Re: [PATCH] x86/paravirt: Fix build PARAVIRT_XXL=y without XEN_PV

[Juergen Gross <jgross@suse.com>]

[-- Attachment #1.1.1: Type: text/plain, Size: 3285 bytes --]

On 19.11.21 11:20, Kirill A. Shutemov wrote:
> On Fri, Nov 19, 2021 at 03:51:44PM +0800, Lai Jiangshan wrote:
>> On Thu, Nov 18, 2021 at 7:14 AM Kirill A. Shutemov <kirill@shutemov.name> wrote:
>>>
>>> TDX is going to use CONFIG_PARAVIRT_XXL, but kernel fails to compile if
>>> XEN_PV is not enabled:
>>>
>>>          ld.lld: error: undefined symbol: xen_iret
>>>
>>> It happens because INTERRUPT_RETURN defined to use xen_iret if
>>> CONFIG_PARAVIRT_XXL enabled regardless of CONFIG_XEN_PV.
>>>
>>> The issue is not visible in the current kernel because CONFIG_XEN_PV is
>>> the only user of CONFIG_PARAVIRT_XXL and there's no way to enable them
>>> separately.
>>>
>>> Rework code to define INTERRUPT_RETURN based on CONFIG_XEN_PV, not
>>> CONFIG_PARAVIRT_XXL.
>>>
>>> Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
>>> Cc: Juergen Gross <jgross@suse.com>
>>> Cc: Deep Shah <sdeep@vmware.com>
>>> Cc: "VMware, Inc." <pv-drivers@vmware.com>
>>> ---
>>>   arch/x86/include/asm/irqflags.h | 7 +++++--
>>>   arch/x86/include/asm/paravirt.h | 5 -----
>>>   2 files changed, 5 insertions(+), 7 deletions(-)
>>>
>>> diff --git a/arch/x86/include/asm/irqflags.h b/arch/x86/include/asm/irqflags.h
>>> index b794b6da3214..3b8ddcb7be76 100644
>>> --- a/arch/x86/include/asm/irqflags.h
>>> +++ b/arch/x86/include/asm/irqflags.h
>>> @@ -118,8 +118,6 @@ static __always_inline unsigned long arch_local_irq_save(void)
>>>   #define SAVE_FLAGS             pushfq; popq %rax
>>>   #endif
>>>
>>> -#define INTERRUPT_RETURN       jmp native_iret
>>> -
>>>   #endif
>>>
>>>   #endif /* __ASSEMBLY__ */
>>> @@ -147,8 +145,13 @@ static __always_inline void arch_local_irq_restore(unsigned long flags)
>>>   #ifdef CONFIG_X86_64
>>>   #ifdef CONFIG_XEN_PV
>>>   #define SWAPGS ALTERNATIVE "swapgs", "", X86_FEATURE_XENPV
>>> +#define INTERRUPT_RETURN                                               \
>>> +       ANNOTATE_RETPOLINE_SAFE;                                        \
>>> +       ALTERNATIVE_TERNARY("jmp *paravirt_iret(%rip);",                \
>>> +               X86_FEATURE_XENPV, "jmp xen_iret;", "jmp native_iret;")
>>
>> It is part of what CONFIG_PARAVIRT_XXL was designed for to enable
>> pv-aware INTERRUPT_RETURN.
> 
> That's very vague statement.
> 
> Could you elaborate on what is wrong with proposed fix?
> 
>> I would prefer xen_iret is defined as a weak symbol unconditionally.
>> Like:
>>
>> diff --git a/arch/x86/entry/entry_64.S b/arch/x86/entry/entry_64.S
>> index e38a4cf795d9..c0953f1b4559 100644
>> --- a/arch/x86/entry/entry_64.S
>> +++ b/arch/x86/entry/entry_64.S
>> @@ -635,6 +635,7 @@ SYM_INNER_LABEL_ALIGN(native_iret, SYM_L_GLOBAL)
>>          jnz     native_irq_return_ldt
>>   #endif
>>
>> +SYM_INNER_LABEL(xen_iret, SYM_L_WEAK) /* placeholder */
>>   SYM_INNER_LABEL(native_irq_return_iret, SYM_L_GLOBAL)
>>          /*
>>           * This may fault.  Non-paranoid faults on return to userspace are
>>
>> It will work when !CONFIG_XEN_PV
> 
> It pollutes namespace for no particular reason. I don't see it justified.
> 

I agree. IMO weak symbols should be avoided if easily possible, and the
patch here shows that it IS possible without any problem.


Juergen

[-- Attachment #1.1.2: OpenPGP public key --]
[-- Type: application/pgp-keys, Size: 3135 bytes --]

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 495 bytes --]