On 29.04.22 11:56, Borislav Petkov wrote:
> On Thu, Apr 21, 2022 at 10:10:53PM +0800, Lai Jiangshan wrote:
>> diff --git a/arch/x86/entry/entry_64_compat.S b/arch/x86/entry/entry_64_compat.S
>> index 4fdb007cddbd..c5aeb0819707 100644
>> --- a/arch/x86/entry/entry_64_compat.S
>> +++ b/arch/x86/entry/entry_64_compat.S
>> @@ -50,7 +50,7 @@ SYM_CODE_START(entry_SYSENTER_compat)
>>   	UNWIND_HINT_EMPTY
>>   	ENDBR
>>   	/* Interrupts are off on entry. */
>> -	SWAPGS
>> +	swapgs
>>   
>>   	pushq	%rax
>>   	SWITCH_TO_KERNEL_CR3 scratch_reg=%rax
> 
> I'm not sure about this: why can't XENPV do a 32-bit syscall through the vdso?

The syscall will land in the hypervisor, which then will activate
the related registered callback (xen_sysenter_target).

> Also, looking at this, Jürgen, it looks kinda spaghetti to me:
> 
> entry_SYSENTER_compat
> 
> ...
> 
>          /* XEN PV guests always use IRET path */
>          ALTERNATIVE "testl %eax, %eax; jz swapgs_restore_regs_and_return_to_usermode", \
>                      "jmp swapgs_restore_regs_and_return_to_usermode", X86_FEATURE_XENPV
> 
> 
> then at that swapgs_restore_regs_and_return_to_usermode label:
> 
> #ifdef CONFIG_XEN_PV
>          ALTERNATIVE "", "jmp xenpv_restore_regs_and_return_to_usermode", X86_FEATURE_XENPV
> #endif
> 
> 
> Can we simply jump directly to xenpv_restore_regs_and_return_to_usermode
> from entry_SYSENTER_compat or is that CONFIG_DEBUG_ENTRY chunk there
> needed?

I wouldn't insist on the CONFIG_DEBUG_ENTRY chunk.


Juergen