Re: [PATCH] CorebootPayloadPkg: Use correct BytesPerScanLine

["You, Benjamin" <benjamin.you@intel.com>]

Hi Arthur,

> -----Original Message-----
> From: Arthur Heymans [mailto:arthur@aheymans.xyz]
> Sent: Thursday, January 25, 2018 5:03 PM
> To: You, Benjamin <benjamin.you@intel.com>
> Cc: edk2-devel@lists.01.org
> Subject: Re: [edk2] [PATCH] CorebootPayloadPkg: Use correct BytesPerScanLine
> 
> "You, Benjamin" <benjamin.you@intel.com> writes:
> 
> > Hi Arthur,
> >
> > Could you please give more details about your case that
> > HorizontalResolution * (BitsPerPixel / 8) and pFbInfo->BytesPerScanLine
> > don't match?
> >
> 
> On many devices, notably Intel hardware, the STRIDE needs to be 64 byte
> aligned when used in linear memory mode, which coreboot does. STRIDE is
> value that used to determine the line to line increment for the
> display. So what coreboot does when initializing the hardware to align
> (HorizontalResolution * (BitsPerPixel / 8)), 64 bytes up.
> 
> > I did a brief search in Coreboot source and found following comments in
> > coreboot-4.6\src\lib\edid.c:
> >
> >   /* In the case of (e.g.) 24 framebuffer bits per pixel, the convention
> >    * nowadays seems to be to round it up to the nearest reasonable
> >    * boundary, because otherwise the byte-packing is hideous.
> >
> > So it appears framebuffer BitsPerPixel will likely be 16 or 32, and the
> > following statement in the same file calculates:
> >
> >   edid->x_resolution = edid->bytes_per_line / (fb_bpp / 8);
> >
> > which results in bytes_per_line (already rounded up to be 32 or 64 byte
> > aligned) matching x_resolution * (fb_bpp / 8).
> >
> > There are cases that even if panel bits_per_pixel is 24, the framebuffer
> > bits_per_pixel is still 32, as shown in
> > coreboot-4.6\src\drivers\emulation\qemu\bochs.c:
> >
> >   edid.panel_bits_per_pixel = 24;
> >   edid_set_framebuffer_bits_per_pixel(&edid, 32, 0);
> >
> > It would be good if you could help with more details on how the mismatch
> > happened in your case as I may have missed something.
> >
> 
> So long story short 'bytes_per_line' reflects how the actual hardware is
> set up, while using '(HorizontalResolution * (BitsPerPixel / 8)' is a
> guess which is only sometimes correct.
> 
> To give an example (on which this was actually a problem):
> let's say we have a display 1366 pixel horizontal resolution with 32
> bits per pixel.
> 
> HorizontalResolution * BitsPerPixel / 8 = 5464
> 
> But this value is not 64 byte aligned which the hardware expects so.
> 
> aligned value = ((HorizontalResolution * (BitsPerPixel / 8) + 63) & ~63
> = 5504
>

I noticed in coreboot-4.7\src\include\edid.h, there are following comments:

/* 3 variables needed for coreboot framebuffer.
	 * In most cases, they are the same as the ha
	 * and va variables, but not always, as in the
	 * case of a 1366 wide display.
	 */
	u32 x_resolution;
	u32 y_resolution;
	u32 bytes_per_line;

And in coreboot-4.7\src\lib\edid.c:

	edid->bytes_per_line = ALIGN_UP(edid->mode.ha *
		div_round_up(fb_bpp, 8), row_byte_alignment);
	edid->x_resolution = edid->bytes_per_line / (fb_bpp / 8);

Above calculations derive x_resolution from the roundup value of 
bytes_per_line. In case of 1366 display, it would produce a x_resolution of
1376, which is larger than 1366 but satisfies the equation of 
bytes_per_line == (HorizontalResolution * (BitsPerPixel / 8)

It appears this is what Coreboot produces right now. Not sure if there are 
other cases leading to Coreboot producing framebuffer parameters NOT 
satisfying the above equation.

BTW, do you think the above calculation of x_resolution hides the 
information of display and should be fixed?

> So tianocore should use the value coreboot provides it instead of trying
> to compute/guess it.
> 
> > Thanks,
> >
> > - ben
> >
> 
> I hope this clarifies it.
> 
> Arthur
> 
> >> -----Original Message-----
> >> From: edk2-devel [mailto:edk2-devel-bounces@lists.01.org] On Behalf Of
> >> arthur@aheymans.xyz
> >> Sent: Wednesday, January 24, 2018 6:58 PM
> >> To: edk2-devel@lists.01.org
> >> Cc: Arthur Heymans <arthur@aheymans.xyz>
> >> Subject: [edk2] [PATCH] CorebootPayloadPkg: Use correct BytesPerScanLine
> >>
> >> From: Arthur Heymans <arthur@aheymans.xyz>
> >>
> >> Fetch BytesPerScanLine from coreboot table to reflect how the actual
> >> framebuffer is set up instead of guessing it from the horizontal
> >> resolution.
> >>
> >> This fixes a garbled display when HorizontalResolution * (BitsPerPixel
> >> / 8) and pFbInfo->BytesPerScanLine don't match.
> >>
> >> Contributed-under: TianoCore Contribution Agreement 1.1
> >> Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
> >>
> >> diff --git a/CorebootPayloadPkg/FbGop/FbGop.c
> >> b/CorebootPayloadPkg/FbGop/FbGop.c
> >> index 37d6def7f7..6790617033 100644
> >> --- a/CorebootPayloadPkg/FbGop/FbGop.c
> >> +++ b/CorebootPayloadPkg/FbGop/FbGop.c
> >> @@ -822,7 +822,7 @@ FbGopCheckForVbe (
> >>    BitsPerPixel         = pFbInfo->BitsPerPixel;
> >>    HorizontalResolution = pFbInfo->HorizontalResolution;
> >>    VerticalResolution   = pFbInfo->VerticalResolution;
> >> -  BytesPerScanLine     = HorizontalResolution * (BitsPerPixel / 8);
> >> +  BytesPerScanLine     = pFbInfo->BytesPerScanLine;
> >>
> >>    ModeBuffer = (FB_VIDEO_MODE_DATA *) AllocatePool (
> >>
> >>
> >> 				ModeNumber * sizeof
> >> (FB_VIDEO_MODE_DATA)
> >> --
> >> 2.16.1
> >>
> >> _______________________________________________
> >> edk2-devel mailing list
> >> edk2-devel@lists.01.org
> >> https://lists.01.org/mailman/listinfo/edk2-devel
> <#secure method=pgpmime mode=sign>
> 
> --
> Arthur Heymans

Thanks,

- ben