Palacios Public Git Repository

To checkout Palacios execute

  git clone http://v3vee.org/palacios/palacios.web/palacios.git
This will give you the master branch. You probably want the devel branch or one of the release branches. To switch to the devel branch, simply execute
  cd palacios
  git checkout --track -b devel origin/devel
The other branches are similar.


VGA enhancements for graphics writes and rendering
Peter Dinda [Wed, 27 Apr 2011 00:05:23 +0000 (19:05 -0500)]
   - now successfully renders vgatest
   - corrected handling of bit mask in mode 0/2/3 writes
   - corrected latch loading for chain4 reads
   - corrected expansion of SR bit (compiler bug workaround?)

palacios/src/devices/vga.c

index de6bc66..ce28940 100644 (file)
 
 #include "vga_regs.h"
 
+#ifndef CONFIG_DEBUG_VGA
+#undef PrintDebug
+#define PrintDebug(fmt, args...)
+#endif
+
+#define DEBUG_MEM_DATA    0
+#define DEBUG_DEEP_MEM    0
+#define DEBUG_DEEP_RENDER 0
+
+
 #define MEM_REGION_START 0xa0000
 #define MEM_REGION_END   0xc0000
 #define MEM_REGION_NUM_PAGES (((MEM_REGION_END)-(MEM_REGION_START))/4096)
@@ -365,6 +375,7 @@ static void find_text_data_start(struct vga_internal *vga, void **data)
 
 }
 
+
 static void find_text_attr_start(struct vga_internal *vga, void **data)
 {
     uint32_t offset;
@@ -446,6 +457,29 @@ static int blinking(struct vga_internal *vga)
 }
 
 
+static void find_graphics_data_starting_offset(struct vga_internal *vga, uint32_t *offset)
+{
+
+    *offset = vga->vga_crt_controller.vga_start_address_high;
+    *offset <<= 8;
+    *offset += vga->vga_crt_controller.vga_start_address_low;
+}
+
+
+static void find_shift_mode(struct vga_internal *vga, shift_mode_t *mode)
+{
+    if (vga->vga_graphics_controller.vga_graphics_mode.c256) { 
+       *mode=C256_SHIFT;
+    } else {
+       if (vga->vga_graphics_controller.vga_graphics_mode.shift_reg_mode) {
+           *mode=PACKED_SHIFT;
+       } else {
+           *mode=PLANAR_SHIFT;
+       }
+    }
+}
+
+
 static void find_graphics_res(struct vga_internal *vga, uint32_t *width, uint32_t *height)
 {
     uint32_t vert_lsb, vert_msb;
@@ -460,6 +494,31 @@ static void find_graphics_res(struct vga_internal *vga, uint32_t *width, uint32_
        + (vga->vga_crt_controller.vga_overflow.vertical_disp_enable_end8);
               
     *height  = ( (vert_msb << 8) + vert_lsb + 1) ; // pixels high (scanlines)
+
+    // At this point we have the resolution in dot clocks across and scanlines top-to-bottom
+    // This is usually the resolution in pixels, but it can be monkeyed with
+    // at least in the following ways
+
+    // vga sequencer dot clock divide by two 
+    if (vga->vga_sequencer.vga_clocking_mode.dot_clock) { 
+       *width/=2;
+       *height/=2;
+    }
+
+    // crt_controller.max_row_scan.double_scan => each row twice for 200=>400
+    if (vga->vga_crt_controller.vga_max_row_scan.double_scan) { 
+       *height/=2;
+    }
+    
+    // crt_controller.crt_mode_control.count_by_two => pixels twice as wide as normal
+    if (vga->vga_crt_controller.vga_crt_mode_control.count_by_two) { 
+       *width /= 2;
+    }
+
+    // crt_controller.crt_mode_control.horizontal_retrace_select => pixels twice as tall as normal
+    if (vga->vga_crt_controller.vga_crt_mode_control.horizontal_retrace_select) { 
+       *height /= 2;
+    }
     
 }
 
@@ -471,19 +530,6 @@ static void find_graphics_cursor_pos(struct vga_internal *vga, uint32_t *x, uint
 }
 
 
-static void find_shift_mode(struct vga_internal *vga, shift_mode_t *mode)
-{
-    if (vga->vga_graphics_controller.vga_graphics_mode.c256) { 
-       *mode=C256_SHIFT;
-    } else {
-       if (vga->vga_graphics_controller.vga_graphics_mode.shift_reg_mode) {
-           *mode=PACKED_SHIFT;
-       } else {
-           *mode=PLANAR_SHIFT;
-       }
-    }
-}
-
 static void dac_lookup_24bit_color(struct vga_internal *vga,
                                   uint8_t entry,
                                   uint8_t *red,
@@ -589,26 +635,28 @@ static void render_graphics(struct vga_internal *vga, void *fb)
 
     find_graphics_cursor_pos(vga,&cur_x,&cur_y);
 
+    find_graphics_data_starting_offset(vga,&offset);
+
     fw = spec->width;
     fh = spec->height;
 
 
     PrintDebug("vga: attempting graphics render (%s): graphics_res=(%u,%u), fb_res=(%u,%u), "
-               "fb=0x%p (16 color mode)\n",
+               "fb=0x%p offset=0x%x\n",
               sm == PLANAR_SHIFT ? "planar shift" : 
               sm == PACKED_SHIFT ? "packed shift" : 
               sm == C256_SHIFT ? "color256 shift" : "UNKNOWN",
-              gw,gh,fw,fh, fb);
+              gw,gh,fw,fh,fb,offset);
 
     // First we need to clip to what we can actually show
     rgw = gw < fw ? gw : fw;
-    rgh = gh < fh ? gh : gw;
+    rgh = gh < fh ? gh : fh;
 
     if (gw%8) { 
        PrintError("vga: warning: graphics width is not a multiple of 8\n");
     }
 
-    offset=0;  // do we always start at zero in the map?
+
 
     // Now we scan across by row
     for (fy=0;fy<gh;fy++) { 
@@ -616,7 +664,7 @@ static void render_graphics(struct vga_internal *vga, void *fb)
        for (fx=0;fx<gw;
             fx += (sm==C256_SHIFT ? 4 : 8) , offset++ ) { 
 
-           // if any of these pixels are in the renger region
+           // if any of these pixels are in the rendger region
            if (fy < rgh && fx < rgw) {
                // assemble all 4 or 8 pixels
                
@@ -630,10 +678,10 @@ static void render_graphics(struct vga_internal *vga, void *fb)
                    case PLANAR_SHIFT:
                        for (p=0;p<8;p++) { 
                            pb[p]= 
-                               ( db[0] >> 7)  |
-                               ( db[1] >> 6)  |
-                               ( db[2] >> 5)  |
-                               ( db[3] >> 4) ;
+                               (( db[0] >> 7) & 0x1) |
+                               (( db[1] >> 6) & 0x2) |
+                               (( db[2] >> 5) & 0x4) |
+                               (( db[3] >> 4) & 0x8) ;
                            
                            for (m=0;m<4;m++) { 
                                db[m] <<= 1;
@@ -645,8 +693,8 @@ static void render_graphics(struct vga_internal *vga, void *fb)
                        // first 4 pixels use planes 0 and 2
                        for (p=0;p<4;p++) { 
                            pb[p] = 
-                               ( db[2] >> 4) |
-                               ( db[0] >> 6) ;
+                               ((db[2] >> 4) & 0xc) |
+                               ((db[0] >> 6) & 0x3) ;
                            db[2] <<= 2;
                            db[0] <<= 2;
                        }
@@ -655,8 +703,8 @@ static void render_graphics(struct vga_internal *vga, void *fb)
                        // next 4 pixels use planes 1 and 3
                        for (p=4;p<8;p++) { 
                            pb[p] = 
-                               ( db[3] >> 4) |
-                               ( db[1] >> 6) ;
+                               ((db[3] >> 4) & 0xc) |
+                               ((db[1] >> 6) & 0x3) ;
                            db[3] <<= 2;
                            db[1] <<= 2;
                        }
@@ -674,7 +722,7 @@ static void render_graphics(struct vga_internal *vga, void *fb)
                for (p=0;p< (sm==C256_SHIFT ? 4 : 8);p++) { 
                    
                    // find its color
-                   find_24bit_color(vga,db[p],&r,&g,&b);
+                   find_24bit_color(vga,pb[p],&r,&g,&b);
                
                    // find its position in the framebuffer;
                    pixel =  fb + (((fx + p) + (fy*spec->width)) * spec->bytes_per_pixel);
@@ -909,6 +957,15 @@ static void render_test(struct vga_internal *vga, void *fb)
     }
 }
 
+static void render_black(struct vga_internal *vga, void *fb)
+{
+    struct v3_frame_buffer_spec *s;
+
+    s=&(vga->target_spec);
+
+    memset(fb,0,s->height*s->width*s->bytes_per_pixel);
+}
+
 static void render_maps(struct vga_internal *vga, void *fb)
 {
 
@@ -956,15 +1013,20 @@ static int render(struct vga_internal *vga)
 
        fb = v3_graphics_console_get_frame_buffer_data_rw(vga->host_cons,&(vga->target_spec));
 
-       if (vga->vga_graphics_controller.vga_misc.graphics_mode) { 
-           render_graphics(vga,fb);
+       if (!(vga->vga_sequencer.vga_clocking_mode.screen_off)) {
+           if (vga->vga_attribute_controller.vga_attribute_mode_control.graphics) { 
+               render_graphics(vga,fb);
+           } else {
+               render_text(vga,fb);
+               render_text_cursor(vga,fb);
+           }
        } else {
-           render_text(vga,fb);
-           render_text_cursor(vga,fb);
+           render_black(vga,fb);
        }
 
        if (0) { render_test(vga,fb); }
 
+       // always render maps for now 
        render_maps(vga,fb);
 
        v3_graphics_console_release_frame_buffer_data_rw(vga->host_cons);
@@ -1065,7 +1127,7 @@ static int vga_write(struct guest_info * core,
        memcpy(V3_VAddr((void*)guest_addr),src,length);
     }
     
-#if 0
+#if DEBUG_MEM_DATA
     int i;
     PrintDebug("vga: data written was 0x");
     for (i=0;i<length;i++) {
@@ -1107,7 +1169,9 @@ static int vga_write(struct guest_info * core,
            
            offset = find_offset_write(vga, guest_addr);
 
+#if DEBUG_DEEP_MEM
            PrintDebug("vga: mode 0 write, offset=0x%llx, ror=%u, func=%u\n", offset,ror,func);
+#endif
 
            for (i=0;i<length;i++,offset+=find_increment_write(vga,guest_addr+i)) { 
                // now for each map
@@ -1116,21 +1180,36 @@ static int vga_write(struct guest_info * core,
                uint8_t bm = vga->vga_graphics_controller.vga_bit_mask;
                uint8_t mm = find_map_write(vga,guest_addr+i);
 
-               //PrintDebug("vga: write i=%u, mm=0x%x, offset=0x%x\n",i,(unsigned int)mm,(unsigned int)offset);
+#if DEBUG_DEEP_MEM
+               PrintDebug("vga: write i=%u, mm=0x%x, bm=0x%x sr=0x%x esr=0x%x offset=0x%x\n",i,(unsigned int)mm,(unsigned int)bm, (unsigned int)sr, (unsigned int)esr,(unsigned int)offset);
+#endif
 
-               for (mapnum=0;mapnum<4;mapnum++, sr>>=1, esr>>=1, bm>>=1, mm>>=1) { 
+               for (mapnum=0;mapnum<4;mapnum++, sr>>=1, esr>>=1, mm>>=1) { 
                    vga_map map = vga->map[mapnum];
                    uint8_t data = ((uint8_t *)src)[i];
                    uint8_t latchval = vga->latch[mapnum];
-                       
+                   
+#if DEBUG_DEEP_MEM
+                   PrintDebug("vga: raw data=0x%x\n",data);
+#endif
                    // rotate data right
-                   data = (data>>ror) | data<<(8-ror);
-
+                   if (ror) { 
+                       data = (data>>ror) | data<<(8-ror);
+                   }
+                   
+#if DEBUG_DEEP_MEM
+                   PrintDebug("vga: data after ror=0x%x\n",data);
+#endif
                    // use SR bit if ESR is on for this map
-                   if (esr & 0x1) { 
-                       data = (uint8_t)((((sint8_t)(sr&0x1))<<7)>>7);  // expand sr bit
+                   if (esr & 0x1) {
+                       data = (sr&0x1) * -1;
+                       
                    }
                    
+#if DEBUG_DEEP_MEM
+                   PrintDebug("vga: data after esrr=0x%x\n",data);
+#endif
+                   
                    // Apply function
                    switch (func) { 
                        case 0: // NOP
@@ -1145,19 +1224,26 @@ static int vga_write(struct guest_info * core,
                            data ^= latchval;
                            break;
                    }
-                           
-                   // mux between latch and alu output
-                   if (bm & 0x1) { 
-                       // use alu output, which is in data
-                   } else {
-                       // use latch value
-                       data=latchval;
-                   }
+                   
+#if DEBUG_DEEP_MEM
+                   PrintDebug("vga: data after func=0x%x\n",data);
+#endif
+                   
+                   // mux between the data byte and the latch byte on
+                   // a per-bit basis
+                   data = (bm & data) | ((~bm) & latchval);
+                   
+
+#if DEBUG_DEEP_MEM
+                   PrintDebug("vga: data after bm mux=0x%x\n",data);
+#endif
                    
                    // selective write
                    if (mm & 0x1) { 
                        // write to this map
-                       //PrintDebug("vga: write map %u offset 0x%p map=0x%p pointer=0x%p\n",mapnum,(void*)offset,map,&(map[offset]));
+#if DEBUG_DEEP_MEM
+                       PrintDebug("vga: write map %u offset 0x%p map=0x%p pointer=0x%p\n",mapnum,(void*)offset,map,&(map[offset]));
+#endif
                        map[offset] = data;
                    } else {
                        // skip this map
@@ -1180,7 +1266,9 @@ static int vga_write(struct guest_info * core,
 
            uint64_t offset = find_offset_write(vga,guest_addr);
 
+#if DEBUG_DEEP_MEM
            PrintDebug("vga: mode 1 write, offset=0x%llx\n", offset);
+#endif
 
            for (i=0;i<length;i++,offset+=find_increment_write(vga,guest_addr+i)) { 
 
@@ -1221,7 +1309,9 @@ static int vga_write(struct guest_info * core,
            
            offset = find_offset_write(vga, guest_addr);
 
+#if DEBUG_DEEP_MEM
            PrintDebug("vga: mode 2 write, offset=0x%llx, func=%u\n", offset,func);
+#endif
 
            for (i=0;i<length;i++,offset+=find_increment_write(vga,guest_addr+i)) { 
                // now for each map
@@ -1234,9 +1324,9 @@ static int vga_write(struct guest_info * core,
                    uint8_t latchval = vga->latch[mapnum];
                        
                    // expand relevant bit to 8 bit
-                   // it's basically esr=1, sr=bit from write
-                   data = (uint8_t)(((sint8_t)(((data>>mapnum)&0x1)<<7))>>7);
-                   
+                   // it's basically esr=1, sr=bit from mode 0 write
+                   data = ((data>>mapnum)&0x1) * -1;
+                       
                    // Apply function
                    switch (func) { 
                        case 0: // NOP
@@ -1251,14 +1341,9 @@ static int vga_write(struct guest_info * core,
                            data ^= latchval;
                            break;
                    }
-                           
+
                    // mux between latch and alu output
-                   if (bm & 0x1) { 
-                       // use alu output, which is in data
-                   } else {
-                       // use latch value
-                       data=latchval;
-                   }
+                   data = (bm & data) | ((~bm) & latchval);
                    
                    // selective write
                    if (mm & 0x1) { 
@@ -1298,7 +1383,9 @@ static int vga_write(struct guest_info * core,
                // now for each map
                uint8_t data = ((uint8_t *)src)[i];
 
-               data = (data>>ror) | data<<(8-ror);
+               if (ror) {
+                   data = (data>>ror) | data<<(8-ror);
+               }
 
                uint8_t bm = vga->vga_graphics_controller.vga_bit_mask & data;
                uint8_t sr = vga->vga_graphics_controller.vga_set_reset.val & 0xf;
@@ -1308,16 +1395,11 @@ static int vga_write(struct guest_info * core,
                    vga_map map = vga->map[mapnum];
                    uint8_t latchval = vga->latch[mapnum];
                        
-                   data = (uint8_t)((((sint8_t)(sr&0x1))<<7)>>7);  // expand sr bit
-                   
-                   
+                   // expand SR bit
+                   data = (sr&0x1) * -1;
+
                    // mux between latch and alu output
-                   if (bm & 0x1) { 
-                       // use alu output, which is in data
-                   } else {
-                       // use latch value
-                       data=latchval;
-                   }
+                   data = (bm & data) | ((~bm) & latchval);
                    
                    // selective write
                    if (mm & 0x1) { 
@@ -1403,7 +1485,15 @@ static int vga_read(struct guest_info * core,
                // address bytes select the map
                for (i=0;i<length;i++,offset+=find_increment_read(vga,guest_addr+i)) { 
                    mapnum = (guest_addr+i) % 4;
-                   ((uint8_t*)dst)[i] = vga->latch[mapnum] = *(vga->map[mapnum]+offset);
+#if DEBUG_DEEP_MEM
+                   PrintDebug("vga: mode 0 read, chain4, offset=0x%llx, mapnum=%u\n",offset,mapnum);
+#endif
+                   ((uint8_t*)dst)[i] = *(vga->map[mapnum]+offset);
+
+                   // presumably all the latches are to be reloaded, not just the selected one?
+                   for (mapnum=0;mapnum<4;mapnum++) { 
+                       vga->latch[mapnum] = *(vga->map[mapnum]+offset);
+                   }
                }
            } else {
                mapnum = vga->vga_graphics_controller.vga_read_map_select.map_select;
@@ -1413,6 +1503,10 @@ static int vga_read(struct guest_info * core,
                    PrintError("vga: read to offset=%llu map=%u (%u bytes)\n",offset,mapnum,length);
                }
                
+#if DEBUG_DEEP_MEM
+               PrintDebug("vga: mode 0 read, not-chain4, offset=0x%llx, mapnum=%u\n",offset,mapnum);
+#endif
+
                memcpy(dst,(vga->map[mapnum])+offset,length);
                
                // load the latches with the last item read
@@ -1443,6 +1537,11 @@ static int vga_read(struct guest_info * core,
            uint8_t  bits;
            
            offset = find_offset_read(vga,guest_addr);
+
+#if DEBUG_DEEP_MEM
+           PrintDebug("vga: mode 1 read, offset=0x%llx, cc=0x%x, dc-0x%x\n",offset,cc,dc);
+#endif
+               
            
            for (i=0;i<length;i++,offset++) { 
                vga_map map;
@@ -1480,7 +1579,7 @@ static int vga_read(struct guest_info * core,
     }
 
 
-#if 0
+#if DEBUG_MEM_DATA
     int i;
     PrintDebug("vga: data read is 0x");
     for (i=0;i<length;i++) {