Message-ID: <86a3e271-86b6-49f5-aaab-3cb35ec2eb7e@suse.de>
Date:   Mon, 13 Nov 2023 09:39:07 +0100
From:   Thomas Zimmermann <tzimmermann@...e.de>
To:     Javier Martinez Canillas <javierm@...hat.com>,
        Andrew Worsley <amworsley@...il.com>
Cc:     open list <linux-kernel@...r.kernel.org>,
        "open list:DRM DRIVER FOR FIRMWARE FRAMEBUFFERS" 
        <dri-devel@...ts.freedesktop.org>,
        Maxime Ripard <mripard@...nel.org>
Subject: Re: [PATCH] Fix failure of simpledrm probe when trying to grab FB
 from the EFI-based Framebuffer

Hi Javier

Am 12.11.23 um 11:35 schrieb Javier Martinez Canillas:
> Andrew Worsley <amworsley@...il.com> writes:
> 
> Hello Andrew,
> 
>> On Sat, 11 Nov 2023 at 20:10, Javier Martinez Canillas
>> <javierm@...hat.com> wrote:
>>>
>> ....
>>>> On Sat, 11 Nov 2023 at 15:30, Andrew Worsley <amworsley@...il.com> wrote:
>>>>>
>>>>>     The simpledrm.c does not call aperture_remove_conflicting_devices() in it's probe
>>>>>     function as the drivers/video/aperture.c documentation says it should. Consequently
>>>>>     it's request for the FB memory fails.
>>>>>
>>>
>>> The current behaviour is correct since aperture_remove_conflicting_devices()
>>> is for native drivers to remove simple framebuffer devices such as simpledrm,
>>> simplefb, efifb, etc.
>>
>> The efifb is the driver that has "grabbed" the FB earlier
>>
>> Here's a debug output with a dump_stack() call in the devm_aperture_acquire()
>> % grep --color -A14 -B4 devm_aperture_acquire ~/dmesg2.txt
>> [    0.055752] efifb: framebuffer at 0xbd58dc000, using 16000k, total 16000k
>> [    0.055755] efifb: mode is 2560x1600x32, linelength=10240, pages=1
>> [    0.055758] efifb: scrolling: redraw
>> [    0.055759] efifb: Truecolor: size=2:10:10:10, shift=30:20:10:0
>> [    0.055771] devm_aperture_acquire: dump stack for debugging
>> [    0.055775] CPU: 2 PID: 1 Comm: swapper/0 Tainted: G S
> 
> I see. This is the problem then. Your platform then is using a Device Tree
> that contains a "simple-framebuffer" node but also doing a UEFI boot and
> providing an EFI GOP table that is picked by the Linux EFI stub on boot.
> 
> [...]
> 
>>>
>>>>> ...
>>>>> [    3.085302] simple-framebuffer bd58dc000.framebuffer: [drm] *ERROR* could not acquire memory range [??? 0xffff6e1d8629d580-0x2a5000001a7 flags 0x0]: -16
>>>>> [    3.086433] simple-framebuffer: probe of bd58dc000.framebuffer failed with error -16
>>>>> ...
>>>>>
>>>
>>> This is -EBUSY. What is your kernel configuration? Can you share it please.
>>
>> Attached - hope that's acceptable...
>>
>>
> 
> Thanks a lot for providing this. It was very helpful to understand the issue.
> 
> [...]
> 
>>>
>>> I would rather try to understand what is going on in your setup and why
>>> the acquire is returning -EBUSY.
>>>
>>
>> Ok - thanks - let me know where to go from here.
>>
> 
> I think that what we should do instead is to prevent both the EFI GOP and
> "simple-framebuffer" to provide a system framebuffer information and the
> kernel to register two devices (a "simple-framebuffer" by the OF core and
> an "efi-framebuffer" by the sysfb infrastructure).
> 
> In my opinion, the DTB is the best source of truth on an DT platform and
> so is the sysfb that should be disabled if there's a "simple-framebuffer"
> DT node found.
> 
> Can you please try the following (untested) patch?
> 
>  From 7bf4a7917962c24c9f15aaf6e798db9d652c6806 Mon Sep 17 00:00:00 2001
> From: Javier Martinez Canillas <javierm@...hat.com>
> Date: Sun, 12 Nov 2023 11:06:22 +0100
> Subject: [PATCH] of/platform: Disable sysfb if a simple-framebuffer node is
>   found
> 
> Some DT platforms use EFI to boot and in this case the EFI Boot Services
> may register a EFI_GRAPHICS_OUTPUT_PROTOCOL handle, that will later be
> queried by the Linux EFI stub to fill the global struct screen_info data.
> 
> The data is used by the Generic System Framebuffers (sysfb) framework to
> add a platform device with platform data about the system framebuffer.
> 
> But if there is a "simple-framebuffer" node in the DT, the OF core will
> also do the same and add another device for the system framebuffer.
> 
> This could lead for example, to two platform devices ("simple-framebuffer"
> and "efi-framebuffer") to be added and matched with their corresponding
> drivers. So both efifb and simpledrm will be probed, leading to following:
> 
> [    0.055752] efifb: framebuffer at 0xbd58dc000, using 16000k, total 16000k
> [    0.055755] efifb: mode is 2560x1600x32, linelength=10240, pages=1
> [    0.055758] efifb: scrolling: redraw
> [    0.055759] efifb: Truecolor: size=2:10:10:10, shift=30:20:10:0
> ...
> [    3.295896] simple-framebuffer bd58dc000.framebuffer: [drm] *ERROR*
> could not acquire memory range [??? 0xffff79f30a29ee40-0x2a5000001a7
> flags 0x0]: -16
> [    3.298018] simple-framebuffer: probe of bd58dc000.framebuffer
> failed with error -16
> 
> To prevent the issue, make the OF core to disable sysfb if there is a node
> with a "simple-framebuffer" compatible. That way only this device will be
> registered and sysfb would not attempt to register another one using the
> screen_info data even if this has been filled.
> 
> Reported-by: Andrew Worsley <amworsley@...il.com>
> Signed-off-by: Javier Martinez Canillas <javierm@...hat.com>

Reviewed-by: Thomas Zimmermann <tzimmermann@...e.de>

It's a known problem and a know workaround. I don't find it particularly 
elegant, but have no other solution either. It would be much nicer to 
have a way for the sysfb code to detect whether it should create a 
device for the framebuffer.

Best regards
Thomas

> ---
>   drivers/of/platform.c | 17 +++++++++++++++--
>   1 file changed, 15 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/of/platform.c b/drivers/of/platform.c
> index f235ab55b91e..a9fd91e6a6df 100644
> --- a/drivers/of/platform.c
> +++ b/drivers/of/platform.c
> @@ -621,8 +621,21 @@ static int __init of_platform_default_populate_init(void)
>   		}
>   
>   		node = of_get_compatible_child(of_chosen, "simple-framebuffer");
> -		of_platform_device_create(node, NULL, NULL);
> -		of_node_put(node);
> +		if (node) {
> +			/*
> +			 * Since a "simple-framebuffer" device is already added
> +			 * here, disable the Generic System Framebuffers (sysfb)
> +			 * to prevent it from registering another device for the
> +			 * system framebuffer later (e.g: using the screen_info
> +			 * data that may had been filled as well).
> +			 *
> +			 * This can happen for example on DT systems that do EFI
> +			 * booting and may provide a GOP table to the EFI stub.
> +			 */
> +			sysfb_disable();
> +			of_platform_device_create(node, NULL, NULL);
> +			of_node_put(node);
> +		}
>   
>   		/* Populate everything else. */
>   		of_platform_default_populate(NULL, NULL, NULL);

-- 
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Frankenstrasse 146, 90461 Nuernberg, Germany
GF: Ivo Totev, Andrew Myers, Andrew McDonald, Boudien Moerman
HRB 36809 (AG Nuernberg)

Download attachment "OpenPGP_signature.asc" of type "application/pgp-signature" (841 bytes)

Powered by blists - more mailing lists