[PATCH 2/2] efi: earlycon: Add support for generic framebuffers and move to fbdev subsystem

[Markuss Broks]

Hi Andy,

On 7/29/22 00:19, Andy Shevchenko wrote:
> On Thu, Jul 28, 2022 at 4:32 PM Markuss Broks <markuss.broks at gmail.com> wrote:
>>
>> Add early console support for generic linear framebuffer devices.
>> This driver supports probing from cmdline early parameters
>> or from the device-tree using information in simple-framebuffer node.
>> The EFI functionality should be retained in whole.
>> The driver was disabled on ARM because of a bug in early_ioremap
>> implementation on ARM.
> 
> ...
> 
>> -               efifb,[options]
>> +               efifb
>>                          Start an early, unaccelerated console on the EFI
>> -                       memory mapped framebuffer (if available). On cache
>> -                       coherent non-x86 systems that use system memory for
>> -                       the framebuffer, pass the 'ram' option so that it is
>> -                       mapped with the correct attributes.
>> +                       memory mapped framebuffer (if available).
> 
> If somebody passes those (legacy) options, what will happen?

Those would be ignored. Instead it would be automatically determined if 
framebuffer is located in RAM or in the I/O space.

> 
> ...
> 
>>   config EFI_EARLYCON
>> -       def_bool y
>> -       depends on SERIAL_EARLYCON && !ARM && !IA64
>> -       select FONT_SUPPORT
>> -       select ARCH_USE_MEMREMAP_PROT
>> +       bool "EFI early console support"
>> +       depends on FB_EARLYCON && !IA64
> 
> This doesn't sound right. Previously on my configuration it was
> selected automatically, now I need to select it explicitly? I mean
> that for me EFI_EARLYCON should be selected by default as it was
> before.

The problem I had with EFI_EARLYCON selected by default was that it 
would also carry fbdev with itself. Luckily, that's solved if it's moved 
to console subsystem.

> 
> ...
> 
>> +static int __init simplefb_earlycon_remap_fb(void)
>> +{
>> +       int is_ram;
> 
> + blank line.
> 
>> +       /* bail if there is no bootconsole or it has been disabled already */
>> +       if (!earlycon_console || !(earlycon_console->flags & CON_ENABLED))
>> +               return 0;
>> +
>> +       is_ram = region_intersects(info.phys_base, info.size,
>> +                                  IORESOURCE_SYSTEM_RAM, IORES_DESC_NONE);
>> +       is_ram = is_ram == REGION_INTERSECTS;
> 
> Was it in the original code? Otherwise, I would go with plain conditional:
> 
>    if (region_intersects())
>      base = ...
>    else
>      base = ...

It wasn't in original code. Original code assumed MEMREMAP_WC always
unless "ram" was specified as an option to efifb (e.g.
earlycon=efifb,ram). This code instead checks if the framebuffer is in 
RAM, and sets the mapping accordingly.

Another issue is that region_intersects() returns REGION_INTERSECTS 
(defined as 0) when true. I suppose we could use something like:

if (region_intersects() == REGION_INTERSECTS)
...

> 
>> +       info.virt_base = memremap(info.phys_base, info.size,
>> +                                 is_ram ? MEMREMAP_WB : MEMREMAP_WC);
>> +
>> +       return info.virt_base ? 0 : -ENOMEM;
>> +}
> 
> ...
> 
>> +static void simplefb_earlycon_write_char(u8 *dst, unsigned char c, unsigned int h)
>> +{
>> +       const u8 *src;
>> +       int m, n, bytes;
>> +       u8 x;
>> +
>> +       bytes = BITS_TO_BYTES(font->width);
>> +       src = font->data + c * font->height * bytes + h * bytes;
>> +
>> +       for (m = 0; m < font->width; m++) {
>> +               n = m % 8;
>> +               x = *(src + m / 8);
>> +               if ((x >> (7 - n)) & 1)
>> +                       memset(dst, 0xff, (info.depth / 8));
>> +               else
>> +                       memset(dst, 0, (info.depth / 8));
>> +               dst += (info.depth / 8);
>> +       }
>> +}
> 
> Wondering if we already have something like this in DRM/fbdev and can
> split into a generic helper.
> 
> ...
> 
>> +       ret = sscanf(device->options, "%u,%u,%u", &info.x, &info.y, &info.depth);
>> +       if (ret != 3)
>> +               return -ENODEV;
> 
> Don't we have some standard template of this, something like XxYxD,
> where X, Y, and D are respective decimal numbers?

I'm not aware of this.

> 

-Markuss