Syslinux - Aug 2015 - [PATCH] EFI booting may hang indefinitely when no keyboard is present

EFI booting may hang indefinitely when no keyboard is present on a 
machine (or when UEFI has chosen to not init USB due to fast boot). 
Reproducible at least with "EFI v2.31 by American Megatrends" on an 
Asrock H81M-ITX board.

Reason is the input draining loop near the end in efi/main.c; it calls 
ReadKeyStroke() as long as EFI_NOT_READY is not returned. On machines 
without keyboard, EFI_DEVICE_ERROR may be returned instead, causing 
infinite looping. Positively checking for EFI_SUCCESS keeps the intent 
while avoiding this behavior.

Patch is against 6.03; the issue is still present in 6.04 as per 
e466d2498604c8eea055a8e98284d65311073b39

Best Regards,
Olli

--- syslinux-6.03/efi/main.c.orig       2015-08-15 12:31:33.987766223 +0200
+++ syslinux-6.03/efi/main.c    2015-08-15 12:31:44.427993051 +0200
@@ -1324,7 +1324,7 @@ EFI_STATUS efi_main(EFI_HANDLE image, EF
         in = ST->ConIn;
         do {
                 status = uefi_call_wrapper(in->ReadKeyStroke, 2, in,
&key);
-       } while (status != EFI_NOT_READY);
+       } while (status == EFI_SUCCESS);

         if (!setjmp(load_error_buf))
                 load_env32(NULL);

> EFI booting may hang indefinitely when no keyboard is present on a 
> machine (or when UEFI has chosen to not init USB due to fast boot). 
> Reproducible at least with "EFI v2.31 by American Megatrends" on
an
> Asrock H81M-ITX board.
> 
> Reason is the input draining loop near the end in efi/main.c; it calls 
> ReadKeyStroke() as long as EFI_NOT_READY is not returned. On machines 
> without keyboard, EFI_DEVICE_ERROR may be returned instead, causing 
> infinite looping. Positively checking for EFI_SUCCESS keeps the intent 
> while avoiding this behavior.
> 
> Patch is against 6.03; the issue is still present in 6.04 as per 
> e466d2498604c8eea055a8e98284d65311073b39
> 
> Best Regards,
> Olli
 
Thank you for sharing patch.

I assume this patch was tested on systems with no keyboard connected.

I wonder whether it was (also) tested on systems that do have a 
keyboard connected. Among such systems, we could mention different 
types of connections (not just USB), and even those with a KVM switch 
(not to be confused with Kernel-based Virtual Machine).

In addition to being able to boot Syslinux in all those systems, I also 
wonder whether this patch could affect any of the special keys and 
their related directives' behavior (and if so, hopefully for good?). I 
am referring mainly to:
 http://www.syslinux.org/wiki/index.php/Directives/special_keys 

TIA,
Ady.

On 15.08.2015 14:31, Ady via Syslinux wrote:
>   
> Thank you for sharing patch.
>
> I assume this patch was tested on systems with no keyboard connected.
>
> I wonder whether it was (also) tested on systems that do have a
> keyboard connected. Among such systems, we could mention different
> types of connections (not just USB), and even those with a KVM switch
> (not to be confused with Kernel-based Virtual Machine).
The patch was tested with and without keyboard.

Please note that it doesn't touch any of the actual input handling; it 
solely modifies the input-discarding-loop immediately before jumping 
into the kernel.

Best Regards,
Olli

On Sat, Aug 15, 2015 at 7:03 AM, Oliver Wagner via Syslinux
<syslinux at zytor.com> wrote:
> EFI booting may hang indefinitely when no keyboard is present on a machine
> (or when UEFI has chosen to not init USB due to fast boot). Reproducible at
> least with "EFI v2.31 by American Megatrends" on an Asrock
H81M-ITX board.
>
> Reason is the input draining loop near the end in efi/main.c; it calls
> ReadKeyStroke() as long as EFI_NOT_READY is not returned. On machines
> without keyboard, EFI_DEVICE_ERROR may be returned instead, causing
infinite
> looping. Positively checking for EFI_SUCCESS keeps the intent while
avoiding
> this behavior.
>
> Patch is against 6.03; the issue is still present in 6.04 as per
> e466d2498604c8eea055a8e98284d65311073b39
>
> Best Regards,
> Olli
>
> --- syslinux-6.03/efi/main.c.orig       2015-08-15 12:31:33.987766223 +0200
> +++ syslinux-6.03/efi/main.c    2015-08-15 12:31:44.427993051 +0200
> @@ -1324,7 +1324,7 @@ EFI_STATUS efi_main(EFI_HANDLE image, EF
>         in = ST->ConIn;
>         do {
>                 status = uefi_call_wrapper(in->ReadKeyStroke, 2, in,
&key);
> -       } while (status != EFI_NOT_READY);
> +       } while (status == EFI_SUCCESS);
>
>         if (!setjmp(load_error_buf))
>                 load_env32(NULL);
Applied.  Thanks!

-- 
-Gene