Re: [Xen-devel] [PATCH for-4.5] EFI: Ignore EFI commandline, skip console setup when booted from GRUB

[Roy Franz <roy.franz@xxxxxxxxxx>]

On Tue, Nov 4, 2014 at 1:17 PM, Daniel Kiper <daniel.kiper@xxxxxxxxxx> wrote:
> On Tue, Nov 04, 2014 at 10:39:53AM -0800, Roy Franz wrote:
>> On Mon, Nov 3, 2014 at 11:49 PM, Jan Beulich <JBeulich@xxxxxxxx> wrote:
>> >>>> On 04.11.14 at 05:31, <roy.franz@xxxxxxxxxx> wrote:
>> >> On Mon, Nov 3, 2014 at 1:33 AM, Jan Beulich <JBeulich@xxxxxxxx> wrote:
>> >>>>>> On 03.11.14 at 07:13, <roy.franz@xxxxxxxxxx> wrote:
>> >>>> This patch implements what I understand to be the desired behavior when
>> >>>> booting
>> >>>> an EFI Xen image via GRUB based on the thread last week.  The EFI 
>> >>>> command
>> >>>> line
>> >>>> is not used, and the Xen commandline comes via the multiboot protocol 
>> >>>> (and
>> >>>> in the ARM case the multiboot FDT bindings).  This brings the x86 and 
>> >>>> arm64
>> >>>> GRUB EFI boot cases into alignment in not using the EFI commandline.
>> >>>
>> >>> Right, but ...
>> >>>
>> >>>> The current state of the arm64 code takes the Xen commandline from the 
>> >>>> FDT,
>> >>>> but still looks in the EFI commandline for EFI boot specific options.  
>> >>>> If
>> >>>> unexpected options are passed in the EFI commandline, it will generate
>> >>>> "unrecognized option" ouput for all unexpected options.
>> >>>
>> >>> ... why is this?
>> >>
>> >> The EFI boot code did this before any of the arm64 changes, and that
>> >> behavior is unchanged.
>> >> The actual message is "WARNING: Unknown command line option"
>> >>
>> >> I was simply trying to explain the current behavior regarding the EFI
>> >> commandline.
>> >
>> > Argh, I should have stripped that second sentence from the quoted
>> > context, as the question was regarding the apparently special ARM64
>> > behavior you describe.
>>
>> >>>> The current state of the arm64 code takes the Xen commandline from the 
>> >>>> FDT,
>> >>>> but still looks in the EFI commandline for EFI boot specific options.
>> OK, I'll try address the above.  This behavior is the "no config file
>> case", which only
>> exists on arm64 - x86 always uses a config file when booted as an EFI
>> application, since
>> for x86 GRUB does execute Xen as en EFI application when using multiboot2.
>
> I have a feeling that you wrote that by mistake. As I said in my earlier 
> emails
> multiboot2 on x86 EFI platform will not load EFI application as an EFI 
> application.

Grr..  meant "does not execute"  of course.  I do get that now :)
>
>> The case of being executed as an EFI application, loaded by GRUB and provided
>> module information via the multiboot2 (in FDT) protocol is unique to
>> arm64.  All the code
>> is common code, but since efi_arch_use_config_file() is always true
>> for x86 the case
>> described only applies to arm64.
>
> Right now it is true. Probably it will change after introducing
> multiboot2 protocol on x86.

So you expect to be using the efi_start() function mostly as is?

>
> Daniel

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel