VMware Fusion 8.0.0 or 8.5.0 running Ubuntu 16.04.0 LTS 64-bit incorrectly restricts the Linux framebuffer to 4 MB of VRAM. This results in a maximum resolution of 1176 x 885 and a few other problems.
The same code running on non-virtualized Ubuntu 16.04.0 LTS 64-bit has no framebuffer issues.
The same code under Fusion 8.0.0 running Ubuntu 14.04.0 LTS 64-bit or Lubuntu 15.04 64-bit has no framebuffer issues.
The vmware.log file for the Ubuntu 16.04.0 LTS 64-bit virtual machine shows the following (with emphasis added):
vmx | I125: SVGA-GFB: Initial gfbSize=4194304
vmx | I125: SVGA: Number of displays 1
vmx | I125: SVGA: Maximum display topology 2560x1600.
vmx | I125: SVGA: Truncated maximum resolution to VRAM size: 4194304 bytes VRAM, 1176x885 Max Resolution
And later:
vmx| I125: SVGA-GFB: Allocated fgbSize=4194304
vmx| A100: ConfigDB: Setting vmotion.checkpointFBSize = "4194304"
vmx| A100: ConfigDB: Setting vmotion.checkpointSVGAPrimarySize = "67108864"
vmx| I125: SVGA-PCI: BAR gfbSize=134217728, fifoSize=3888608
vmx| I125: SVGA: SVGA_REG_MEMORY_SIZE=4194304
vmx| I125: SVGA: SVGA_REG_VRAM_SIZE=4194304
Adding the following to the .vmx file did not make any difference:
svga.vramSize = "67108864"
svga.gfbSize = "67108864"
svga.autodetect = "FALSE"
The following code demonstrates the error:
#include <linux/fb.h>
#include <stdio.h>
#include <fcntl.h>
#include <errno.h>
int main() {
struct fb_var_screeninfo vinfo;
int result = 0;
int fd = open("/dev/fb0", O_RDWR);
if(fd < 0) return -1;
result = ioctl(fd, FBIOGET_VSCREENINFO, &vinfo);
if(result) return -1;
vinfo.xres = 1920;
vinfo.yres = 1080;
result = ioctl(fd, FBIOPUT_VSCREENINFO, &vinfo);
if(result) {
fprintf(stderr, "DEBUG: result=%d, errno=%d.\n", result, errno);
return -1;
}
return 0;
}
What can be done to prevent truncation of the VRAM and framebuffer memory?
The host is an iMac (Retina 5K, 27-inch, 3.3 GHz Intel Core i5, 32 GB 1600 MHz DDR3, AME Radeon R9 M290 2040 MB) running OS X 10.10.2.
