Re: [Linux-fbdev-devel] Portrait display mode

[Antonino D. <ad...@po...>]

On Wed, 2002-09-11 at 14:38, Geert Uytterhoeven wrote: 

> > > turnable display I rather see this feature implemented. I even may help but can
> > > spend much time. How difficult would it be?
> > > 
> > It should not be too difficult to add per-driver support once the
> 
> Can't this be moved up to generic code, so we don't have to modify all drivers?
> 
I have thought about that, but since the patch was preliminary, I chose
not to disrupt the framework yet.  The tricky part in making the support
generic is display panning. 

Here is my train of thought.  Moving support into fbgen might involve a
few assumptions.  Firstly, info->var is assumed to be device specific
(ie, it does not care about the orientation). Secondly, the display->var
as well as those received from other sources (such as in an ioctl call)
are partly device independent.  When a structure var is passed to the
driver, it will undergo conversion so the structure is sane from the
driver's perspective.  The same is done vice-versa.  I've introduced two
(exportable) functions, convert_var_to_physical(), and
convert_var_to_logical() that should do that.  

All fbops functions should be passed with device-specific var except for
fb_set_var since I'm not too sure where to look for specific entry
points for fb_set_var.  I guess fb_check_var and fb_pan_display are the
most critical ones, since the other fbops functions do not look at the
framelengths, nor the offsets. 

With the changes, drivers that support fbcon_accel and gen_* functions
should have support for console rotation automatically.  

For those that support fbcon_accel, but not gen_* functions, the driver
should not look at display->var anymore, but look instead into
info->var. If possible, use the convert_to_var functions when
passing/getting information to/from structure display. 

For those, that don't support fbcon_accel, behaviour should remain as
usual. 

I'm not too sure if this is the correct way of doing this, but it's the
best I can think of at the moment without causing very intrusive
changes. 

Attached is a patch (fb_rotate2.diff).  Please apply the patches I have
sent previously before applying this patch. 

Tony