quoted

-----Original Message-----
From: Thomas Abraham [mailto:thomas.abraham@linaro.org]
Sent: Tuesday, March 06, 2012 2:26 PM
To: Jingoo Han
Cc: linux-fbdev@vger.kernel.org; FlorianSchandinat@gmx.de; linux-samsung-soc@vger.kernel.org;
kgene.kim@samsung.com; ben-linux@fluff.org; patches@linaro.org
Subject: Re: [PATCH 1/3] video: s3c-fb: move video interface timing out of window setup data

On 6 March 2012 10:15, Jingoo Han [off-list ref] wrote:

quoted

quoted

-----Original Message-----
From: Thomas Abraham [mailto:thomas.abraham@linaro.org]
Sent: Sunday, March 04, 2012 12:50 AM
To: linux-fbdev@vger.kernel.org
Cc: FlorianSchandinat@gmx.de; linux-samsung-soc@vger.kernel.org; kgene.kim@samsung.com;
jg1.han@samsung.com; ben-linux@fluff.org; patches@linaro.org
Subject: [PATCH 1/3] video: s3c-fb: move video interface timing out of window setup data

The video interface timing is independent of the window setup data.
The resolution of the window can be smaller than that of the lcd
panel to which the video data is output.

So move the video timing data from the per-window setup data to the
platform specific section in the platform data. This also removes
the restriction that atleast one window should have the same
resolution as that of the panel attached.

Cc: Ben Dooks <ben-linux@fluff.org>
Cc: Jingoo Han <redacted>
Signed-off-by: Thomas Abraham <redacted>
---
 arch/arm/plat-samsung/include/plat/fb.h |    9 ++-
 drivers/video/s3c-fb.c                  |  106 +++++++++++++++++--------------
 2 files changed, 63 insertions(+), 52 deletions(-)

diff --git a/arch/arm/plat-samsung/include/plat/fb.h b/arch/arm/plat-samsung/include/plat/fb.h
index 0fedf47..39d6bd7 100644
--- a/arch/arm/plat-samsung/include/plat/fb.h
+++ b/arch/arm/plat-samsung/include/plat/fb.h

@@ -24,15 +24,16 @@

 /**
  * struct s3c_fb_pd_win - per window setup data
- * @win_mode: The display parameters to initialise (not for window 0)
+ * @xres     : The window X size.
+ * @yres     : The window Y size.
  * @virtual_x: The virtual X size.
  * @virtual_y: The virtual Y size.
  */
 struct s3c_fb_pd_win {
-     struct fb_videomode     win_mode;
-
      unsigned short          default_bpp;
      unsigned short          max_bpp;
+     unsigned short          xres;
+     unsigned short          yres;
      unsigned short          virtual_x;
      unsigned short          virtual_y;
 };

@@ -45,6 +46,7 @@ struct s3c_fb_pd_win {

  * @default_win: default window layer number to be used for UI layer.
  * @vidcon0: The base vidcon0 values to control the panel data format.
  * @vidcon1: The base vidcon1 values to control the panel data output.
+ * @vtiming: Video timing when connected to a RGB type panel.

fb_videomode can be set, even if it is not RGB type panel.
In my opinion, it would be better.
+ * @vtiming: The video timing values to set the interface timing of the panel.

The other interface that is supported is the i80 interface. Can these
timing values be used for i80 interface as well ?

No, you're right. The i80 is not supported.
Please ignore my comment.

quoted

quoted

quoted

  * @win: The setup data for each hardware window, or NULL for unused.
  * @display_mode: The LCD output display mode.
  *

@@ -58,6 +60,7 @@ struct s3c_fb_platdata {

      void    (*setup_gpio)(void);

      struct s3c_fb_pd_win    *win[S3C_FB_MAX_WIN];
+     struct fb_videomode     *vtiming;

      u32                      default_win;

diff --git a/drivers/video/s3c-fb.c b/drivers/video/s3c-fb.c
index 1fb7ddf..8e05d4d 100644
--- a/drivers/video/s3c-fb.c
+++ b/drivers/video/s3c-fb.c

@@ -495,7 +495,6 @@ static int s3c_fb_set_par(struct fb_info *info)

      u32 alpha = 0;
      u32 data;
      u32 pagewidth;
-     int clkdiv;

      dev_dbg(sfb->dev, "setting framebuffer parameters\n");

@@ -532,46 +531,9 @@ static int s3c_fb_set_par(struct fb_info *info)

      /* disable the window whilst we update it */
      writel(0, regs + WINCON(win_no));

-     /* use platform specified window as the basis for the lcd timings */
-
-     if (win_no = sfb->pdata->default_win) {
-             clkdiv = s3c_fb_calc_pixclk(sfb, var->pixclock);
-
-             data = sfb->pdata->vidcon0;
-             data &= ~(VIDCON0_CLKVAL_F_MASK | VIDCON0_CLKDIR);
-
-             if (clkdiv > 1)
-                     data |= VIDCON0_CLKVAL_F(clkdiv-1) | VIDCON0_CLKDIR;
-             else
-                     data &= ~VIDCON0_CLKDIR;        /* 1:1 clock */
-
-             /* write the timing data to the panel */
-
-             if (sfb->variant.is_2443)
-                     data |= (1 << 5);
-
-             writel(data, regs + VIDCON0);
-
+     if (win_no = sfb->pdata->default_win)
              s3c_fb_enable(sfb, 1);

-             data = VIDTCON0_VBPD(var->upper_margin - 1) |
-                    VIDTCON0_VFPD(var->lower_margin - 1) |
-                    VIDTCON0_VSPW(var->vsync_len - 1);
-
-             writel(data, regs + sfb->variant.vidtcon);
-
-             data = VIDTCON1_HBPD(var->left_margin - 1) |
-                    VIDTCON1_HFPD(var->right_margin - 1) |
-                    VIDTCON1_HSPW(var->hsync_len - 1);
-
-             /* VIDTCON1 */
-             writel(data, regs + sfb->variant.vidtcon + 4);
-
-             data = VIDTCON2_LINEVAL(var->yres - 1) |
-                    VIDTCON2_HOZVAL(var->xres - 1);
-             writel(data, regs + sfb->variant.vidtcon + 8);
-     }
-

It looks good.
VIDTCON registers don't need to be written whenever s3c_fb_set_par is called.

quoted

      /* write the buffer address */

      /* start and end registers stride is 8 */

@@ -1136,11 +1098,11 @@ static int __devinit s3c_fb_alloc_memory(struct s3c_fb *sfb,

      dev_dbg(sfb->dev, "allocating memory for display\n");

-     real_size = windata->win_mode.xres * windata->win_mode.yres;
+     real_size = windata->xres * windata->yres;
      virt_size = windata->virtual_x * windata->virtual_y;

      dev_dbg(sfb->dev, "real_size=%u (%u.%u), virt_size=%u (%u.%u)\n",
-             real_size, windata->win_mode.xres, windata->win_mode.yres,
+             real_size, windata->xres, windata->yres,
              virt_size, windata->virtual_x, windata->virtual_y);

      size = (real_size > virt_size) ? real_size : virt_size;

@@ -1222,7 +1184,7 @@ static int __devinit s3c_fb_probe_win(struct s3c_fb *sfb, unsigned int win_no,

                                    struct s3c_fb_win **res)
 {
      struct fb_var_screeninfo *var;
-     struct fb_videomode *initmode;
+     struct fb_videomode initmode;

*initmode cannot be used???
Can you tell me why pointer type should be changed?

The initmode is used to pass video timing to the fb_videomode_to_var()
function. Each window setup data in platform data included video
timing information. Since, the video timing data is now moved out of
per-window data, the xres and yres values have to be setup based on
the window for which the fb_videomode_to_var() is called. Hence, the
common timing values is first copied into initmode and then the window
specific xres and yres are set. If initmode is a maintained as a
pointer (to the video timing data in platform data), then any xres and
yres update to initmode would overwrite the 'constant' video timing.

My point is that variable type 'initmode' can be not changed by using pointer type as follows:

@@ -1243,11 +1243,11 @@ static int __devinit s3c_fb_probe_win(struct s3c_fb *sfb, unsigned int win_no,

       }

       windata = sfb->pdata->win[win_no];
-       initmode = &windata->win_mode;
+       initmode = sfb->pdata->vtiming;

       WARN_ON(windata->max_bpp = 0);
-       WARN_ON(windata->win_mode.xres = 0);
-       WARN_ON(windata->win_mode.yres = 0);
+       WARN_ON(windata->xres = 0);
+       WARN_ON(windata->yres = 0);

       win = fbinfo->par;
       *res = win;

@@ -1286,6 +1286,8 @@ static int __devinit s3c_fb_probe_win(struct s3c_fb *sfb, unsigned int win_no,

       }

       /* setup the initial video mode from the window */
+       initmode->xres = windata->xres;
+       initmode->yres = windata->yres;