xorg-video-intel: Implement support for 24 bit pixel format

From: Mike Isely <isely@pobox.com>

The Intel driver appears to be coded to only work with displays
expecting 18 bit pixels.  However I have an application using a LCD
display that expects pixel data in 24 bit format.  The difference is
only 2 bits in a single GPU register.  This patch implements that
change, controlled by a new driver option, "LVDS24Bit".  The default
value is false, which is the previous behavior.  When set to true,
then 24 bit panels should work (at least the one I'm testing here
does).  A couple notes:

1. I noticed a comment in the source (right where I made the change)
suggesting the desire to support 24 bit pixels, but expressing some
concern about understanding how to do this correctly.  I can attest to
this patch working for my situation, and it's done in a
non-interfering manner (i.e. controlled via driver option) so it
should be safe to apply.

2. One other thought I had was that perhaps rather than controlling
this with a driver option that instead we simply notice if the user is
asking for a 24 bit (or 32 bit) display depth.  If the bit depth is
deep enough, then enable the 24 bit pixel format.  However that really
isn't right.  The display depth setting is really a user preference
not a setting that must match some esoteric hardware attribute.  I
think to drive this setting magically from the specified bit depth is
just going to ask for a lot of confusion going forwards.  Besides
there are good reasons for example to use a 16 bit frame buffer while
still needing the 24 bit pixel format (like performance).  So the
pixel format really should be separate from the frame buffer display
depth.

3. It seems still that there should be a way to automatically control
this.  I have no idea however what would be.  I'm just a hack at this.
But I do strongly believe that at least implementing a manual control
is better than no control at all.  Thus this patch.

Signed-off-by: Mike Isely <isely@pobox.com>

---

Note: This patch was built against the 2.2.1 driver as patched in
Debian 2.2.1-1.  However it should apply reasonably well against the
upstream tree (I hope).

 src/i830.h         |    2 ++
 src/i830_display.c |   19 +++++++++++++++----
 src/i830_driver.c  |    8 ++++++++
 3 files changed, 25 insertions(+), 4 deletions(-)

Index: src/i830_driver.c
===================================================================
--- src/i830_driver.c	(.../debian-2.2.1-1)	(revision 40)
+++ src/i830_driver.c	(.../patch-lvds24bit)	(revision 40)
@@ -292,6 +292,7 @@
    OPTION_COLOR_KEY,
    OPTION_CHECKDEVICES,
    OPTION_MODEDEBUG,
+   OPTION_LVDS24BITMODE,
    OPTION_FBC,
    OPTION_TILING,
 #ifdef XF86DRI_MM
@@ -315,6 +316,7 @@
    {OPTION_VIDEO_KEY,	"VideoKey",	OPTV_INTEGER,	{0},	FALSE},
    {OPTION_CHECKDEVICES, "CheckDevices",OPTV_BOOLEAN,	{0},	FALSE},
    {OPTION_MODEDEBUG,	"ModeDebug",	OPTV_BOOLEAN,	{0},	FALSE},
+   {OPTION_LVDS24BITMODE, "LVDS24Bit",	OPTV_BOOLEAN,	{0},	FALSE},
    {OPTION_FBC,		"FramebufferCompression", OPTV_BOOLEAN, {0}, TRUE},
    {OPTION_TILING,	"Tiling",	OPTV_BOOLEAN,	{0},	TRUE},
 #ifdef XF86DRI_MM
@@ -1195,6 +1197,12 @@
       pI830->debug_modes = FALSE;
    }
 
+   if (xf86ReturnOptValBool(pI830->Options, OPTION_LVDS24BITMODE, FALSE)) {
+      pI830->lvds_24_bit_mode = TRUE;
+   } else {
+      pI830->lvds_24_bit_mode = FALSE;
+   }
+
    if (xf86ReturnOptValBool(pI830->Options, OPTION_FORCEENABLEPIPEA, FALSE))
        pI830->quirk_flag |= QUIRK_PIPEA_FORCE;
 
Index: src/i830_display.c
===================================================================
--- src/i830_display.c	(.../debian-2.2.1-1)	(revision 40)
+++ src/i830_display.c	(.../patch-lvds24bit)	(revision 40)
@@ -1221,10 +1221,21 @@
 	else
 	    lvds &= ~(LVDS_B0B3_POWER_UP | LVDS_CLKB_POWER_UP);
 
-	/* It would be nice to set 24 vs 18-bit mode (LVDS_A3_POWER_UP)
-	 * appropriately here, but we need to look more thoroughly into how
-	 * panels behave in the two modes.
-	 */
+	if (pI830->lvds_24_bit_mode) {
+	    /* Option set which requests 24-bit mode
+	     * (LVDS_A3_POWER_UP, as opposed to 18-bit mode) here; we
+	     * still need to look more thoroughly into how panels
+	     * behave in the two modes.  This option enables that
+	     * experimentation.
+	     */
+	    xf86DrvMsg(pScrn->scrnIndex, X_INFO,
+		       "Selecting less common 24 bit TMDS pixel format.\n");
+	    lvds |= LVDS_A3_POWER_UP;
+	    lvds |= LVDS_DATA_FORMAT_DOT_ONE;
+	} else {
+	    xf86DrvMsg(pScrn->scrnIndex, X_INFO,
+		       "Selecting standard 18 bit TMDS pixel format.\n");
+	}
 
 	/* Enable dithering if we're in 18-bit mode. */
 	if (IS_I965G(pI830))
Index: src/i830.h
===================================================================
--- src/i830.h	(.../debian-2.2.1-1)	(revision 40)
+++ src/i830.h	(.../patch-lvds24bit)	(revision 40)
@@ -526,6 +526,8 @@
    /* Broken-out options. */
    OptionInfoPtr Options;
 
+   Bool lvds_24_bit_mode;
+
    Bool StolenOnly;
 
    Bool swfSaved;