Re: Accesing vt(4) framebuffer

On Wed, 3 Sept 2025 at 09:36, Ahmad Khalifa <ahmadkhalifa570@gmail.com>
wrote:

> On Wed Sep 3, 2025 at 6:23 PM +0300, Adrian Chadd wrote:
> > hi!
> >
> > We put code review requests in the tool at https://reviews.freebsd.org/
> .
> > If you've not heard back about this
> > then please feel free to poke me directly.
> >
> > Having direct fb access from userland would be great. I thought we
> already
> > did?!
>
> We can access it through mmap (see vd_fb_mmap). This patch just seems to
> expose some framebuffer attributes. Although I think (?) that this info
> is available through FBIOGTYPE.


i tried this tonight (as I'm debugging some fb font stuff, and i want to
try writing to the framebuffer from userland).

FBIOGTYPE however just exposes/copies a kernel struct, without scrubbing it
(pointers and all), and its inside
#ifdef KERNEL.

So i do think it's worth creating a struct thats specifically the useful
userland contract bits.



-adrian



> >
> >
> >
> > -adrian
> >
> > On Tue, 26 Aug 2025 at 07:10, Eduardo Morrás <emorras@emorras.eu> wrote:
> >
> >>
> >> Hello, I'm developing a gui/widget over vt(4) framebuffer, no X11 nor
> >> Wayland.
> >>
> >> For this I have created a new ioctl that allows access to framebuffer
> >> information and change it's content. For this I modified
> >> /usr/src/sys/dev/vt/hw/fb/vt_fb.c and .h, patches at end of this mail.
> >>
> >> Is this a correct way? Should I do it different? Any security problems?
> >>
> >>
> >> Thanks, Good Day.
> >>
> >>
> >> --- vt_fb.h_ORIG        2025-08-21 19:01:01.192050000 +0200
> >> +++ vt_fb.h     2025-08-21 19:04:58.447415000 +0200
> >> @@ -49,4 +49,20 @@
> >>  vd_fb_ioctl_t          vt_fb_ioctl;
> >>  vd_fb_mmap_t           vt_fb_mmap;
> >>
> >> +#ifdef _KERNEL
> >> +#include <sys/ioccom.h>
> >> +#endif
> >> +
> >> +/* Struct to expose framebuffer to userland*/
> >> +struct vt_fb_userinfo {
> >> +    uint32_t width;
> >> +    uint32_t height;
> >> +    uint32_t depth;
> >> +    uint32_t stride;
> >> +    uint32_t size;
> >> +};
> >> +
> >> +/* IOCTL code to work with vt_fb_userinfo */
> >> +#define VT_FB_GETINFO _IOR('v', 1, struct vt_fb_userinfo)
> >> +
> >>  #endif /* _DEV_VT_HW_FB_VT_FB_H_ */
> >>
> >>
> >> --- vt_fb.c_ORIG        2025-08-21 19:01:07.096190000 +0200
> >> +++ vt_fb.c     2025-08-21 19:10:22.228787000 +0200
> >> @@ -42,6 +42,8 @@
> >>  #include <vm/vm.h>
> >>  #include <vm/pmap.h>
> >>
> >> +#include <sys/ioccom.h> /* Userland access framebuffer */
> >> +
> >>  static struct vt_driver vt_fb_driver = {
> >>         .vd_name = "fb",
> >>         .vd_init = vt_fb_init,
> >> @@ -87,6 +89,24 @@
> >>         *(uint32_t *)(sc->fb_vbase + o) = v;
> >>  }
> >>
> >> +static int
> >> +vt_fb_userioctl(struct vt_device *vd, u_long cmd, caddr_t data)
> >> +{
> >> +       struct vt_fb_userinfo *ui = (struct vt_fb_userinfo *)data;
> >> +       struct fb_info *info = vd->vd_softc;
> >> +
> >> +       if (cmd != VT_FB_GETINFO)
> >> +               return (ENOIOCTL);
> >> +
> >> +       ui->width  = info->fb_width;
> >> +       ui->height = info->fb_height;
> >> +       ui->depth  = info->fb_bpp * 8;
> >> +       ui->stride = info->fb_stride;
> >> +       ui->size   = info->fb_size;
> >> +
> >> +       return (0);
> >> +}
> >> +
> >>  int
> >>  vt_fb_ioctl(struct vt_device *vd, u_long cmd, caddr_t data, struct
> thread
> >> *td)
> >>  {
> >> @@ -125,6 +145,10 @@
> >>                         return (ENOTTY);
> >>                 memcpy((struct fb_rgboffs *)data, &info->fb_rgboffs,
> >>                     sizeof(struct fb_rgboffs));
> >> +               break;
> >> +
> >> +       case  VT_FB_GETINFO:   /* userland framebuffer */
> >> +               return vt_fb_userioctl(vd, cmd, data);
> >>                 break;
> >>
> >>         default:
> >>
> >>
> >> ---   ---
> >> Eduardo Morrás <emorras@emorras.eu>
> >>
> >>
>