMBOL(cuMemcpyAsync, tcuMemcpyAsync, "cuMemcpyAsync");
+    LOAD_SYMBOL(cuMemcpy2D, tcuMemcpy2D_v2, "cuMemcpy2D_v2");
+    LOAD_SYMBOL(cuMemcpy2DAsync, tcuMemcpy2DAsync_v2, "cuMemcpy2DAsync_v2");
+    LOAD_SYMBOL(cuGetErrorName, tcuGetErrorName, "cuGetErrorName");
+    LOAD_SYMBOL(cuGetErrorString, tcuGetErrorString, "cuGetErrorString");
+    LOAD_SYMBOL(cuCtxGetDevice, tcuCtxGetDevice, "cuCtxGetDevice");
+
+    LOAD_SYMBOL(cuDevicePrimaryCtxRetain, tcuDevicePrimaryCtxRetain, "cuDevicePrimaryCtxRetain");
+    LOAD_SYMBOL(cuDevicePrimaryCtxRelease, tcuDevicePrimaryCtxRelease, "cuDevicePrimaryCtxRelease");
+    LOAD_SYMBOL(cuDevicePrimaryCtxSetFlags, tcuDevicePrimaryCtxSetFlags, "cuDevicePrimaryCtxSetFlags");
+    LOAD_SYMBOL(cuDevicePrimaryCtxGetState, tcuDevicePrimaryCtxGetState, "cuDevicePrimaryCtxGetState");
+    LOAD_SYMBOL(cuDevicePrimaryCtxReset, tcuDevicePrimaryCtxReset, "cuDevicePrimaryCtxReset");
+
+    LOAD_SYMBOL(cuStreamCreate, tcuStreamCreate, "cuStreamCreate");
+    LOAD_SYMBOL(cuStreamQuery, tcuStreamQuery, "cuStreamQuery");
+    LOAD_SYMBOL(cuStreamSynchronize, tcuStreamSynchronize, "cuStreamSynchronize");
+    LOAD_SYMBOL(cuStreamDestroy, tcuStreamDestroy_v2, "cuStreamDestroy_v2");
+    LOAD_SYMBOL(cuStreamAddCallback, tcuStreamAddCallback, "cuStreamAddCallback");
+    LOAD_SYMBOL(cuEventCreate, tcuEventCreate, "cuEventCreate");
+    LOAD_SYMBOL(cuEventDestroy, tcuEventDestroy_v2, "cuEventDestroy_v2");
+    LOAD_SYMBOL(cuEventSynchronize, tcuEventSynchronize, "cuEventSynchronize");
+    LOAD_SYMBOL(cuEventQuery, tcuEventQuery, "cuEventQuery");
+    LOAD_SYMBOL(cuEventRecord, tcuEventRecord, "cuEventRecord");
+
+    LOAD_SYMBOL(cuLaunchKernel, tcuLaunchKernel, "cuLaunchKernel");
+    LOAD_SYMBOL(cuLinkCreate, tcuLinkCreate, "cuLinkCreate");
+    LOAD_SYMBOL(cuLinkAddData, tcuLinkAddData, "cuLinkAddData");
+    LOAD_SYMBOL(cuLinkComplete, tcuLinkComplete, "cuLinkComplete");
+    LOAD_SYMBOL(cuLinkDestroy, tcuLinkDestroy, "cuLinkDestroy");
+    LOAD_SYMBOL(cuModuleLoadData, tcuModuleLoadData, "cuModuleLoadData");
+    LOAD_SYMBOL(cuModuleUnload, tcuModuleUnload, "cuModuleUnload");
+    LOAD_SYMBOL(cuModuleGetFunction, tcuModuleGetFunction, "cuModuleGetFunction");
+    LOAD_SYMBOL(cuModuleGetGlobal, tcuModuleGetGlobal, "cuModuleGetGlobal");
+    LOAD_SYMBOL(cuTexObjectCreate, tcuTexObjectCreate, "cuTexObjectCreate");
+    LOAD_SYMBOL(cuTexObjectDestroy, tcuTexObjectDestroy, "cuTexObjectDestroy");
+
+    LOAD_SYMBOL(cuGLGetDevices, tcuGLGetDevices_v2, "cuGLGetDevices_v2");
+    LOAD_SYMBOL(cuGraphicsGLRegisterImage, tcuGraphicsGLRegisterImage, "cuGraphicsGLRegisterImage");
+    LOAD_SYMBOL(cuGraphicsUnregisterResource, tcuGraphicsUnregisterResource, "cuGraphicsUnregisterResource");
+    LOAD_SYMBOL(cuGraphicsMapResources, tcuGraphicsMapResources, "cuGraphicsMapResources");
+    LOAD_SYMBOL(cuGraphicsUnmapResources, tcuGraphicsUnmapResources, "cuGraphicsUnmapResources");
+    LOAD_SYMBOL(cuGraphicsSubResourceGetMappedArray, tcuGraphicsSubResourceGetMappedArray, "cuGraphicsSubResourceGetMappedArray");
+
+    LOAD_SYMBOL_OPT(cuDeviceGetUuid, tcuDeviceGetUuid, "cuDeviceGetUuid");
+    LOAD_SYMBOL_OPT(cuImportExternalMemory, tcuImportExternalMemory, "cuImportExternalMemory");
+    LOAD_SYMBOL_OPT(cuDestroyExternalMemory, tcuDestroyExternalMemory, "cuDestroyExternalMemory");
+    LOAD_SYMBOL_OPT(cuExternalMemoryGetMappedBuffer, tcuExternalMemoryGetMappedBuffer, "cuExternalMemoryGetMappedBuffer");
+    LOAD_SYMBOL_OPT(cuExternalMemoryGetMappedMipmappedArray, tcuExternalMemoryGetMappedMipmappedArray, "cuExternalMemoryGetMappedMipmappedArray");
+    LOAD_SYMBOL_OPT(cuMipmappedArrayGetLevel, tcuMipmappedArrayGetLevel, "cuMipmappedArrayGetLevel");
+    LOAD_SYMBOL_OPT(cuMipmappedArrayDestroy, tcuMipmappedArrayDestroy, "cuMipmappedArrayDestroy");
+
+    LOAD_SYMBOL_OPT(cuImportExternalSemaphore, tcuImportExternalSemaphore, "cuImportExternalSemaphore");
+    LOAD_SYMBOL_OPT(cuDestroyExternalSemaphore, tcuDestroyExternalSemaphore, "cuDestroyExternalSemaphore");
+    LOAD_SYMBOL_OPT(cuSignalExternalSemaphoresAsync, tcuSignalExternalSemaphoresAsync, "cuSignalExternalSemaphoresAsync");
+    LOAD_SYMBOL_OPT(cuWaitExternalSemaphoresAsync, tcuWaitExternalSemaphoresAsync, "cuWaitExternalSemaphoresAsync");
+
+    GENERIC_LOAD_FUNC_FINALE(cuda);
+}
+#endif
+
+static inline int nvenc_load_functions(NvencFunctions **functions)
+{
+    GENERIC_LOAD_FUNC_PREAMBLE(NvencFunctions, nvenc, NVENC_LIBNAME);
+
+    LOAD_SYMBOL(NvEncodeAPICreateInstance, tNvEncodeAPICreateInstance, "NvEncodeAPICreateInstance");
+    LOAD_SYMBOL(NvEncodeAPIGetMaxSupportedVersion, tNvEncodeAPIGetMaxSupportedVersion, "NvEncodeAPIGetMaxSupportedVersion");
+
+    GENERIC_LOAD_FUNC_FINALE(nvenc);
+}
+
+#undef GENERIC_LOAD_FUNC_PREAMBLE
+#undef LOAD_LIBRARY
+#undef LOAD_SYMBOL
+#undef GENERIC_LOAD_FUNC_FINALE
+#undef GENERIC_FREE_FUNC
+#undef CUDA_LIBNAME
+#undef NVCUVID_LIBNAME
+#undef NVENC_LIBNAME
+
+#endif
+

common/rfb/mp4.c

@@ -0,0 +1,98 @@
+#include "mp4.h"
+
+#include <string.h>
+
+uint32_t default_sample_size = 40000;
+
+enum BufError create_header();
+
+void set_sps(struct Mp4Context *ctx, const uint8_t* nal_data, const uint32_t nal_len) {
+    memcpy(ctx->buf_sps, nal_data, nal_len);
+    ctx->buf_sps_len = nal_len;
+    create_header(ctx);
+}
+
+void set_pps(struct Mp4Context *ctx, const uint8_t* nal_data, const uint32_t nal_len) {
+    memcpy(ctx->buf_pps, nal_data, nal_len);
+    ctx->buf_pps_len = nal_len;
+    create_header(ctx);
+}
+
+enum BufError create_header(struct Mp4Context *ctx) {
+    if (ctx->buf_header.offset > 0) return BUF_OK;
+    if (ctx->buf_sps_len == 0) return BUF_OK;
+    if (ctx->buf_pps_len == 0) return BUF_OK;
+
+    struct MoovInfo moov_info;
+    memset(&moov_info, 0, sizeof(struct MoovInfo));
+    moov_info.width = ctx->w;
+    moov_info.height = ctx->h;
+    moov_info.horizontal_resolution = 0x00480000; // 72 dpi
+    moov_info.vertical_resolution = 0x00480000; // 72 dpi
+    moov_info.creation_time = 0;
+    moov_info.timescale = default_sample_size * ctx->framerate;
+    moov_info.sps = (uint8_t *) ctx->buf_sps;
+    moov_info.sps_length = ctx->buf_sps_len;
+    moov_info.pps = (uint8_t *) ctx->buf_pps;
+    moov_info.pps_length = ctx->buf_pps_len;
+
+    ctx->buf_header.offset = 0;
+    enum BufError err = write_header(&ctx->buf_header, &moov_info);  chk_err
+    return BUF_OK;
+}
+
+enum BufError get_header(struct Mp4Context *ctx, struct BitBuf *ptr) {
+    ptr->buf =  ctx->buf_header.buf;
+    ptr->size =  ctx->buf_header.size;
+    ptr->offset =  ctx->buf_header.offset;
+    return BUF_OK;
+}
+
+enum BufError set_slice(struct Mp4Context *ctx, const uint8_t* nal_data, const uint32_t origlen, const uint32_t nal_len, const enum NalUnitType unit_type) {
+    enum BufError err;
+
+    const uint32_t samples_info_len = 1;
+    struct SampleInfo samples_info[1];
+    memset(&samples_info[0], 0, sizeof(struct SampleInfo));
+    samples_info[0].size = nal_len + 4; // add size of sample
+    samples_info[0].composition_offset = default_sample_size;
+    samples_info[0].decode_time = default_sample_size;
+    samples_info[0].duration = default_sample_size;
+    samples_info[0].flags = unit_type == NalUnitType_CodedSliceIdr ? 0 : 65536;
+
+    ctx->buf_moof.offset = 0;
+    err = write_moof(&ctx->buf_moof, 0, 0, 0, default_sample_size, samples_info, samples_info_len); chk_err
+
+    ctx->buf_mdat.offset = 0;
+    // printf("nal_data %d\n", nal_data);
+    // printf("      slice: "); for (int i = 0; i < 32; ++i) printf(" 0x%02hhX", nal_data[i]); printf("\n");
+
+    err = write_mdat(&ctx->buf_mdat, nal_data, origlen, nal_len); chk_err
+
+    return BUF_OK;
+}
+
+enum BufError set_mp4_state(struct Mp4Context *ctx, struct Mp4State *state) {
+    enum BufError err = BUF_OK;
+    if (pos_sequence_number > 0) err = put_u32_be_to_offset(&ctx->buf_moof, pos_sequence_number, state->sequence_number); chk_err
+    if (pos_base_data_offset > 0) err = put_u64_be_to_offset(&ctx->buf_moof, pos_base_data_offset, state->base_data_offset); chk_err
+    if (pos_base_media_decode_time > 0) err = put_u64_be_to_offset(&ctx->buf_moof, pos_base_media_decode_time, state->base_media_decode_time); chk_err
+    state->sequence_number++;
+    state->base_data_offset += ctx->buf_moof.offset + ctx->buf_mdat.offset;
+    state->base_media_decode_time += state->default_sample_duration;
+    return BUF_OK;
+}
+enum BufError get_moof(struct Mp4Context *ctx, struct BitBuf *ptr) {
+    ptr->buf =  ctx->buf_moof.buf;
+    ptr->size =  ctx->buf_moof.size;
+    ptr->offset =  ctx->buf_moof.offset;
+    return BUF_OK;
+}
+enum BufError get_mdat(struct Mp4Context *ctx, struct BitBuf *ptr) {
+    ptr->buf =  ctx->buf_mdat.buf;
+    ptr->size =  ctx->buf_mdat.size;
+    ptr->offset =  ctx->buf_mdat.offset;
+    // int o = 4+4;
+    // printf("   get_mdat: "); for (int i = 0; i < 32; ++i) printf(" 0x%02hhX", ptr->buf[o+i]); printf("\n");
+    return BUF_OK;
+}

common/rfb/mp4.h

@@ -0,0 +1,197 @@
+#pragma once
+#include <stdint.h>
+#include <stdbool.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define chk_err  if (err != BUF_OK) { printf("Error buf: %s     %s(...)    %s:%d\n", buf_error_to_str(err), __func__, __FILE__, __LINE__); return err; }
+#define chk_err_continue  if (err != BUF_OK) { printf("Error buf: %s     %s(...)    %s:%d\n", buf_error_to_str(err), __func__, __FILE__, __LINE__); continue; }
+
+enum BufError {
+    BUF_OK = 0,
+    BUF_ENDOFBUF_ERROR,
+    BUF_MALLOC_ERROR,
+    BUF_INCORRECT
+};
+char* buf_error_to_str(const enum BufError err);
+
+struct BitBuf {
+    char *buf;
+    uint32_t size;
+    uint32_t offset;
+};
+
+enum BufError put_skip(struct BitBuf *ptr, const uint32_t count);
+enum BufError put_to_offset(struct BitBuf *ptr, const uint32_t offset, const uint8_t* data, const uint32_t size);
+enum BufError put(struct BitBuf *ptr, const uint8_t* data, const uint32_t size);
+enum BufError put_u8_to_offset(struct BitBuf *ptr, const uint32_t offset, const uint8_t val);
+enum BufError put_u8(struct BitBuf *ptr, uint8_t val);
+enum BufError put_u16_be_to_offset(struct BitBuf *ptr, const uint32_t offset, const uint16_t val);
+enum BufError put_u16_be(struct BitBuf *ptr, const uint16_t val);
+enum BufError put_u16_le_to_offset(struct BitBuf *ptr, const uint32_t offset, const uint16_t val);
+enum BufError put_u16_le(struct BitBuf *ptr, const uint16_t val);
+enum BufError put_u32_be_to_offset(struct BitBuf *ptr, const uint32_t offset, const uint32_t val);
+enum BufError put_u32_be(struct BitBuf *ptr, const uint32_t val);
+enum BufError put_i32_be(struct BitBuf *ptr, const int32_t val);
+enum BufError put_u64_be_to_offset(struct BitBuf *ptr, const uint32_t offset, const uint64_t val);
+enum BufError put_u64_be(struct BitBuf *ptr, const uint64_t val);
+enum BufError put_u32_le_to_offset(struct BitBuf *ptr, const uint32_t offset, const uint32_t val);
+enum BufError put_u32_le(struct BitBuf *ptr, const uint32_t val);
+enum BufError put_str4_to_offset(struct BitBuf *ptr, const uint32_t offset, const char str[4]);
+enum BufError put_str4(struct BitBuf *ptr, const char str[4]);
+enum BufError put_counted_str_to_offset(struct BitBuf *ptr, const uint32_t offset, const char *str, const uint32_t len);
+enum BufError put_counted_str(struct BitBuf *ptr, const char *str, const uint32_t len);
+
+struct MoovInfo {
+    uint8_t profile_idc;
+    uint8_t level_idc;
+    uint8_t *sps;
+    uint16_t sps_length;
+    uint8_t *pps;
+    uint16_t pps_length;
+    uint16_t width;
+    uint16_t height;
+    uint32_t horizontal_resolution;
+    uint32_t vertical_resolution;
+    uint32_t creation_time;
+    uint32_t timescale;
+};
+
+enum BufError write_header(struct BitBuf *ptr, struct MoovInfo *moov_info);
+
+extern uint32_t pos_sequence_number;
+extern uint32_t pos_base_data_offset;
+extern uint32_t pos_base_media_decode_time;
+
+struct SampleInfo {
+    uint32_t duration;
+    uint32_t decode_time;
+    uint32_t composition_time;
+    uint32_t composition_offset;
+    uint32_t size;
+    uint32_t flags;
+};
+
+enum BufError write_mdat(struct BitBuf *ptr, const uint8_t* data, const uint32_t origlen, const uint32_t len);
+enum BufError write_moof(struct BitBuf *ptr,
+    const uint32_t sequence_number,
+    const uint64_t base_data_offset,
+    const uint64_t base_media_decode_time,
+    const uint32_t default_sample_duration,
+    const struct SampleInfo *samples_info, const uint32_t samples_info_len);
+
+enum NalUnitType { //   Table 7-1 NAL unit type codes
+    NalUnitType_Unspecified = 0,                // Unspecified
+    NalUnitType_CodedSliceNonIdr = 1,           // Coded slice of a non-IDR picture
+    NalUnitType_CodedSliceDataPartitionA = 2,   // Coded slice data partition A
+    NalUnitType_CodedSliceDataPartitionB = 3,   // Coded slice data partition B
+    NalUnitType_CodedSliceDataPartitionC = 4,   // Coded slice data partition C
+    NalUnitType_CodedSliceIdr = 5,              // Coded slice of an IDR picture
+    NalUnitType_SEI = 6,                        // Supplemental enhancement information (SEI)
+    NalUnitType_SPS = 7,                        // Sequence parameter set
+    NalUnitType_PPS = 8,                        // Picture parameter set
+    NalUnitType_AUD = 9,                        // Access unit delimiter
+    NalUnitType_EndOfSequence = 10,             // End of sequence
+    NalUnitType_EndOfStream = 11,               // End of stream
+    NalUnitType_Filler = 12,                    // Filler data
+    NalUnitType_SpsExt = 13,                    // Sequence parameter set extension
+    // 14..18           // Reserved
+            NalUnitType_CodedSliceAux = 19,  // Coded slice of an auxiliary coded picture without partitioning
+    // 20..23           // Reserved
+    // 24..31           // Unspecified
+};
+
+struct NAL {
+    char *data;
+    uint64_t data_size;
+    uint32_t picture_order_count;
+
+    // NAL header
+    bool forbidden_zero_bit;
+    uint8_t ref_idc;
+    uint8_t unit_type_value;
+    enum NalUnitType unit_type;
+};
+
+static inline const char* nal_type_to_str(const enum NalUnitType nal_type) {
+    switch (nal_type) {
+        case NalUnitType_Unspecified: return "Unspecified";
+        case NalUnitType_CodedSliceNonIdr: return "CodedSliceNonIdr";
+        case NalUnitType_CodedSliceDataPartitionA: return "CodedSliceDataPartitionA";
+        case NalUnitType_CodedSliceDataPartitionB: return "CodedSliceDataPartitionB";
+        case NalUnitType_CodedSliceDataPartitionC: return "CodedSliceDataPartitionC";
+        case NalUnitType_CodedSliceIdr: return "CodedSliceIdr";
+        case NalUnitType_SEI: return "SEI";
+        case NalUnitType_SPS: return "SPS";
+        case NalUnitType_PPS: return "PPS";
+        case NalUnitType_AUD: return "AUD";
+        case NalUnitType_EndOfSequence: return "EndOfSequence";
+        case NalUnitType_EndOfStream: return "EndOfStream";
+        case NalUnitType_Filler: return "Filler";
+        case NalUnitType_SpsExt: return "SpsExt";
+        case NalUnitType_CodedSliceAux: return "CodedSliceAux";
+        default: return "Unknown";
+    }
+}
+
+static inline void nal_parse_header(struct NAL *nal, const char first_byte) {
+    nal->forbidden_zero_bit = ((first_byte & 0x80) >> 7) == 1;
+    nal->ref_idc = (first_byte & 0x60) >> 5;
+    nal->unit_type = (enum NalUnitType) ((first_byte & 0x1f) >> 0);
+}
+
+static inline bool nal_chk4(const uint8_t *buf, const uint32_t offset) {
+    if (buf[offset] == 0x00 && buf[offset+1] == 0x00 && buf[offset+2] == 0x01) { return true; }
+    if (buf[offset] == 0x00 && buf[offset+1] == 0x00 && buf[offset+2] == 0x00 && buf[offset+3] == 0x01) { return true; }
+    return false;
+}
+
+static inline bool nal_chk3(const uint8_t *buf, const uint32_t offset) {
+    if (buf[offset] == 0x00 && buf[offset+1] == 0x00 && buf[offset+2] == 0x01) { return true; }
+    return false;
+}
+
+extern uint32_t default_sample_size;
+
+struct Mp4State {
+    bool header_sent;
+
+    uint32_t sequence_number;
+    uint64_t base_data_offset;
+    uint64_t base_media_decode_time;
+    uint32_t default_sample_duration;
+
+    uint32_t nals_count;
+};
+
+struct Mp4Context {
+    char buf_sps[128];
+    uint16_t buf_sps_len;
+    char buf_pps[128];
+    uint16_t buf_pps_len;
+
+    uint16_t w, h, framerate;
+
+    struct BitBuf buf_header;
+    struct BitBuf buf_moof;
+    struct BitBuf buf_mdat;
+};
+
+
+enum BufError set_slice(struct Mp4Context *ctx, const uint8_t* nal_data, const uint32_t origlen, const uint32_t nal_len, const enum NalUnitType unit_type);
+void set_sps(struct Mp4Context *ctx, const uint8_t* nal_data, const uint32_t nal_len);
+void set_pps(struct Mp4Context *ctx, const uint8_t* nal_data, const uint32_t nal_len);
+
+enum BufError get_header(struct Mp4Context *ctx, struct BitBuf *ptr);
+
+enum BufError set_mp4_state(struct Mp4Context *ctx, struct Mp4State *state);
+enum BufError get_moof(struct Mp4Context *ctx, struct BitBuf *ptr);
+enum BufError get_mdat(struct Mp4Context *ctx, struct BitBuf *ptr);
+
+#ifdef __cplusplus
+} // extern C
+#endif

common/rfb/mp4_bitbuf.c

@@ -0,0 +1,193 @@
+#include "mp4.h"
+
+#define chk_ptr if (!ptr) return BUF_INCORRECT;
+#define chk_realloc { enum BufError err; err = try_to_realloc(ptr, pos); if (err != BUF_OK) return err; }
+
+char* buf_error_to_str(const enum BufError err) {
+    switch (err) {
+        case BUF_OK: return "BUF_OK";
+        case BUF_ENDOFBUF_ERROR: return "BUF_ENDOFBUF_ERROR";
+        case BUF_MALLOC_ERROR: return "BUF_MALLOC_ERROR";
+        case BUF_INCORRECT: return "BUF_INCORRECT";
+        default: { static char str[32]; sprintf(str, "Unknown(%d)", err); return str; }
+    }
+}
+
+enum BufError try_to_realloc(struct BitBuf *ptr, const uint32_t min_size) {
+    chk_ptr
+    uint32_t new_size = ptr->size + min_size + 1024;
+    char* new_buf = realloc(ptr->buf, new_size);
+    if (new_buf == NULL) return BUF_MALLOC_ERROR;
+    ptr->buf = new_buf;
+    ptr->size = new_size;
+    return BUF_OK;
+}
+
+enum BufError put_skip(struct BitBuf *ptr, const uint32_t count) {
+    chk_ptr
+    uint32_t pos = ptr->offset + count;
+    if (pos >= ptr->size) chk_realloc
+    for (uint32_t i = 0; i < count; i++) ptr->buf[ptr->offset + i] = 0;
+    ptr->offset = pos;
+    return BUF_OK;
+}
+
+enum BufError put_to_offset(struct BitBuf *ptr, const uint32_t offset, const uint8_t* data, const uint32_t size) {
+    chk_ptr
+    uint32_t pos = offset + size;
+    if (pos >= ptr->size) chk_realloc
+    for (uint32_t i = 0; i < size; i++) ptr->buf[offset + i] = data[i];
+    return BUF_OK;
+}
+enum BufError put(struct BitBuf *ptr, const uint8_t* data, const uint32_t size) {
+    chk_ptr
+    enum BufError err = put_to_offset(ptr, ptr->offset, data, size); chk_err
+    ptr->offset += size;
+    return BUF_OK;
+}
+
+enum BufError put_u8_to_offset(struct BitBuf *ptr, const uint32_t offset, const uint8_t val) {
+    chk_ptr
+    uint32_t pos = offset + sizeof(uint8_t);
+    if (pos >= ptr->size) chk_realloc
+    ptr->buf[offset + 0] = val & 0xff;
+    return BUF_OK;
+}
+enum BufError put_u8(struct BitBuf *ptr, uint8_t val) {
+    chk_ptr
+    enum BufError err = put_u8_to_offset(ptr, ptr->offset, val); chk_err
+    ptr->offset += sizeof(uint8_t);
+    return BUF_OK;
+}
+
+enum BufError put_u16_be_to_offset(struct BitBuf *ptr, const uint32_t offset, const uint16_t val) {
+    chk_ptr
+    uint32_t pos = offset + sizeof(uint16_t);
+    if (pos >= ptr->size) chk_realloc
+    ptr->buf[offset + 0] = (val >> 8) & 0xff;
+    ptr->buf[offset + 1] = (val >> 0) & 0xff;
+    return BUF_OK;
+}
+enum BufError put_u16_be(struct BitBuf *ptr, const uint16_t val) {
+    chk_ptr
+    enum BufError err = put_u16_be_to_offset(ptr, ptr->offset, val); chk_err
+    ptr->offset += sizeof(uint16_t);
+    return BUF_OK;
+}
+
+enum BufError put_u16_le_to_offset(struct BitBuf *ptr, const uint32_t offset, const uint16_t val) {
+    chk_ptr
+    uint32_t pos = offset + sizeof(uint16_t);
+    if (pos >= ptr->size) chk_realloc
+    ptr->buf[offset + 0] = (val >> 0) & 0xff;
+    ptr->buf[offset + 1] = (val >> 8) & 0xff;
+    return BUF_OK;
+}
+enum BufError put_u16_le(struct BitBuf *ptr, const uint16_t val) {
+    chk_ptr
+    enum BufError err = put_u16_le_to_offset(ptr, ptr->offset, val); chk_err
+    ptr->offset += sizeof(uint16_t);
+    return BUF_OK;
+}
+
+enum BufError put_u32_be_to_offset(struct BitBuf *ptr, const uint32_t offset, const uint32_t val) {
+    chk_ptr
+    uint32_t pos = offset + sizeof(uint32_t);
+    if (pos >= ptr->size) chk_realloc
+    ptr->buf[offset + 0] = (val >> 24) & 0xff;
+    ptr->buf[offset + 1] = (val >> 16) & 0xff;
+    ptr->buf[offset + 2] = (val >>  8) & 0xff;
+    ptr->buf[offset + 3] = (val >>  0) & 0xff;
+    return BUF_OK;
+}
+enum BufError put_u32_be(struct BitBuf *ptr, const uint32_t val) {
+    chk_ptr
+    enum BufError err = put_u32_be_to_offset(ptr, ptr->offset, val); chk_err
+    ptr->offset += sizeof(uint32_t);
+    return BUF_OK;
+}
+enum BufError put_i32_be(struct BitBuf *ptr, const int32_t val) {
+    chk_ptr
+    enum BufError err = put_u32_be_to_offset(ptr, ptr->offset, val); chk_err
+    ptr->offset += sizeof(int32_t);
+    return BUF_OK;
+}
+
+enum BufError put_u64_be_to_offset(struct BitBuf *ptr, const uint32_t offset, const uint64_t val) {
+    chk_ptr
+    uint32_t pos = offset + sizeof(uint64_t);
+    if (pos > ptr->size) chk_realloc
+    ptr->buf[offset + 0] = (val >> 56) & 0xff;
+    ptr->buf[offset + 1] = (val >> 48) & 0xff;
+    ptr->buf[offset + 2] = (val >> 40) & 0xff;
+    ptr->buf[offset + 3] = (val >> 32) & 0xff;
+    ptr->buf[offset + 4] = (val >> 24) & 0xff;
+    ptr->buf[offset + 5] = (val >> 16) & 0xff;
+    ptr->buf[offset + 6] = (val >>  8) & 0xff;
+    ptr->buf[offset + 7] = (val >>  0) & 0xff;
+    return BUF_OK;
+}
+enum BufError put_u64_be(struct BitBuf *ptr, const uint64_t val) {
+    chk_ptr
+    enum BufError err = put_u64_be_to_offset(ptr, ptr->offset, val); chk_err
+    ptr->offset += sizeof(uint64_t);
+    return BUF_OK;
+}
+
+enum BufError put_u32_le_to_offset(struct BitBuf *ptr, const uint32_t offset, const uint32_t val) {
+    chk_ptr
+    uint32_t pos = offset + 4;
+    if (pos >= ptr->size) chk_realloc
+    ptr->buf[offset + 0] = (val >>  0) & 0xff;
+    ptr->buf[offset + 1] = (val >>  8) & 0xff;
+    ptr->buf[offset + 2] = (val >> 16) & 0xff;
+    ptr->buf[offset + 3] = (val >> 24) & 0xff;
+    return BUF_OK;
+}
+enum BufError put_u32_le(struct BitBuf *ptr, const uint32_t val) {
+    chk_ptr
+    enum BufError err = put_u32_le_to_offset(ptr, ptr->offset, val); chk_err
+    ptr->offset += sizeof(uint32_t);
+    return BUF_OK;
+}
+
+enum BufError put_str4_to_offset(struct BitBuf *ptr, const uint32_t offset, const char str[4]) {
+    chk_ptr
+    uint32_t pos = offset + 4;
+    if (pos >= ptr->size) chk_realloc
+    for (uint8_t i = 0; i < 4; i++) { ptr->buf[offset + i] = str[i]; }
+    return BUF_OK;
+}
+enum BufError put_str4(struct BitBuf *ptr, const char str[4]) {
+    chk_ptr
+    enum BufError err = put_str4_to_offset(ptr, ptr->offset, str); chk_err
+    ptr->offset += 4;
+    return BUF_OK;
+}
+enum BufError put_counted_str_to_offset(struct BitBuf *ptr, const uint32_t offset, const char *str, const uint32_t len) {
+    chk_ptr
+    uint32_t pos = offset + len + 1;
+    if (pos >= ptr->size) chk_realloc
+    for (uint32_t i = 0; i < len+1; i++) { ptr->buf[offset + i] = str[i]; }
+    ptr->buf[pos] = 0;
+    return BUF_OK;
+    }
+enum BufError put_counted_str(struct BitBuf *ptr, const char *str, const uint32_t len) {
+    chk_ptr
+    enum BufError err = put_counted_str_to_offset(ptr, ptr->offset, str, len); chk_err
+    ptr->offset += len+1;
+    return BUF_OK;
+}
+
+// enum BufError hexStr(struct BitBuf *ptr, char *str) : string {
+//     let str = '';
+//     for(let i = 0; i < this.offset; i++) {
+//         if (i % 40 === 0) str += '\n';
+//         if (i % 4 === 0 && i+1 < this.data.length) {
+//             if (i > 0 && i % 40 !== 0) str += ' ';
+//             str += '0x';
+//         }
+//         str += decimalToHex(this.data[i]);
+//     }
+//     return str;
+// }

common/rfb/mp4_moof.c

@@ -0,0 +1,194 @@
+#include "mp4.h"
+#include <string.h>
+
+uint32_t pos_sequence_number = 0;
+uint32_t pos_base_data_offset = 0;
+uint32_t pos_base_media_decode_time = 0;
+
+struct DataOffsetPos {
+    bool data_offset_present;
+    uint32_t offset;
+};
+
+enum BufError write_mfhd(struct BitBuf *ptr, const uint32_t sequence_number);
+enum BufError write_traf(struct BitBuf *ptr,
+                         const uint32_t sequence_number,
+                         const uint64_t base_data_offset,
+                         const uint64_t base_media_decode_time,
+                         const uint32_t default_sample_duration,
+                         const struct SampleInfo *samples_info, const uint32_t samples_info_len,
+                         struct DataOffsetPos *data_offset);
+enum BufError write_tfhd(struct BitBuf *ptr,
+                         const uint32_t sequence_number,
+                         const uint64_t base_data_offset,
+                         const uint64_t base_media_decode_time,
+                         const uint32_t default_sample_size,
+                         const uint32_t default_sample_duration,
+                         const struct SampleInfo *samples_info, const uint32_t samples_info_len,
+                         struct DataOffsetPos *data_offset);
+enum BufError write_tfdt(struct BitBuf *ptr, const uint64_t base_media_decode_time);
+enum BufError write_trun(struct BitBuf *ptr,
+                         const struct SampleInfo *samples_info, const uint32_t samples_info_count,
+                         struct DataOffsetPos *data_offset);
+
+enum BufError write_mdat(struct BitBuf *ptr, const uint8_t* data, const uint32_t origlen, const uint32_t len) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "mdat"); chk_err
+    err = put_u32_be(ptr, origlen); chk_err   // todo
+    // printf(" write_mdat: "); for (int i = 0; i < 32; ++i) printf(" 0x%02hhX", data[i]); printf("\n");
+    err = put(ptr, data, len); chk_err
+    // printf("mdat len %d     ptr->offset %d     start_atom: %d \n", len, ptr->offset, start_atom);
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_moof(struct BitBuf *ptr,
+    const uint32_t sequence_number,
+    const uint64_t base_data_offset,
+    const uint64_t base_media_decode_time,
+    const uint32_t default_sample_duration,
+    const struct SampleInfo *samples_info, const uint32_t samples_info_len)
+{
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "moof"); chk_err
+    err = write_mfhd(ptr, sequence_number); chk_err
+    struct DataOffsetPos data_offset;
+    data_offset.offset = 0;
+    err = write_traf(ptr, sequence_number, base_data_offset, base_media_decode_time, default_sample_duration, samples_info, samples_info_len, &data_offset); chk_err
+    if (data_offset.data_offset_present)
+        err = put_u32_be_to_offset(ptr, data_offset.offset, ptr->offset + 4 /*mdat size*/ + 4 /*mdat id*/); chk_err
+
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_mfhd(struct BitBuf *ptr, const uint32_t sequence_number) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "mfhd"); chk_err
+    err = put_u8(ptr, 0);  // 1 version
+    err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err  // 3 flags
+    pos_sequence_number = ptr->offset;
+    err = put_u32_be(ptr, sequence_number); chk_err  // 4 sequence_number
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_traf(struct BitBuf *ptr,
+    const uint32_t sequence_number,
+    const uint64_t base_data_offset,
+    const uint64_t base_media_decode_time,
+    const uint32_t default_sample_duration,
+    const struct SampleInfo *samples_info, const uint32_t samples_info_len,
+    struct DataOffsetPos *data_offset)
+{
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "traf"); chk_err
+    err = write_tfhd(ptr, sequence_number, base_data_offset, base_media_decode_time, samples_info[0].size, default_sample_duration, samples_info, samples_info_len, data_offset); chk_err
+    err = write_tfdt(ptr, base_media_decode_time); chk_err
+    err = write_trun(ptr, samples_info, samples_info_len, data_offset); chk_err
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_tfhd(struct BitBuf *ptr,
+    const uint32_t sequence_number,
+    const uint64_t base_data_offset,
+    const uint64_t base_media_decode_time,
+    const uint32_t default_sample_size,
+    const uint32_t default_sample_duration,
+    const struct SampleInfo *samples_info, const uint32_t samples_info_len,
+    struct DataOffsetPos *data_offset)
+{
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "tfhd"); chk_err
+
+    err = put_u8(ptr, 0); chk_err  // 1 byte version
+    uint64_t flags = 0x0;
+    const bool base_data_offset_present = false;
+    const bool sample_description_index_present = false;
+    const bool default_sample_duration_present = true;
+    const bool default_sample_size_present = true;
+    const bool default_sample_flags_present = true;
+    const bool duration_is_empty = false;
+    const bool default_base_is_moof = false;
+
+    if (base_data_offset_present)         { flags = flags | 0x000001; } // base-data-offset-present
+    if (sample_description_index_present) { flags = flags | 0x000002; } // sample-description-index-present
+    if (default_sample_duration_present)  { flags = flags | 0x000008; } // default-sample-duration-present
+    if (default_sample_size_present)      { flags = flags | 0x000010; } // default-sample-size-present
+    if (default_sample_flags_present)     { flags = flags | 0x000020; } // default-sample-flags-present
+    if (duration_is_empty)                { flags = flags | 0x010000; } // duration-is-empty
+    if (default_base_is_moof)             { flags = flags | 0x020000; } // default-base-is-moof
+    // buf.put_u8(0); buf.put_u8(0); buf.put_u8(0x39);  // 3 flags
+    // println!("tfhd flags: 0x{:06x}        0x{:02x}: 0x{:02x}: 0x{:02x}", flags, (flags >> 16) as u8, (flags >> 8) as u8, (flags >> 0) as u8);
+    err = put_u8(ptr, flags >> 16); chk_err; err = put_u8(ptr, flags >> 8); chk_err; err = put_u8(ptr, flags >> 0); chk_err // 3 flags
+
+    err = put_u32_be(ptr, 1); chk_err // 4 track_ID
+    if (base_data_offset_present) { pos_base_data_offset = ptr->offset; err = put_u64_be(ptr, base_data_offset); chk_err }
+    // if sample_description_index_present { buf.put_u32_be(0); } // 4 default_sample_description_index
+    if (default_sample_duration_present) { err = put_u32_be(ptr, default_sample_duration); chk_err }
+    if (default_sample_size_present) { err = put_u32_be(ptr, default_sample_size); chk_err }
+    if (default_sample_flags_present) { err = put_u32_be(ptr, 16842752); chk_err }
+
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_tfdt(struct BitBuf *ptr, const uint64_t base_media_decode_time) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "tfdt"); chk_err
+    err = put_u8(ptr, 1);  // 1 version
+    err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err  // 3 flags
+    pos_base_media_decode_time = ptr->offset;
+    err = put_u64_be(ptr, base_media_decode_time); chk_err  // 4 baseMediaDecodeTime
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_trun(struct BitBuf *ptr,
+    const struct SampleInfo *samples_info, const uint32_t samples_info_count,
+    struct DataOffsetPos *data_offset)
+{
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "trun"); chk_err
+
+    err = put_u8(ptr, 0); chk_err  // 1 version
+    const bool data_offset_present = true;
+    const bool first_sample_flags_present = false;
+    const bool sample_duration_present = true;
+    const bool sample_size_present = true;
+    const bool sample_flags_present = true;
+    const bool sample_composition_time_offsets_present = true;
+    {
+        uint64_t flags = 0x0;
+        if (data_offset_present)        { flags = flags | 0x000001; } // 0x000001 data-offset-present.
+        if (first_sample_flags_present) { flags = flags | 0x000004; } // 0x000004 first-sample-flags-present
+        if (sample_duration_present)    { flags = flags | 0x000100; } // 0x000100 sample-duration-present
+        if (sample_size_present)        { flags = flags | 0x000200; } // 0x000200 sample-size-present
+        if (sample_flags_present)       { flags = flags | 0x000400; } // 0x000400 sample-flags-present
+        if (sample_composition_time_offsets_present) { flags = flags | 0x000800; } // 0x000800 sample-composition-time-offsets-present
+        // println!("trup flags: 0x{:06x}        0x{:02x}: 0x{:02x}: 0x{:02x}", flags, (flags >> 16) as u8, (flags >> 8) as u8, (flags >> 0) as u8);
+        err = put_u8(ptr, flags >> 16); chk_err; err = put_u8(ptr, flags >> 8); chk_err; err = put_u8(ptr, flags >> 0); chk_err // 3 flags
+    }
+    err = put_u32_be(ptr, samples_info_count); chk_err  // 4 sample_count
+
+    data_offset->data_offset_present = data_offset_present;
+    data_offset->offset = ptr->offset; // save pointer to this place. we will change size after moof atom will created
+    if (data_offset_present) { err = put_i32_be(ptr, 0); chk_err } // 4 fake data_offset
+
+    if (first_sample_flags_present) { err = put_u32_be(ptr, 33554432); chk_err } // 4 first_sample_flags
+    for (uint32_t i = 0; i < samples_info_count; ++i) {
+        const struct SampleInfo sample_info = samples_info[i];
+        if (sample_duration_present) { err = put_u32_be(ptr, sample_info.duration); chk_err } // 4 sample_duration
+        if (sample_size_present)     { err = put_u32_be(ptr, sample_info.size); chk_err }    // 4 sample_size
+        if (sample_flags_present)    { err = put_u32_be(ptr, sample_info.flags); chk_err }    // 4 sample_flags
+        if (sample_composition_time_offsets_present) {
+//             if version == 0 { err = put_u32_be(ptr, sample_info.composition_offset as u32); chk_err }
+//             else
+            { err = put_i32_be(ptr, sample_info.composition_offset); chk_err }
+        }
+    }
+
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}

common/rfb/mp4_moov.c

@@ -0,0 +1,420 @@
+#include "mp4.h"
+
+#include <string.h>
+
+enum BufError write_ftyp(struct BitBuf *ptr);
+enum BufError write_moov(struct BitBuf *ptr, const struct MoovInfo *moov_info);
+
+enum BufError write_mvhd(struct BitBuf *ptr, const struct MoovInfo *moov_info);
+enum BufError write_trak(struct BitBuf *ptr, const struct MoovInfo *moov_info);
+enum BufError write_tkhd(struct BitBuf *ptr, const struct MoovInfo *moov_info);
+enum BufError write_mdia(struct BitBuf *ptr, const struct MoovInfo *moov_info);
+enum BufError write_mdhd(struct BitBuf *ptr, const struct MoovInfo *moov_info);
+enum BufError write_minf(struct BitBuf *ptr, const struct MoovInfo *moov_info);
+enum BufError write_dinf(struct BitBuf *ptr);
+enum BufError write_dref(struct BitBuf *ptr);
+enum BufError write_url(struct BitBuf *ptr);
+enum BufError write_vmhd(struct BitBuf *ptr);
+enum BufError write_stbl(struct BitBuf *ptr, const struct MoovInfo *moov_info);
+enum BufError write_stsd(struct BitBuf *ptr, const struct MoovInfo *moov_info);
+enum BufError write_avc1(struct BitBuf *ptr, const struct MoovInfo *moov_info);
+enum BufError write_avcC(struct BitBuf *ptr, const struct MoovInfo *moov_info);
+enum BufError write_stts(struct BitBuf *ptr);
+enum BufError write_stsc(struct BitBuf *ptr);
+enum BufError write_stsz(struct BitBuf *ptr);
+enum BufError write_stco(struct BitBuf *ptr);
+enum BufError write_mvex(struct BitBuf *ptr);
+enum BufError write_trex(struct BitBuf *ptr);
+enum BufError write_udta(struct BitBuf *ptr);
+enum BufError write_meta(struct BitBuf *ptr);
+enum BufError write_hdlr(struct BitBuf *ptr, const char name[4], const char manufacturer[4], const char *value, const uint32_t value_len);
+enum BufError write_ilst(struct BitBuf *ptr, const uint8_t *array, const uint32_t len);
+
+enum BufError write_header(struct BitBuf *ptr, struct MoovInfo *moov_info) {
+    enum BufError err;
+    err = write_ftyp(ptr); chk_err
+    err = write_moov(ptr, moov_info); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_ftyp(struct BitBuf *ptr) {
+    enum BufError err;
+    // atom header  <fake size><id>
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "ftyp"); chk_err
+
+    err = put_str4(ptr, "isom"); chk_err          // major_brand
+    err = put_u32_be(ptr, 0x00000200); chk_err    // minor_version
+    err = put_str4(ptr, "isom"); chk_err
+    err = put_str4(ptr, "iso2"); chk_err
+    err = put_str4(ptr, "avc1"); chk_err
+    err = put_str4(ptr, "iso6"); chk_err
+    err = put_str4(ptr, "mp41"); chk_err
+
+    // write atom size
+    uint32_t atom_size = ptr->offset - start_atom;
+    err = put_u32_be_to_offset(ptr, start_atom, atom_size); chk_err
+    return BUF_OK;
+}
+
+
+enum BufError write_moov(struct BitBuf *ptr, const struct MoovInfo *moov_info) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "moov"); chk_err
+    err = write_mvhd(ptr, moov_info); chk_err
+    err = write_trak(ptr, moov_info); chk_err
+    err = write_mvex(ptr); chk_err
+    err = write_udta(ptr); chk_err
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+
+enum BufError write_mvhd(struct BitBuf *ptr, const struct MoovInfo *moov_info) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "mvhd"); chk_err
+
+    err = put_u8(ptr, 0); chk_err  // 1 version
+    err = put_u8(ptr, 0); chk_err; err=put_u8(ptr, 0);chk_err; err=put_u8(ptr, 0);chk_err  // 3 flags
+    err = put_u32_be(ptr, moov_info->creation_time); chk_err // 4 creation_time
+    err = put_u32_be(ptr, 0); chk_err // 4 modification_time
+    err = put_u32_be(ptr, moov_info->timescale); chk_err // 4 timescale
+    err = put_u32_be(ptr, 0); chk_err // 4 duration
+    err = put_u32_be(ptr, 65536); chk_err // 4 preferred rate
+    err = put_u16_le(ptr, 1); chk_err // 2 preferred volume
+    err = put_skip(ptr, 10); chk_err // 10 reserved
+    {   // 36 matrix
+        err = put_u32_be(ptr, 65536); chk_err
+        err = put_u32_be(ptr, 0); chk_err
+        err = put_u32_be(ptr, 0); chk_err
+        err = put_u32_be(ptr, 0); chk_err
+        err = put_u32_be(ptr, 65536); chk_err
+        err = put_u32_be(ptr, 0); chk_err
+        err = put_u32_be(ptr, 0); chk_err
+        err = put_u32_be(ptr, 0); chk_err
+        err = put_u32_be(ptr, 1073741824); chk_err
+    }
+    err = put_u32_be(ptr, 0); chk_err // 4 Preview time
+    err = put_u32_be(ptr, 0); chk_err // 4 Preview duration
+    err = put_u32_be(ptr, 0); chk_err // 4 Poster time
+    err = put_u32_be(ptr, 0); chk_err // 4 Selection time
+    err = put_u32_be(ptr, 0); chk_err // 4 Selection duration
+    err = put_u32_be(ptr, 0); chk_err // 4 Current time
+    err = put_u32_be(ptr, 2); chk_err // 4 Next track ID
+
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_trak(struct BitBuf *ptr, const struct MoovInfo *moov_info) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "trak"); chk_err
+    err = write_tkhd(ptr, moov_info); chk_err
+    err = write_mdia(ptr, moov_info); chk_err
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_tkhd(struct BitBuf *ptr, const struct MoovInfo *moov_info) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "tkhd"); chk_err
+
+    err = put_u8(ptr, 0);  // 1 version
+    err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 3); chk_err  // 3 flags
+    err = put_u32_be(ptr, moov_info->creation_time);  // 4 creation_time
+    err = put_u32_be(ptr, 0);  // 4 modification_time
+    err = put_u32_be(ptr, 1);  // 4 track id
+    err = put_u32_be(ptr, 0);  // 4 reserved
+    err = put_u32_be(ptr, 0);  // 4 duration
+    err = put_skip(ptr, 8);        // 8 reserved
+    err = put_u16_be(ptr, 0);  // 2 layer
+    err = put_u16_be(ptr, 0);  // 2 Alternate group
+    err = put_u16_be(ptr, 0);  // 2 Volume
+    err = put_u16_be(ptr, 0);  // 2 Reserved
+    {   // 36 Matrix structure
+        err = put_u32_be(ptr, 65536); chk_err
+        err = put_u32_be(ptr, 0); chk_err
+        err = put_u32_be(ptr, 0); chk_err
+        err = put_u32_be(ptr, 0); chk_err
+        err = put_u32_be(ptr, 65536); chk_err
+        err = put_u32_be(ptr, 0); chk_err
+        err = put_u32_be(ptr, 0); chk_err
+        err = put_u32_be(ptr, 0); chk_err
+        err = put_u32_be(ptr, 1073741824); chk_err
+    }
+    err = put_u32_be(ptr, moov_info->width * 65536); chk_err  // 4 Track width
+    err = put_u32_be(ptr, moov_info->height * 65536); chk_err  // 4 Track height
+
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+
+enum BufError write_mdia(struct BitBuf *ptr, const struct MoovInfo *moov_info) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "mdia"); chk_err
+    err = write_mdhd(ptr, moov_info); chk_err
+    char *str = "VideoHandler";
+    err = write_hdlr(ptr, "vide", "\0\0\0\0", str, strlen(str)); chk_err
+    err = write_minf(ptr, moov_info); chk_err
+
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_mdhd(struct BitBuf *ptr, const struct MoovInfo *moov_info) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "mdhd"); chk_err
+    err = put_u8(ptr, 0);  // 1 version
+    err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err  // 3 flags
+    err = put_u32_be(ptr, 0); chk_err  // 4 creation_time
+    err = put_u32_be(ptr, 0); chk_err  // 4 modification_time
+    err = put_u32_be(ptr, moov_info->timescale); chk_err  // 4 timescale
+    err = put_u32_be(ptr, 0); chk_err  // 4 duration
+    err = put_u16_be(ptr, 21956); chk_err  // 2 language
+    err = put_u16_be(ptr, 0); chk_err  // 2 quality
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_minf(struct BitBuf *ptr, const struct MoovInfo *moov_info) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "minf"); chk_err
+    err = write_vmhd(ptr); chk_err
+    err = write_dinf(ptr); chk_err
+    err = write_stbl(ptr, moov_info); chk_err
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_dinf(struct BitBuf *ptr) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "dinf"); chk_err
+    err = write_dref(ptr); chk_err
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_dref(struct BitBuf *ptr) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "dref"); chk_err
+    err = put_u8(ptr, 0); chk_err  // 1 version
+    err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err  // 3 flags
+    err = put_u32_be(ptr, 1); chk_err // 4 Component flags mask
+    err = write_url(ptr); chk_err
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_url(struct BitBuf *ptr) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "url "); chk_err
+    err = put_u8(ptr, 0); chk_err  // 1 version
+    err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 1); chk_err  // 3 flags
+    //err = put_u8(ptr, 0); chk_err // <counted string> end
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_vmhd(struct BitBuf *ptr) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "vmhd"); chk_err
+
+    err = put_u8(ptr, 0); chk_err  // 1 version
+    err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 1); chk_err  // 3 flags
+    err = put_u16_be(ptr, 0); chk_err  // 2 Graphics mode
+    err = put_u16_be(ptr, 0); chk_err  // 2 Opcolor
+    err = put_u16_be(ptr, 0); chk_err  // 2 Opcolor
+    err = put_u16_be(ptr, 0); chk_err  // 2 Opcolor
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_stbl(struct BitBuf *ptr, const struct MoovInfo *moov_info) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "stbl"); chk_err
+    err = write_stsd(ptr, moov_info); chk_err
+    err = write_stts(ptr); chk_err
+    err = write_stsc(ptr); chk_err
+    err = write_stsz(ptr); chk_err
+    err = write_stco(ptr); chk_err
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_stsd(struct BitBuf *ptr, const struct MoovInfo *moov_info) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "stsd"); chk_err
+    err = put_u8(ptr, 0); chk_err  // 1 version
+    err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err  // 3 flags
+    err = put_u32_be(ptr, 1); chk_err // 4  Number of entries
+    err = write_avc1(ptr, moov_info); chk_err
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_avc1(struct BitBuf *ptr, const struct MoovInfo *moov_info) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "avc1"); chk_err
+
+    err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err  // reserved
+    err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err  // reserved
+    err = put_u16_be(ptr, 1); chk_err // data_reference_index
+    err = put_u16_be(ptr, 0); chk_err // pre_defined
+    err = put_u16_be(ptr, 0); chk_err // reserved
+    err = put_u32_be(ptr, 0); chk_err
+    err = put_u32_be(ptr, 0); chk_err
+    err = put_u32_be(ptr, 0); chk_err // pre_defined
+    err = put_u16_be(ptr, moov_info->width); chk_err // 2 width
+    err = put_u16_be(ptr, moov_info->height); chk_err // 2 height
+    err = put_u32_be(ptr, moov_info->horizontal_resolution); chk_err // 4 horizontal_resolution
+    err = put_u32_be(ptr, moov_info->vertical_resolution); chk_err // 4 vertical_resolution
+    err = put_u32_be(ptr, 0); chk_err // reserved
+    err = put_u16_be(ptr, 1); chk_err // 2 frame_count
+    err = put_u8(ptr, 0); chk_err
+//    uint8_t *compressorname = { 0, 0, 0, 0, // dailymotion/hls.js
+//                                0, 0, 0, 0,
+//                                0, 0, 0, 0,
+//                                0, 0, 0, 0,
+//                                0, 0, 0, 0,
+//                                0, 0, 0, 0,
+//                                0, 0, 0, 0,
+//                                0, 0, 0 };
+    char compressorname[50] = "OpenIPC project                    ";
+
+    err = put(ptr, (uint8_t *) compressorname, 31); chk_err // compressorname
+    err = put_u16_be(ptr, 24); chk_err // 2 depth
+    err = put_u16_be(ptr, 0xffff); chk_err // 2 color_table_id
+    err = write_avcC(ptr, moov_info); chk_err
+
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+
+enum BufError write_avcC(struct BitBuf *ptr, const struct MoovInfo *moov_info) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "avcC"); chk_err
+
+    err = put_u8(ptr, 1); chk_err  // 1 version
+    err = put_u8(ptr, moov_info->sps[1]); chk_err  // 1 profile
+    err = put_u8(ptr, moov_info->sps[2]); chk_err  // 1 compatibility
+    err = put_u8(ptr, moov_info->sps[3]); chk_err  // 1 level
+    err = put_u8(ptr, 0xFF); chk_err  // 6 bits reserved (111111) + 2 bits nal size length - 1 (11)
+    err = put_u8(ptr, 0xE1); chk_err  // 3 bits reserved (111) + 5 bits number of sps (00001)
+    err = put_u16_be(ptr, moov_info->sps_length); chk_err
+    err = put(ptr, (const uint8_t *) moov_info->sps, moov_info->sps_length); chk_err // SPS
+    err = put_u8(ptr, 1); chk_err  // 1 num pps
+    err = put_u16_be(ptr, moov_info->pps_length); chk_err
+    err = put(ptr, (const uint8_t *)moov_info->pps, moov_info->pps_length); chk_err // pps
+
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_stts(struct BitBuf *ptr) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "stts"); chk_err
+
+    err = put_u8(ptr, 0); chk_err  // 1 version
+    err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err  // 3 flags
+    err = put_u32_be(ptr, 0); chk_err // Number of entries
+    // Time-to-sample table
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_stsc(struct BitBuf *ptr) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "stsc"); chk_err
+    err = put_u8(ptr, 0); chk_err  // 1 version
+    err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err  // 3 flags
+    err = put_u32_be(ptr, 0); chk_err // Number of entries
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_stsz(struct BitBuf *ptr) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "stsz"); chk_err
+    err = put_u8(ptr, 0); chk_err  // 1 version
+    err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err  // 3 flags
+    err = put_u32_be(ptr, 0); chk_err // Sample size
+    err = put_u32_be(ptr, 0); chk_err // Number of entries
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_stco(struct BitBuf *ptr) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "stco"); chk_err
+    err = put_u8(ptr, 0); chk_err  // 1 version
+    err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err  // 3 flags
+    err = put_u32_be(ptr, 0); chk_err // Number of entries
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_mvex(struct BitBuf *ptr) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "mvex"); chk_err
+    err = write_trex(ptr); chk_err
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_trex(struct BitBuf *ptr) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "trex"); chk_err
+    err = put_u8(ptr, 0); chk_err  // 1 version
+    err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err  // 3 flags
+    err = put_u32_be(ptr, 1); chk_err // track_ID
+    err = put_u32_be(ptr, 1); chk_err // default_sample_description_index
+    err = put_u32_be(ptr, 0); chk_err // default_sample_duration
+    err = put_u32_be(ptr, 0); chk_err // default_sample_size
+    err = put_u32_be(ptr, 0); chk_err // default_sample_flags
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_udta(struct BitBuf *ptr) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "udta"); chk_err
+    err = write_meta(ptr); chk_err
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_meta(struct BitBuf *ptr) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "meta"); chk_err
+    err = put_u8(ptr, 0); chk_err  // 1 version
+    err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err  // 3 flags
+    err = write_hdlr(ptr, "mdir", "appl", "", 0); chk_err
+    uint8_t array[37] = {0,0,0,37,169,116,111,111,0,0,0,29,100,97,116,97,0,0,0,1,0,0,0,0,76,97,118,102,53,55,46,56,51,46,49,48,48};
+    err = write_ilst(ptr, array, 37); chk_err
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_hdlr(struct BitBuf *ptr, const char name[4], const char manufacturer[4], const char *value, const uint32_t value_len) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "hdlr"); chk_err
+    err = put_u8(ptr, 0); chk_err  // 1 version
+    err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err; err = put_u8(ptr, 0); chk_err  // 3 flags
+    err = put_u32_be(ptr, 0); chk_err // 4 Predefined
+    err = put_str4(ptr, name); chk_err // 4 Component subtype
+    err = put_str4(ptr, manufacturer); chk_err // 4 Component manufacturer
+    err = put_u32_be(ptr, 0); chk_err // 4 Component flags
+    err = put_u32_be(ptr, 0); chk_err // 4 Component flags mask
+    err = put_counted_str(ptr, value, value_len); chk_err // <counted string> Component name
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}
+
+enum BufError write_ilst(struct BitBuf *ptr, const uint8_t *array, const uint32_t len) {
+    enum BufError err;
+    uint32_t start_atom = ptr->offset; err = put_u32_be(ptr, 0); chk_err; err = put_str4(ptr, "ilst"); chk_err
+    err = put(ptr, array, len); chk_err // <counted string> Component name
+    err = put_u32_be_to_offset(ptr, start_atom, ptr->offset - start_atom); chk_err
+    return BUF_OK;
+}

common/rfb/nvidia.cxx

@@ -0,0 +1,453 @@
+/*
+ * H.264/HEVC hardware encoding using nvidia nvenc
+ * Copyright (c) 2016 Timo Rothenpieler <timo@rothenpieler.org>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <rfb/LogWriter.h>
+
+#include <dlfcn.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "nvidia.h"
+
+using namespace rfb;
+
+static LogWriter vlog("nvidia");
+
+#define FFNV_LOG_FUNC(logctx, msg, ...) vlog.info((msg), __VA_ARGS__)
+#define FFNV_DEBUG_LOG_FUNC(logctx, msg, ...)
+
+#include "dynlink_loader.h"
+
+#define NUM_SURF 4
+
+typedef struct NvencSurface
+{
+	NV_ENC_INPUT_PTR input_surface;
+	int reg_idx;
+	int width;
+	int height;
+	int pitch;
+
+	NV_ENC_OUTPUT_PTR output_surface;
+	NV_ENC_BUFFER_FORMAT format;
+} NvencSurface;
+
+typedef struct NvencDynLoadFunctions
+{
+	CudaFunctions *cuda_dl;
+	NvencFunctions *nvenc_dl;
+
+	void *nvenc_ctx;
+	NV_ENCODE_API_FUNCTION_LIST nvenc_funcs;
+
+	NV_ENC_INITIALIZE_PARAMS init_enc_parms;
+	NV_ENC_CONFIG enc_cfg;
+	CUdevice cu_dev;
+	CUcontext cu_ctx;
+
+	NvencSurface surf[NUM_SURF];
+	uint8_t cursurf;
+} NvencDynLoadFunctions;
+
+static NvencDynLoadFunctions nvenc;
+
+/*
+Recommended settings for streaming
+Low-Latency High Quality preset
+Rate control mode = Two-pass CBR
+Very low VBV buffer size (Single frame)
+No B Frames
+Infinite GOP length
+Adaptive Quantization enabled
+*/
+
+static int loadfuncs() {
+	int ret;
+	NVENCSTATUS err;
+	uint32_t nvenc_max_ver;
+
+	ret = cuda_load_functions(&nvenc.cuda_dl);
+	if (ret < 0)
+		return ret;
+
+	ret = nvenc_load_functions(&nvenc.nvenc_dl);
+	if (ret < 0)
+		return ret;
+
+	err = nvenc.nvenc_dl->NvEncodeAPIGetMaxSupportedVersion(&nvenc_max_ver);
+	if (err != NV_ENC_SUCCESS)
+		return -1;
+
+	vlog.info("Loaded nvenc version %u.%u", nvenc_max_ver >> 4, nvenc_max_ver & 0xf);
+
+	if ((NVENCAPI_MAJOR_VERSION << 4 | NVENCAPI_MINOR_VERSION) > nvenc_max_ver) {
+		vlog.error("Your Nvidia driver is too old. Nvenc %u.%u required",
+			NVENCAPI_MAJOR_VERSION, NVENCAPI_MINOR_VERSION);
+		return -1;
+	}
+
+	nvenc.nvenc_funcs.version = NV_ENCODE_API_FUNCTION_LIST_VER;
+
+	err = nvenc.nvenc_dl->NvEncodeAPICreateInstance(&nvenc.nvenc_funcs);
+	if (err != NV_ENC_SUCCESS)
+		return -1;
+	return 0;
+}
+
+static int nvenc_check_cap(NV_ENC_CAPS cap) {
+	NV_ENC_CAPS_PARAM params;
+	memset(&params, 0, sizeof(NV_ENC_CAPS_PARAM));
+
+	params.version = NV_ENC_CAPS_PARAM_VER;
+	params.capsToQuery = cap;
+
+	int ret, val = 0;
+
+	ret = nvenc.nvenc_funcs.nvEncGetEncodeCaps(nvenc.nvenc_ctx,
+						nvenc.init_enc_parms.encodeGUID,
+						&params, &val);
+	if (ret == NV_ENC_SUCCESS)
+		return val;
+	return 0;
+}
+
+static int setupdevice() {
+	int ret;
+
+	nvenc.init_enc_parms.encodeGUID = NV_ENC_CODEC_H264_GUID;
+	nvenc.init_enc_parms.presetGUID = NV_ENC_PRESET_P7_GUID;
+
+	ret = nvenc.cuda_dl->cuInit(0);
+	if (ret < 0)
+		return ret;
+
+	ret = nvenc.cuda_dl->cuDeviceGet(&nvenc.cu_dev, 0);
+	if (ret < 0)
+		return ret;
+
+	ret = nvenc.cuda_dl->cuCtxCreate(&nvenc.cu_ctx, CU_CTX_SCHED_BLOCKING_SYNC,
+					nvenc.cu_dev);
+	if (ret < 0)
+		return ret;
+
+	CUcontext dummy;
+	nvenc.cuda_dl->cuCtxPopCurrent(&dummy);
+
+	// cuda stream is NULL to use the default
+
+	// open session
+	NV_ENC_OPEN_ENCODE_SESSION_EX_PARAMS params;
+	memset(&params, 0, sizeof(NV_ENC_OPEN_ENCODE_SESSION_EX_PARAMS));
+	NVENCSTATUS err;
+
+	params.version = NV_ENC_OPEN_ENCODE_SESSION_EX_PARAMS_VER;
+	params.apiVersion = NVENCAPI_VERSION;
+	params.device = nvenc.cu_ctx;
+	params.deviceType = NV_ENC_DEVICE_TYPE_CUDA;
+
+	err = nvenc.nvenc_funcs.nvEncOpenEncodeSessionEx(&params, &nvenc.nvenc_ctx);
+	if (err != NV_ENC_SUCCESS)
+		return -1;
+
+	// check caps
+	const int maxw = nvenc_check_cap(NV_ENC_CAPS_WIDTH_MAX);
+	const int maxh = nvenc_check_cap(NV_ENC_CAPS_HEIGHT_MAX);
+	const int minw = nvenc_check_cap(NV_ENC_CAPS_WIDTH_MIN);
+	const int minh = nvenc_check_cap(NV_ENC_CAPS_HEIGHT_MIN);
+
+	vlog.info("Max enc resolution %ux%u, min %ux%u", maxw, maxh, minw, minh);
+
+	return 0;
+}
+
+static int setupenc(const unsigned w, const unsigned h, const unsigned kbps,
+			const unsigned fps) {
+	NVENCSTATUS err;
+
+	nvenc.enc_cfg.version = NV_ENC_CONFIG_VER;
+	nvenc.init_enc_parms.version = NV_ENC_INITIALIZE_PARAMS_VER;
+	nvenc.init_enc_parms.darWidth =
+	nvenc.init_enc_parms.encodeWidth = w;
+	nvenc.init_enc_parms.darHeight =
+	nvenc.init_enc_parms.encodeHeight = h;
+
+	nvenc.init_enc_parms.frameRateNum = fps;
+	nvenc.init_enc_parms.frameRateDen = 1;
+
+	nvenc.init_enc_parms.encodeConfig = &nvenc.enc_cfg;
+	nvenc.init_enc_parms.tuningInfo = NV_ENC_TUNING_INFO_LOW_LATENCY;
+
+	NV_ENC_PRESET_CONFIG preset_cfg;
+	memset(&preset_cfg, 0, sizeof(NV_ENC_PRESET_CONFIG));
+
+	preset_cfg.version = NV_ENC_PRESET_CONFIG_VER;
+	preset_cfg.presetCfg.version = NV_ENC_CONFIG_VER;
+
+	err = nvenc.nvenc_funcs.nvEncGetEncodePresetConfigEx(nvenc.nvenc_ctx,
+			nvenc.init_enc_parms.encodeGUID,
+			nvenc.init_enc_parms.presetGUID,
+			nvenc.init_enc_parms.tuningInfo,
+			&preset_cfg);
+	if (err != NV_ENC_SUCCESS)
+		return -1;
+
+	memcpy(&nvenc.enc_cfg, &preset_cfg.presetCfg, sizeof(nvenc.enc_cfg));
+
+	nvenc.enc_cfg.version = NV_ENC_CONFIG_VER;
+
+	nvenc.init_enc_parms.enableEncodeAsync = 0;
+	nvenc.init_enc_parms.enablePTD = 1;
+
+	nvenc.enc_cfg.frameIntervalP = 0;
+	nvenc.enc_cfg.gopLength = 1;
+
+	// use 4 surfaces
+
+	// setup rate control
+	nvenc.enc_cfg.rcParams.multiPass = NV_ENC_TWO_PASS_FULL_RESOLUTION;
+	nvenc.enc_cfg.rcParams.averageBitRate = kbps * 1024;
+	nvenc.enc_cfg.rcParams.rateControlMode = NV_ENC_PARAMS_RC_CBR;
+	nvenc.enc_cfg.rcParams.lowDelayKeyFrameScale = 1;
+
+	nvenc.enc_cfg.rcParams.enableAQ = 1;
+	nvenc.enc_cfg.rcParams.aqStrength = 4; // 1 - 15, 0 would be auto
+
+	nvenc.enc_cfg.frameFieldMode = NV_ENC_PARAMS_FRAME_FIELD_MODE_FRAME;
+
+	// setup_codec_config
+	nvenc.enc_cfg.encodeCodecConfig.h264Config.h264VUIParameters.videoFullRangeFlag = 1;
+	nvenc.enc_cfg.encodeCodecConfig.h264Config.outputBufferingPeriodSEI = 1;
+	nvenc.enc_cfg.encodeCodecConfig.h264Config.adaptiveTransformMode = NV_ENC_H264_ADAPTIVE_TRANSFORM_ENABLE;
+	nvenc.enc_cfg.encodeCodecConfig.h264Config.fmoMode = NV_ENC_H264_FMO_DISABLE;
+	nvenc.enc_cfg.profileGUID = NV_ENC_H264_PROFILE_MAIN_GUID;
+
+	nvenc.cuda_dl->cuCtxPushCurrent(nvenc.cu_ctx);
+
+	err = nvenc.nvenc_funcs.nvEncInitializeEncoder(nvenc.nvenc_ctx,
+			&nvenc.init_enc_parms);
+	if (err != NV_ENC_SUCCESS)
+		return -1;
+
+	// custream?
+
+	CUcontext dummy;
+	nvenc.cuda_dl->cuCtxPopCurrent(&dummy);
+
+	return 0;
+}
+
+static int setupsurf(const unsigned w, const unsigned h) {
+
+	nvenc.cuda_dl->cuCtxPushCurrent(nvenc.cu_ctx);
+
+	int i;
+	for (i = 0; i < NUM_SURF; i++) {
+		NVENCSTATUS err;
+		NV_ENC_CREATE_BITSTREAM_BUFFER allocOut;
+		memset(&allocOut, 0, sizeof(NV_ENC_CREATE_BITSTREAM_BUFFER));
+		allocOut.version = NV_ENC_CREATE_BITSTREAM_BUFFER_VER;
+
+		NV_ENC_CREATE_INPUT_BUFFER allocSurf;
+		memset(&allocSurf, 0, sizeof(NV_ENC_CREATE_INPUT_BUFFER));
+
+		nvenc.surf[i].format = NV_ENC_BUFFER_FORMAT_ARGB; // doesn't have RGBA!
+		allocSurf.version = NV_ENC_CREATE_INPUT_BUFFER_VER;
+		allocSurf.width = w;
+		allocSurf.height = h;
+		allocSurf.bufferFmt = nvenc.surf[i].format;
+
+		err = nvenc.nvenc_funcs.nvEncCreateInputBuffer(nvenc.nvenc_ctx, &allocSurf);
+		if (err != NV_ENC_SUCCESS)
+			return -1;
+
+		nvenc.surf[i].input_surface = allocSurf.inputBuffer;
+		nvenc.surf[i].width = allocSurf.width;
+		nvenc.surf[i].height = allocSurf.height;
+
+		// output
+		err = nvenc.nvenc_funcs.nvEncCreateBitstreamBuffer(nvenc.nvenc_ctx, &allocOut);
+		if (err != NV_ENC_SUCCESS)
+			return -1;
+
+		nvenc.surf[i].output_surface = allocOut.bitstreamBuffer;
+	}
+
+	CUcontext dummy;
+	nvenc.cuda_dl->cuCtxPopCurrent(&dummy);
+
+	return 0;
+}
+
+int nvenc_frame(const uint8_t *data, unsigned pts, uint8_t *out, uint32_t &outlen) {
+	NVENCSTATUS err;
+
+	NV_ENC_PIC_PARAMS params;
+	memset(&params, 0, sizeof(NV_ENC_PIC_PARAMS));
+	params.version = NV_ENC_PIC_PARAMS_VER;
+	params.encodePicFlags = NV_ENC_PIC_FLAG_FORCEINTRA | NV_ENC_PIC_FLAG_OUTPUT_SPSPPS;
+
+	nvenc.cuda_dl->cuCtxPushCurrent(nvenc.cu_ctx);
+
+	NV_ENC_LOCK_INPUT_BUFFER lockBufferParams;
+	memset(&lockBufferParams, 0, sizeof(NV_ENC_LOCK_INPUT_BUFFER));
+	lockBufferParams.version = NV_ENC_LOCK_INPUT_BUFFER_VER;
+	lockBufferParams.inputBuffer = nvenc.surf[nvenc.cursurf].input_surface;
+
+	err = nvenc.nvenc_funcs.nvEncLockInputBuffer(nvenc.nvenc_ctx, &lockBufferParams);
+	if (err != NV_ENC_SUCCESS)
+		return -1;
+
+	nvenc.surf[nvenc.cursurf].pitch = lockBufferParams.pitch;
+	//vlog.info("pitch %u", lockBufferParams.pitch);
+
+	// copy frame
+	unsigned y;
+	uint8_t *dst = (uint8_t *) lockBufferParams.bufferDataPtr;
+	const unsigned linelen = nvenc.surf[nvenc.cursurf].width * 4;
+	for (y = 0; y < (unsigned) nvenc.surf[nvenc.cursurf].height; y++) {
+		memcpy(dst, data, linelen);
+		data += linelen;
+		dst += lockBufferParams.pitch;
+	}
+
+	err = nvenc.nvenc_funcs.nvEncUnlockInputBuffer(nvenc.nvenc_ctx,
+				nvenc.surf[nvenc.cursurf].input_surface);
+	if (err != NV_ENC_SUCCESS)
+		return -1;
+
+	CUcontext dummy;
+	nvenc.cuda_dl->cuCtxPopCurrent(&dummy);
+
+	params.inputBuffer = nvenc.surf[nvenc.cursurf].input_surface;
+	params.bufferFmt = nvenc.surf[nvenc.cursurf].format;
+	params.inputWidth = nvenc.surf[nvenc.cursurf].width;
+	params.inputHeight = nvenc.surf[nvenc.cursurf].height;
+	params.inputPitch = nvenc.surf[nvenc.cursurf].pitch;
+	params.outputBitstream = nvenc.surf[nvenc.cursurf].output_surface;
+	params.pictureStruct = NV_ENC_PIC_STRUCT_FRAME;
+	params.inputTimeStamp = pts;
+
+	nvenc.cuda_dl->cuCtxPushCurrent(nvenc.cu_ctx);
+
+	err = nvenc.nvenc_funcs.nvEncEncodePicture(nvenc.nvenc_ctx, &params);
+
+	nvenc.cuda_dl->cuCtxPopCurrent(&dummy);
+
+	if (err != NV_ENC_SUCCESS)
+		return -1;
+
+
+	nvenc.cuda_dl->cuCtxPushCurrent(nvenc.cu_ctx);
+
+	// Get output
+	NV_ENC_LOCK_BITSTREAM lock_params;
+	memset(&lock_params, 0, sizeof(NV_ENC_LOCK_BITSTREAM));
+
+	lock_params.version = NV_ENC_LOCK_BITSTREAM_VER;
+	lock_params.doNotWait = 0;
+	lock_params.outputBitstream = nvenc.surf[nvenc.cursurf].output_surface;
+	// lock_params.sliceOffsets = slice_offsets; TODO?
+
+	err = nvenc.nvenc_funcs.nvEncLockBitstream(nvenc.nvenc_ctx, &lock_params);
+	if (err != NV_ENC_SUCCESS)
+		return -1;
+
+	memcpy(out, lock_params.bitstreamBufferPtr, lock_params.bitstreamSizeInBytes);
+	outlen = lock_params.bitstreamSizeInBytes;
+
+	err = nvenc.nvenc_funcs.nvEncUnlockBitstream(nvenc.nvenc_ctx,
+				nvenc.surf[nvenc.cursurf].output_surface);
+	if (err != NV_ENC_SUCCESS)
+		return -1;
+
+	nvenc.cuda_dl->cuCtxPopCurrent(&dummy);
+
+	//vlog.info("Pic type %x, idr %x i %x", lock_params.pictureType, NV_ENC_PIC_TYPE_IDR,
+	//	NV_ENC_PIC_TYPE_I);
+
+	return 0;
+}
+
+void nvidia_unload() {
+	NV_ENC_PIC_PARAMS params;
+	memset(&params, 0, sizeof(NV_ENC_PIC_PARAMS));
+	params.version = NV_ENC_PIC_PARAMS_VER;
+	params.encodePicFlags = NV_ENC_PIC_FLAG_EOS;
+
+	nvenc.cuda_dl->cuCtxPushCurrent(nvenc.cu_ctx);
+
+	nvenc.nvenc_funcs.nvEncEncodePicture(nvenc.nvenc_ctx, &params);
+
+	int i;
+	for (i = 0; i < NUM_SURF; i++) {
+		nvenc.nvenc_funcs.nvEncDestroyInputBuffer(nvenc.nvenc_ctx,
+					nvenc.surf[i].input_surface);
+		nvenc.nvenc_funcs.nvEncDestroyBitstreamBuffer(nvenc.nvenc_ctx,
+					nvenc.surf[i].output_surface);
+	}
+
+	nvenc.nvenc_funcs.nvEncDestroyEncoder(nvenc.nvenc_ctx);
+
+	CUcontext dummy;
+	nvenc.cuda_dl->cuCtxPopCurrent(&dummy);
+
+	nvenc.cuda_dl->cuCtxDestroy(nvenc.cu_ctx);
+
+	nvenc_free_functions(&nvenc.nvenc_dl);
+	cuda_free_functions(&nvenc.cuda_dl);
+}
+/*
+int main() {
+
+	unsigned w = 256, h = 256, kbps = 400, fps = 15;
+
+	memset(&nvenc, 0, sizeof(NvencDynLoadFunctions));
+	if (loadfuncs() < 0)
+		return 1;
+	if (setupdevice() < 0)
+		return 1;
+	if (setupenc(w, h, kbps, fps) < 0)
+		return 1;
+	if (setupsurf(w, h) < 0)
+		return 1;
+
+	unload();
+
+	return 0;
+}
+*/
+
+int nvidia_init(const unsigned w, const unsigned h, const unsigned kbps,
+		const unsigned fps) {
+
+	memset(&nvenc, 0, sizeof(NvencDynLoadFunctions));
+	if (loadfuncs() < 0)
+		return 1;
+	if (setupdevice() < 0)
+		return 1;
+	if (setupenc(w, h, kbps, fps) < 0)
+		return 1;
+	if (setupsurf(w, h) < 0)
+		return 1;
+
+	return 0;
+}

common/rfb/nvidia.h

@@ -0,0 +1,11 @@
+#ifndef KASM_NVIDIA_H
+#define KASM_NVIDIA_H
+
+#include <stdint.h>
+
+int nvidia_init(const unsigned w, const unsigned h, const unsigned kbps,
+		const unsigned fps);
+int nvenc_frame(const uint8_t *data, unsigned pts, uint8_t *out, uint32_t &outlen);
+void nvidia_unload();
+
+#endif

debian/control

@@ -3,7 +3,8 @@ Section: x11
 Priority: optional
 Maintainer: Kasm Technologies LLC <info@kasmweb.com>
 Build-Depends: debhelper (>= 11), rsync, libjpeg-dev, libjpeg-dev, libpng-dev,
-  libtiff-dev, libgif-dev, libavcodec-dev, libssl-dev, libgl1, libxfont2, libsm6
+  libtiff-dev, libgif-dev, libavcodec-dev, libssl-dev, libgl1, libxfont2, libsm6,
+  libx264-dev
 Standards-Version: 4.1.3
 Homepage: https://github.com/kasmtech/KasmVNC
 #Vcs-Browser: https://salsa.debian.org/debian/kasmvnc
@@ -12,7 +13,7 @@ Homepage: https://github.com/kasmtech/KasmVNC
 Package: kasmvncserver
 Architecture: amd64
 Depends: ${shlibs:Depends}, ${misc:Depends}, ${perl:Depends}, ssl-cert, xauth,
-  x11-xkb-utils, xkb-data, procps
+  x11-xkb-utils, xkb-data, procps, x264
 Provides: vnc-server
 Description: VNC server accessible from a web browser
  VNC stands for Virtual Network Computing. It is, in essence, a remote

kasmweb

@@ -1 +1 @@
-Subproject commit 67466077c07377178599315b0cba01440ce6fb53
+Subproject commit cd7a460920aa2a958023222d148f29ace77ebe2c

unix/xserver/hw/vnc/Makefile.am

@@ -46,7 +46,7 @@ Xvnc_CPPFLAGS = $(XVNC_CPPFLAGS) -DKASMVNC -DNO_MODULE_EXTS \
 	-I$(top_srcdir)/include ${XSERVERLIBS_CFLAGS} -I$(includedir)
 
 Xvnc_LDADD = $(XVNC_LIBS) libvnccommon.la $(COMMON_LIBS) \
-	$(XSERVER_LIBS) $(XSERVER_SYS_LIBS) $(XVNC_SYS_LIBS) -lX11 -lwebp -lssl -lcrypto -lcrypt
+	$(XSERVER_LIBS) $(XSERVER_SYS_LIBS) $(XVNC_SYS_LIBS) -lX11 -lwebp -lssl -lcrypto -lcrypt -lx264
 
 Xvnc_LDFLAGS = $(LD_EXPORT_SYMBOLS_FLAG) -fopenmp
 

unix/xserver/hw/vnc/XserverDesktop.cc

@@ -74,7 +74,7 @@ XserverDesktop::XserverDesktop(int screenIndex_,
   : screenIndex(screenIndex_),
     server(0), listeners(listeners_),
     directFbptr(true),
-    queryConnectId(0), queryConnectTimer(this)
+    queryConnectId(0), queryConnectTimer(this), resizing(false)
 {
   format = pf;
 
@@ -251,7 +251,7 @@ void XserverDesktop::setCursor(int width, int height, int hotX, int hotY,
   }
 
   try {
-    server->setCursor(width, height, Point(hotX, hotY), cursorData);
+    server->setCursor(width, height, Point(hotX, hotY), cursorData, resizing);
   } catch (rdr::Exception& e) {
     vlog.error("XserverDesktop::setCursor: %s",e.str());
   }
@@ -462,8 +462,11 @@ unsigned int XserverDesktop::setScreenLayout(int fb_width, int fb_height,
   layout.print(buffer, sizeof(buffer));
   vlog.debug("%s", buffer);
 
+  resizing = true;
   vncSetGlueContext(screenIndex);
-  return ::setScreenLayout(fb_width, fb_height, layout, &outputIdMap);
+  const unsigned int ret = ::setScreenLayout(fb_width, fb_height, layout, &outputIdMap);
+  resizing = false;
+  return ret;
 }
 
 void XserverDesktop::handleClipboardRequest()

unix/xserver/hw/vnc/XserverDesktop.h

@@ -132,5 +132,7 @@ private:
   OutputIdMap outputIdMap;
 
   rfb::Point oldCursorPos;
+
+  bool resizing;
 };
 #endif

unix/xserver/hw/vnc/Xvnc.man

@@ -256,6 +256,10 @@ Scaling method to use when in downscaled video mode. 0 = nearest, 1 = bilinear,
 Default \fB2\fP.
 .
 .TP
+.B \-x264Bitrate \fInum\fP
+Enable x264 encoding for full-screen video, in kbps. Default \fB0\fP (off).
+.
+.TP
 .B \-CompareFB \fImode\fP
 Perform pixel comparison on framebuffer to reduce unnecessary updates. Can
 be either \fB0\fP (off), \fB1\fP (always) or \fB2\fP (auto). Default is