Re: [PATCH] arm64: xen_boot: Fix xen boot using Grub on AARCH64

[Andrei Borzenkov]

23.02.2016 17:16, Fu Wei пишет:
> Hi Ian
> 
> On 23 February 2016 at 18:00, Ian Campbell <address@hidden> wrote:
>> On Mon, 2016-02-22 at 17:29 +0800, Fu Wei wrote:
>>> Hi Julien,
>>>
>>> On 20 February 2016 at 00:28, Julien Grall <address@hidden>
>>> wrote:
>>>> Xen is currently crashing because of malformed compatible property for
>>>> the boot module. This is because the property string is not
>>>> null-terminated as requested by the ePAR spec.
>>>> ---
>>>>  grub-core/loader/arm64/xen_boot.c | 2 +-
>>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>>
>>>> diff --git a/grub-core/loader/arm64/xen_boot.c b/grub-
>>>> core/loader/arm64/xen_boot.c
>>>> index a914eb8..8ae43d7 100644
>>>> --- a/grub-core/loader/arm64/xen_boot.c
>>>> +++ b/grub-core/loader/arm64/xen_boot.c
>>>> @@ -156,7 +156,7 @@ prepare_xen_module_params (struct xen_boot_binary
>>>> *module, void *xen_boot_fdt)
>>>>        grub_fdt_add_subnode (xen_boot_fdt, chosen_node, module_name);
>>>>
>>>>    retval = grub_fdt_set_prop (xen_boot_fdt, module_node, "compatible",
>>>> -                             MODULE_CUSTOM_COMPATIBLE,
>>>> sizeof(MODULE_CUSTOM_COMPATIBLE) - 1);
>>>> +                             MODULE_CUSTOM_COMPATIBLE,
>>>> sizeof(MODULE_CUSTOM_COMPATIBLE));
>>>>    if (retval)
>>>>      return grub_error (GRUB_ERR_IO, "failed to update FDT");
>>>>
>>>> --
>>>> 1.9.1
>>>
>>> I have tested it. yes, xen will crash without this patch.
>>> Tested-by: Fu Wei <address@hidden>
>>>
>>> BTW, since Vladimir has simplified the xen boot code , and Now Xen
>>> distinguishes modules by order.
>>> ------
>>> menuentry 'Foundation Model Xen test with initramfs' {
>>>     xen_hypervisor /xen -- no-bootscrub console=dtuart conswitch=x
>>> dtuart=serial0 dom0_mem=512M dom0_max_vcpus=2
>>>     xen_module /dom0_kernel_Image console=hvc0 root=/dev/ram0 ro
>>>     xen_module /dom0_initramfs.cpio
>>>     xen_module /xsm
>>>     devicetree /fvp-base-gicv2-psci.dtb
>>> }
>>> -----
>>>
>>> Maybe we should update doc to  coordinate with this change.
>>>
>>> And another problem is: How does XEN identify XSM?
>>> Test with the config file above, I got "(XEN) MODULE[3]:
>>> 00000008fabb4000 - 00000008fabb65e7 Unknown"
>>
>> I must confess when Vlad proposed this change I thought Xen was able to
>> automatically probe for whether a module was an XSM module. Perhaps it is
>> either broken or that functionality never existed to start with?
>>
> 
> From the test result, it seems that the latest xen can NOT
> automatically probe XSM module, even we load xsm as the third module.
> 
> the test branch is master branch. Please correct me if I tested it in wrong 
> way.
> 

That's exactly what code looks like. I.e. XSM is recognized only if
bootloader provided xen,xsm-policy compatible property.

OTOH looking at xsm_dt_policy_init(), XSM module has defined magic and
multiboot version actually detects module type based on this magic. So I
think it makes sense to extend process_multiboot_node() to check for XSM
magic in case of unknown modules. This will make ARM behavior compatible
with x86. Otherwise we simply won't be able to load XSM on ARM at all,
given current implementation in GRUB.